Esp with thepic16f877

EMBEDDED
SYSTEMS
PROGRAMMING
WITH THE
PIC16F877
Second Edition
By Timothy D. Green
Copyright 2008 by Timothy D. Green
All Rights Reserved.

2
Table of Contents
Preface …………………………………………………………………. 5
List of Figures …………………………………………………………. 6
Abbreviations and Acronyms …………………………………………. 7
Trademarks ……………………………………………………………. 10
Chapter 1 Introduction to ESP and the PIC …………………………. 11
Chapter 2 Microcontrollers and the PIC16F877 ……………………. 15
Section 2.0 Chapter Summary ……………………………….. 15
Section 2.1 Memory and Memory Organization ……………. 15
Section 2.2 The PIC16F877 …………………………………. 16
Section 2.3 Programming the PIC …………………………… 17
Chapter 3 Simple PIC Hardware & Software (“Hello World”) …….. 20
Section 3.1 A Simple Example System ……………………… 20
Section 3.2 Summary of Instructions and Concepts …………. 25
Chapter 4 The PIC Instruction Set (Part I) ………………………….. 27
Section 4.0 Chapter Summary ………………………………. 27
Section 4.1 The PIC16F877 Instruction Set ………………… 27
Section 4.2 Summary of Instructions and Concepts …………. 33
Chapter 5 The PIC Instruction Set (Part II) …………………………. 34
Section 5.1 Introduction ……………………………………… 34
Section 5.2 Keypad and Display Interface …………………… 35
Section 5.3 The STATUS Register and Flag Bits ……………. 39
Section 5.4 The Keypad Software ……………………………. 40
Section 5.5 The LED Display Software ……………………… 43
Section 5.6 Improved Display and Indirect Addressing ……… 46
Section 5.7 Odds & Ends …………………………………….. 50
Section 5.8 Using KEY_SCAN and DISPLAY Together ……. 54
Section 5.9 A Last Look at the Advanced Security System ….. 56
Section 5.10 Summary of Instructions and Concepts ………… 57
Chapter 6 Fundamental ESP Techniques ……………………………. 59
Section 6.0 Chapter Summary ………………………………… 59
Section 6.1 Introduction ………………………………………. 59
Section 6.2 Software Readability …………………………… 59
Section 6.3 Software Maintainability …………………………. 60

3
Chapter 6
Section 6.4 Software Fundamentals ………………………….. 60
Section 6.5 The Background Routine ………………………… 61
Section 6.6 The Watch-Dog Timer …………………………… 61
Section 6.7 Event-Driven Software …………………………... 62
Section 6.8 Interrupts …………………………………………. 65
Section 6.9 Slow Inputs and Outputs …………………………. 65
Section 6.10 Software Time Measurement …………………… 66
Section 6.11 Hashing …………………………………………. 67
Section 6.12 Waveform Encoding ……………………………. 68
Section 6.13 Waveform Decoding …………………………… 71
Section 6.14 RAM, ROM, and Time Tradeoffs ……………… 75
Section 6.15 ROM States ……………………………………... 75
Section 6.16 Limitations of C/C++ …………………………… 76
Chapter 7 Advanced ESP ……………………………………………. 78
Section 7.0 Chapter Summary ………………………………… 78
Section 7.1 Introduction ………………………………………. 78
Section 7.2 Sine Wave Generation …………………………….78
Section 7.3 Dual-Tone-Multi-Frequency (DTMF) Signaling …81
Section 7.4 Pulse-Width Modulation ………………………….82
Section 7.5 ADPCM Data Compression ………………………84
Section 7.6 Test Functions and System Ideas …………………87
Chapter 8 PIC Peripherals and Interrupts ……………………………91
Section 8.0 Chapter Summary ………………………………...91
Section 8.1 Overview of the PIC Peripherals …………………91
Section 8.2 Input/Output Ports ………………………………..93
Section 8.2.1 Port A ………………………………………….. 93
Section 8.2.2 Port B, Port C, Port D ………………………….. 95
Section 8.2.3 Port E ………………………………………….. 95
Section 8.3 Interrupts …………………………………………. 95
Section 8.4 ADC and Analog MUX ………………………….. 98
Section 8.5 Watch-Dog Timer ………………………………... 102
Section 8.6 Timer 0 …………………………………………… 103
Section 8.7 Timer 1 ………………………………………….. 105
Section 8.8 Timer 2 ………………………………………….. 106
Section 8.9 Capture Mode …………………………………… 107
Section 8.10 Compare Mode ………………………………… 109
Section 8.11 Pulse-Width Modulation (PWM) ……………… 111
Section 8.12 Parallel Slave Port ……………………………… 114

4
Chapter 8
Section 8.13 EEPROM Data Memory ………………………… 116
Section 8.14 FLASH Program Memory ………………………. 117
Section 8.15 FLASH Code & Data EEPROM Protection ……. 118
Section 8.16 The CONFIGURATION Word ………………… 119
Section 8.17 Sleep Modes & Reset Modes …………………… 120
Chapter 9 PIC Peripherals, Serial Communications Ports ……………122
Section 9.0 Chapter Summary …………………………………122
Section 9.1 Introduction ……………………………………….122
Section 9.2 USART (Overview) ………………………………122
Section 9.2.1 USART (Asynchronous Mode, Full-Duplex) …..123
Section 9.2.2 USART (Synchronous, Master Mode) ………… 127
Section 9.2.3 USART (Synchronous, Slave Mode) ………….. 128
Section 9.3 Serial Peripheral Interface (Master Mode) ………. 128
Section 9.4 Serial Peripheral Interface (Slave Mode) …………132
Section 9.5 I2C System Overview …………………………….134
Section 9.5.1 I2C Slave Mode …………………………………137
Section 9.5.2 I2C Master Mode ………………………………. 139
Chapter 10 DSP Fundamentals ……………………………………… 147
Section 10.0 Chapter Summary ………………………………. 147
Section 10.1 Introduction …………………………………….. 147
Section 10.2 An Example: A Low-Pass Filter ……………….. 148
Section 10.3 An Example: A High-Pass Filter ………………. 148
Section 10.4 DSP Filters in General ………………………… 149
Section 10.5 Aliasing and the Nyquist Sampling Theorem ……149
Section 10.6 DSP Cookbook I: A Simple LPF/HPF …………. 151
Section 10.7 DSP Cookbook II: A Simple BPF ……………… 152
Section 10.8 DSP Cookbook III: A Median Filter …………… 153
Section 10.9 DSP Example I: Standard DTMF Decoding …….154
Section 10.10 DSP Example II: Alternative DTMF Decoding .. 155
Section 10.11 DSP Application: Speech Compression ………… 159
Appendix A The PIC16F877 Instruction Set ………………………….. 163
Appendix B Useful C++ Programs for PIC ASM Applications ………. 180
Appendix C Special Function Registers (RAM Addresses & Bits) ……185
Appendix D PIC16F877 Register File Map ……………………………191
Appendix E PIC16F877 Pin Function Map ……………………………192
Appendix F Save/Restore Registers on Interrupt ……………………… 193
References ………………………………………………………………. 195

5
Preface
This book is intended for use by Junior-level undergraduates, Senior-level
undergraduates, and Graduate students in electrical engineering as well as practicing
electrical engineers and hobbyists and seeks to provide a gentle introduction to embedded
systems programming with the Microchip PIC16F877 microcontroller. After introducing
the PIC16F877 and its programming, this book covers the fundamental techniques and
advanced level techniques of embedded systems programming in a general sense. The
general sense ESP techniques can be applied to any microcontroller. There is also an
introduction to the fundamentals of digital signal processing (DSP) using the PIC16F877.
I would like to thank Dr. Dan Simon of the Cleveland State University Electrical
Engineering Department for his kind and valuable help and suggestions in the
preparations for this book. I would also like to thank John R. Owerko and James R.
Jackson, both of A.R.F Products, Inc., for their expertise in the security systems market.
I owe them both a great debt for my knowledge of security systems and for expanding my
knowledge of the techniques of embedded systems programming in general.
Special thanks also go to Sister Renee Oliver who proofread the manuscript and
offered many helpful suggestions. Thanks go to my friends Damian Poirier, Jim Strieter,
Greg Glazer, Zarif Bastawros, Brian McGeever, Ted Seman, and Jim Chesebrough who
offered many helpful suggestions.
Any errors that remain in the text are mine and I will correct them in the next
edition.
Timothy D. Green
November 2005
Cleveland, Ohio

6
List of Figures
2-1 uP Internal View Block Diagram
2-2 PIC16F877 Internal Block Diagram
3-1 Simple Hardware View (Ports Only)
3-2 Basic Hardware System Example
4-1 A Simple Security System
5-1 Twelve-Key Matrix Keypad
5-2 PIC Matrix Keypad Interface Circuit
5-3 Seven Segment LED Digit Display in Common Cathode and Common
Anode Forms
5-4 Single LED Digit Drives for Common Cathode/Anode Forms
5-5a Multiplexed LED Digit Drives for Common Cathode Form
5-5b Multiplexed LED Digit Drives for Common Anode Form
5-6 Illustration of Indirect RAM Addressing
5-7 Diagram of RLF and RRF Instructions
6-1 “All Digital” Watch-Dog Timer Circuit
6-2 Event-Driven Push-Button Switch DeBouncing
6-3 Manchester Code Waveform
6-4 Decoding of Manchester Waveform
7-1 Gaussian Probability Density Function and a Set of Sampled Values
8-1 ADCON1 “Analog vs Digital” Selection Codes
8-2 PIC16F877 Interrupt Tree
9-1 Master Mode SPI Mode Timing
9-2 Serial-Out/Serial-In with the 74HC164 and 74HC165
9-3 Serial-Out/Serial-In with Gated Clock to Inhibit Serial Out
9-4 SPI Mode Timing (Slave Mode, CKE = 0)
9-5 SPI Mode Timing (Slave Mode, CKE = 1)
10-1 Example of Aliasing When Sampling an Analog Signal
10-2 “Slow-to-Fast” Mode Speech Compression Process
10-3 “Fast-to-Slow” Mode Speech Compression Process
Appendix A Figure: Diagram of RLF and RRF Instructions

7
Abbreviations and Acronyms
ABS = Absolute Value
ACCUM = Accumulator
ADC = Analog-to-Digital Converter
ADPCM = Adaptive Differential Pulse Code Modulation
ALU = Arithmetic Logic Unit
Arccos = Arc-Cosine
ASCII = American Standard Code for Information Interchange
ATM = Automatic Teller Machine
BOR = Brown Out Reset
BPF = Band Pass Filter
CK = Clock
Cos = Cosine
CPU = Central Processing Unit
D = Data
DAC = Digital-to-Analog Converter
dB = Decibels
DC = Direct Current
DDS = Direct Digital Synthesis
DIP = Dual Inline Package
DPSK = Differential Phase Shift Keying
DSP = Digital Signal Processing
DTMF = Dual Tone Multi-Frequency
EEPROM = Electrically Erasable Programmable Read Only Memory
EMC = Electro-Magnetic Compatibility
EMI = Electro-Magnetic Interference
EPROM = Erasable Programmable Read Only Memory
ESP = Embedded Systems Programming
Fmax = Maximum Frequency
Fosc = Oscillator Frequency (of the PIC)
Freq = Frequency

8
FSK = Frequency Shift Keying
HPF = High Pass Filter
Hz = Hertz
IIC or I2C = Inter-Integrated Circuit
INT = Interrupt
I/O = Input/Output
ISR = Interrupt Service Routine
kHz = Kilohertz
LED = Light Emitting Diode
LPF = Low Pass Filter
Max = Maximum
MHz = Megahertz
Min = Minimum
ms = Milliseconds
MSSP = Master Synchronous Serial Port
OSC = Oscillator
PC = Personal Computer or Program Counter
PIC = Peripheral Interface Controller
PISO = Parallel-In, Serial-Out (Shift Register)
PLL = Phase-Locked Loop
POR = Power-On Reset
PROM = Programmable Read Only Memory
PSP = Parallel Slave Port
PWM = Pulse Width Modulation
Q = Flip-Flop, Counter, or Shift Register Output State (Data Out)
RAM = Random Access Memory (A Read/Write Memory)
RC = Resistor/Capacitor (Time Constant or Circuit)
RF = Radio Frequency
RFI = Radio Frequency Interference
ROM = Read Only Memory
Sin or sin = Sine
SIPO = Serial-In, Parallel-Out (Shift Register)
SPI = Serial Peripheral Interface
sqrt = Square Root
SR = Sampling Rate

9
uC = Microcontroller
uP = Microprocessor
USART = Universal Synchronous/Asynchronous Receiver/Transmitter
WDT = Watch-Dog Timer
XTAL = Quartz Crystal (Sets Oscillator Frequncy)

10
Trademarks
IBM and IBM-PC are registered trademarks of International Business Machines,
Inc.
PIC, PIC16F87X, PIC16F877, MPLAB, MPASM, In-Circuit Debugger, In-
Circuit Serial Programmer are registered trademarks of Microchip Technology, Inc.
CTI and CTI Speech Compressor are registered trademarks of Compressed Time,
Inc.
Touch Tone, Unix, C, and C++ are registered trademarks of AT&T, Inc.
Linux is a registered trademark of the Free Software Foundation.
I2C, IIC, and Inter-Integrated Circuit are registered trademarks of Philips Corp.
The PIC Instruction Set, Assembly Keywords, and Mnemonics are Copyrighted
by Microchip Technology, Inc.
Maxim and MAX690CPA are registered trademarks of Maxim Corporation.
Borland, Turbo, Turbo C++ are registered trademarks of Inprise, Inc.

11
Chapter 1: Introduction to ESP and the PIC
An embedded system is a product which uses a computer to run it but the product,
itself, is not a computer. This is a very broad and very general definition. Embedded
systems programming, therefore, consists of building the software control system of a
computer-based product. ESP encompasses much more than traditional programming
techniques since it actually controls hardware in advance of real time. ESP systems often
have limitations on memory, speed, and peripheral hardware. The goals of ESP
programmers are to get the “maximum function and features in the minimum of space
and in minimum time”.
Embedded systems are everywhere! Name almost any appliance in your home or
office and it may have a microprocessor or a microcomputer to run it. A watch,
microwave oven, telephone, answering machine, washer, dryer, calculator, toy, robot, test
equipment, medical equipment, traffic light, automobile computer, VCR, CD player,
DVD player, TV, radio, and printer all have computers in them to run them.
These examples of embedded systems are simple but the concept of embedded
systems applies to much larger systems as well. Overall, there are four levels of size,
option, and complexity in embedded systems. These levels are:
1) High Level
2) Medium Level
3) Low level with hardware
4) Low level without hardware
A good example of a high level embedded system is an air-traffic control system.
It would use a main-frame computer with many terminals and many users on a time-
sharing basis. It would connect to several smaller computers, run the radar, receive
telemetry, get weather information, have extensive communications sub-systems, and
coordinate all of these function in an orderly, systematic way. It is necessarily a high-
reliability system and may, therefore, have extensive backup systems. It would have a
custom-built operating system that would be completely dedicated to controlling air-
traffic.
An example of a medium level embedded system is a typical automatic teller
machine (ATM) at any bank or bank terminal. It may use a more advanced
microprocessor with many peripheral functions. Consider that it contains a video
terminal, a keyboard, a card-reader, a printer, the money-dispensing unit, a modem, and
many input/output ports. The ATM probably doesn’t use a custom operating system but
would use something off the shelf, like Unix or Linux. The controlling software is
probably written in a high-level language like C or C++.

12
The appliances and other things given on page 11 are all examples of the low-
level-with-hardware embedded systems. They do not use microprocessors but do use
microcontrollers, which are complete computers on a single chip. Microcontrollers have
a CPU, RAM, ROM, and, typically, several peripheral hardware modules which are built-
in and are under software control. The PIC16F877 is such a microcontroller. Any of the
example products and applications on page 11 could be controlled by the PIC. They
could be programmed in C or C++ but care would be needed so as not to use too much
RAM or ROM inadvertently. The process or program also must not need very high speed
operation – it should not be timing-critical. More control, stability, memory
management, and speed can be gained by programming in assembly languages. The
programming at the low-level will interact with the hardware in much finer detail than in
the medium-level or the high-level systems.
The low-level-without-hardware embedded systems are almost identical to the
low-level-with-hardware systems and can run exactly the same products, devices, and
applications. The differences which are present in the low-level-without-hardware
systems are that the microcontroller and the system have an absolute minimum of
hardware peripheral functions. At this level, the software must mimic the desired
hardware peripheral functions. This puts a much greater challenge on the ESP
programmer. (Assembly language is a MUST.)
There are several characteristics in ESP that separate it from traditional
programming techniques. They are as follows:
1) ESP is all about process control and control systems. ESP is what runs a
given product.
2) ESP systems must run in “real-time”. The program must keep pace or stay
ahead of the real world and its timing. For example, a telephone answering
machine may use a complex algorithm to compress, expand, encode, and
decode speech signals. The ESP program must be able to run these processes
as speech is coming in or going out. There must be no delays. A traditional
program would not be sensitive to the requirements of speed that are needed
here.
3) ESP software must run with infinity-loops. If they didn’t, the products could
not run at all! In contrast, infinity-loops are the cardinal sin of traditional
programming.
4) ESP software often uses “event-driven” techniques, especially at the low-
levels. These techniques are highly structured and save operating time.
Traditional programming may also use “event-driven” techniques but it is not
critical.
5) Low-level ESP software systems must sometimes mimic the hardware that the
product needs. There is no parallel to this in traditional programming.
6) Embedded systems usually have far less memory than traditional
programming environments. This eliminates heavy nesting of subroutines and
recursive subroutine calls.

13
7) The arithmetic/logic unit of a microcontroller is much smaller than ones in a
traditional setting, and, consequently, ESP is not as mathematically oriented
as a traditional program.
Embedded systems programming at the low levels is necessarily a multi-disciplinary
field. The programmers and designers must consider hardware issues, manufacturing
issues, electromagnetic interference/electromagnetic compatibility problems, and noise
limitations.
Low-level code doesn’t just consist of algorithms but the code, itself, is, at times,
a great function of the code geometry. Getting optimum performance of low-level
embedded code depends very much on how the instructions are placed in program
memory.
The C/C++ language may be used in low-level embedded systems programming
but not where “fine controls” or “high speeds” are required. The blanket statement that a
“C compiler can produce code that is nearly as good as an assembly language program”
is OK for traditional programs, high-level ESP, and medium-level ESP. It is not true for
low-level ESP! Low-level ESP is special and has very exacting demands on its code.
(This fact will be explained in detail in Chapter 6.)
The overall scope of this book is to show the reader how to program the PIC,
use its peripheral functions, and provide the fundamentals of general ESP techniques.
The plan of this book is as follows:
Chapter 2 introduces microcontrollers in general and details the basic structure of
the PIC16F877.
Chapter 3 introduces the most fundamental elements of programming the PIC in
assembly language using a simple circuit and program. Several instructions are
introduced here. (Chapter 3 is the “Hello World” Chapter.)
Chapter 4 covers the first half of the PIC instruction set by considering the design
of a very simple home security system. This method illustrates not only what the
instructions are but, also, how to use them in a practical way.
Chapter 5 covers the rest of the PIC instruction set in the same way as in
Chapter 4 with a more advanced security system design which includes a keypad and a
digit display interface.
Chapter 6 covers the fundamentals of general ESP and shows the reader the style,
technique and art of ESP. These techniques and ideas apply to all processors and are the
main-stay of all of low-end embedded systems programming. Examples of coding and
programs are given in the PIC assembly language.

14
Chapter 7 covers more advanced ESP techniques, such as sine-wave generation,
DTMF signaling, data compression, pulse-width modulation, and testing techniques.
Chapter 8 covers the PIC peripherals with complete examples of how to use them.
These include counters, timers, interrupts, the analog-to-digital converter, the pulse-width
modulators, measurement hardware, and event generation hardware.
Chapter 9 covers the PIC peripherals for serial data communications. These are
the USART, a shift-register interface, and the “Inter-Integrated Circuit” serial interface.
Chapter 10 is an introduction to the fundamentals of Digital Signal Processing
(DSP) in an intuitive way and with detailed examples. Filter designs and programs are
given in a cookbook fashion. Some advanced and exotic DSP applications and
techniques are given.
Appendix A is a detailed view of the PIC instruction set.
Appendix B is a set of useful C++ programs which help the reader design projects
for the PIC.
Appendix C is a list of special-function registers and their bit-settings.
Appendix D is a register-file map.
Appendix E shows the PIC16F877 external pins and their functions.
Appendix F shows the sequence of instructions to save registers and restore them
when doing a processor interrupt.

15
Chapter 2: Microcontrollers and the PIC16F877
2.0 Chapter Summary
Section 2.1 covers the types of memories used in the general sense and their
organization with respect to the data and addressing. Section 2.2 discusses the memories
used in the PIC. Section 2.3 introduces the most fundamental elements of programming
at the assembly language level.
2.1 Memory and Memory Organization
A microcontroller is a complete computer system on a single chip. It is more than
just a microprocessor: It also contains a Read-Only Memory (ROM), a Read-Write
Memory (RAM), some input/output ports, and some peripherals, such as,
counters/timers, analog-to-digital converters, digital-to-analog converters, and serial
communication ports.
The internal view of a typical microprocessor is shown in Figure 2-1 and is
composed of three things: an arithmetic/logic unit (ALU) which performs calculations on
data; a set of registers which hold the user’s data and the system’s data; and a control unit
which orchestrates everything and interprets and executes the user’s instructions. As far
as the microprocessor is concerned, it assumes that there are sets of data memories and
program memories (RAM and ROM) in the system. The only thing the microprocessor
has to do is run a cycle of getting new instructions and executing them from the
memories.
Both the RAM and the ROM are organized as indexed sets of data words, where
each “index” is the “address” of its corresponding data. Both the data and its address
codes are numbers represented in binary or hexadecimal.
The RAM is a read-write memory which can rapidly read and write the data. It is
a volatile memory which means that it loses its memory when power is removed (turned
off). The ROM is for program memory and is “read-only” except in modern variants,
such as Electrically Erasable Programmable Read Only Memory (EEPROM) and Flash
Memory, which allow data words to be written as well as read. The writing of an
EEPROM is not the same as a RAM since the data-writing time of the EEPROM is about
ten thousand times as long as the data-writing time of the RAM. The ROM and its
variants are non-volatile memories that preserve their memories when the power is
removed (turned off).

16
ALU
Registers
Data Bus Address Bus
Control
Unit
To / From
System
Memory
To Address
System
Memory
Microprocessor
Figure 2-1
uP Internal View
Block Diagram
2.2 The PIC16F877
The PIC16F877`s internal block diagram is shown in Figure 2-2. The PIC
contains an ALU, which does arithmetic and logic operations, the RAM, which is also
called the “register-file”, the program EEPROM (Flash Memory), the data EEPROM, and
the “W” register. The “W” register is not a part of the register-file but is a stand-alone,
working register (also called an “accumulator”).
The ALU, the RAM, the “W” register, and the data EEPROM each manipulate
and hold 8-bit-wide data, which ranges in value from zero to 255 (or, in hexadecimal,
from 0x00 to 0xFF).
The program EEPROM (Flash Memory) works with 14-bit-wide words and
contains each of the user’s instructions.
It is not uncommon for microcontrollers to have different sizes of data memory
and program memory (in the PIC: 8-bits for data and 14-bits for program words). More
than that, the key is that the data and program memories occupy separate spaces. This
allows access to each at the same time.

17
ALU
W Register
Timing &
Control
In / Out Ports &
Other Peripherals
RAM
128 128 128 128
Bytes Bytes Bytes Bytes
0 1 2 3
4- Banks
Program
EEPROM
8192
Words
(14-Bits Long)
Data
EEPROM
256 Bytes
Figure 2-2 PIC16F877 Internal Block Diagram
The PIC’s RAM addresses range from zero to 511 but the user can only access a
RAM byte in a set of four “banks” of 128 bytes each and only one bank at a time. Not all
of this RAM is available to the user as read-write memory, however. Many addresses are
dedicated to special functions within the processor but they “look-like” RAM and are
accessed the same way.
The PIC’s program EEPROM (Flash Memory) has addresses that range from zero
to 8191 (0x1FFF). The user’s program occupies this memory space.
2.3 Programming The PIC
All types of computer programs can be broken-down into four main sets of
actions:
1) Top-Down Execution
2) Conditional Branching
3) Loops
4) Subroutine Calls

18
Programming the PIC in assembly language is no exception but it is more difficult to
work with than high-level languages, like BASIC and C++.
Assembly language uses a one-to-one correspondence of mnemonic words with
the binary machine codes that the processor uses to code the instructions. The user writes
the program using the mnemonic words called the “source” program and gives this to the
program on the PC called the “assembler” which converts it into the machine code of the
PIC in the form of a list of hexadecimal numbers. This set of numbers is called the
“object” program. The user then writes the object program into the PIC in the
downloading process of programming the PIC. When this is done, the PIC is ready to run
its new program.
Understanding how to code a program in assembly language is contingent upon
understanding how the PIC works at the machine level.
The PIC executes instructions from program memory in sequential addresses,
starting from address zero, when the PIC is reset upon power-up. The address of the
current instruction being executed is given in a special register called, the “program-
counter” (PC). The PIC’s control unit automatically increments the program-counter
(PC), gets the next instruction, decodes that instruction, and then executes it. If this is
done on sequential addresses, this is called, “top-down” execution. There are also ways
to do non-sequential-address executions. This is done with special instructions which
load new addresses into the program-counter. This is how conditional-branching, loops,
and subroutines are done at the machine language level.
Each line of source program code in assembly language has up to four parts: A
label, an op-code, an operand, and a comment. This is shown as follows:
LABEL: OPCODE OPERAND(S) ; COMMENT
The label is an arbitrary name the user picks to mark a program address, a RAM address,
or a constant value. If the label has its first character (a letter) that starts in column one of
the text, the colon is optional. Otherwise, the colon separates the label from the “Op-
Code”. The “Op-Code” is short for, “Operation-Code”, and is the mnemonic name for
the instruction to be executed. The operand or operands are the data or the address that
the instruction uses when it is to be executed. This is where labels come into play such as
when an instruction needs a new address. Comments are optional and must begin with a
semi-colon. Comments are for documenting the source program so that it will be easy to
read and understand. For example, the following lines are valid source program code
lines:

19
LOOP: BSF FLAGS,2 ; Set Alarm Flag Bit
MAIN: CALL SORT_SUB ; Sort the Data
DECFSZ TEMP,F
; This is an example of a line which consists of only a comment
MAIN CALL SORT_SUB ; Sort the Data (No Colon in Label)

20
Chapter 3: Simple PIC Hardware
& Software (“Hello World”)
3.0 Chapter Summary
Section 3.1 introduces a simple PIC system and analyzes its controlling program.
It looks at each instruction, examines its structure and coding, and sets a foundation for
proper usage of assembly language. Section 3.2 summarizes the instructions and
concepts covered in Section 3.1.
3.1 A Simple Example System
An external view of the PIC with its pins is shown in Figure 3-1. This is the most
basic view of the PIC’s functions. Most pins have a second use, or, even a third use: to
run the PIC’s peripherals, such as the ADC, the timers, and the serial ports, to name a
few. These other functions will be shown later as they are needed.
The basic function of these pins is for digital inputs and digital outputs. The
individual bits on each of the input/output ports (A-through-E) can each be selected as
“input” or “output” by special configuration registers in the RAM. The software must set
these bits before the ports can be used. This will be shown in more detail shortly.
An example circuit which uses the input/output (I/O) ports “B” and “C” is shown
in Figure 3-2. This is a simple, bare minimum, PIC example circuit that serves to
introduce some simple software and instructions. Port “B” has a set of 8 DIP switches
with resistor pull-ups on it to allow data from these switches to be read into the PIC. Port
“C” drives a set of 8 LEDs through resistors to allow the PIC to send out and display its
data. The power supply must drive both sets of power pins as shown in Figure 3-2. The
clock input, OSC1, is driven by an external oscillator module as shown. Also, a 10K
Ohm pull-up resistor is used on Pin 1 (/MCLR) to keep the PIC out of its “Reset” state.
The following program can be used to run the circuit of Figure 3-2:
LIST P = 16F877
INCLUDE “P16F877.INC”
ORG 0x0000 ; Program starts at address zero.
NOP
BANKSEL PORTC ; Select Bank Zero
MOVLW B’00000000’ ; Reset PORTC
MOVWF PORTC

21
BANKSEL TRISC ; Select Bank One
MOVLW B’00000000’ ; Make PORTC All Outputs
MOVWF TRISC
MOVLW B’11111111’ ; Make PORTB All Inputs
MOVWF TRISB
BANKSEL PORTC ; Select Bank Zero
MAIN:
MOVF PORTB,W ; Read DIP Switches into W-Register
MOVWF PORTC ; Write W-Register to LEDs
GOTO MAIN ; Loop To Address “MAIN”
END
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20 21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40/MCLR or Vpp
RA0
RA1
RA2
RA3
RA4
RA5
PORT A
RE0
RE1
RE2
PORT E
Vdd = +5 Volts
Vdd = +5 Volts
Vss = Ground
Vss = Ground
OSC1
OSC2
RC0
RC1
RC2
RC3
PORT C
Low Half
RD0
RD1
RD2
RD3
PORT D
PORT DLow Half
Low Half
RC4
RC5
RC6
RC7
PORT C
High Half
RD4
RD5
RD6
RD7
PORT D
High Half
RB0
RB1
RB2
RB3
RB4
RB5
RB6
RB7
PORT B
PIC16F877
Figure 3-1 Simple Hardware View (Ports Only)
After setting up Port “C” as an “output” and Port “B” as an “input”, this program
reads the value of the switches on Port “B” and sends it back out to the LEDs on Port “C”
in a continuous loop. This may seem like a very trivial program but it is still useful as a
test and a simple demonstration of the PIC.
This program is an example of an assembly language program. The MPLAB
assembler on the IBM-PC takes this ASCII coded text and converts it into machine code
for the PIC to use.

22
PIC16F877
RB7
RB6
RB5
RB4
RB3
RB2
RB1
RB0
PORT C
Vdd
Vss
OSC1
/MCLR
10 K Ohm+5 Volts
+5 Volts
+5 Volts
OSC
4 MHz
Figure 3-2 Basic Hardware System Example
LEDs
LEDs470 Ohms 470 Ohms
LSB
MSB
RC0
RC1
RC2
RC3
RC4
RC5
RC6
RC7
MSB
LSB
+5 Volts
10 K Ohms
The major part of this program uses the I/O ports and their configuration registers
and we can now look at the RAM or register-file bytes in extensive detail. Appendix D
shows the addresses and names of the registers and RAM bytes used by the PIC.
The RAM or Register-File is divided into four banks of 128 bytes each. Only one
bank can be used at a time. Not all of the bytes in a bank can be used as user memory
because some bytes have special purposes, such as I/O ports. However, all of these bytes
work like RAM: They can be read from, and written to, just like a memory byte.
Each register-file byte has a unique address within its bank which ranges from
zero to 127 (0x00 to 0x7F). The register, PORTC, for example, is located in Bank zero
and has the address 0x07. Notice that some register-file bytes in Appendix D are
repeated across each of the banks at the same corresponding addresses. For example,
STATUS, PCL, FSR, PCLATH, and INTCON all occupy the same line addresses in each
of the four banks. This is so that they can be accessed and have the same values
contained in them independent of the bank which is currently selected. For the other
registers which are not repeated, the proper bank must be selected before they can be
used. For example, storing data in the register byte at address 0x30 in Bank Zero,
changing the current bank to Bank One, and then trying to read the right data in the
register byte at address 0x30 will not give the same data that was stored even though the
addresses were both 0x30. Bank Zero must be selected again as the current bank to get

23
the right data. However, the RAM addresses from 0x70 through 0x7F can be used in any
bank to reference the same data without switching banks. This is a very valuable feature!
The usage of the register files will be made more clear as we analyze the program
and consider how the instructions are coded and used.
The program starts with the statements, “LIST” and “INCLUDE”. These are not
instructions in that they are not translated into machine code. They are “assembler
directives” which tell the assembler (MPLAB) where to get information about the PIC
chip being used (PIC16F877). All of your programs must start with these statements, in
this order, as shown.
The statement, “ORG 0x0000”, is an assembler directive which tells the
assembler at what address the following instructions will start. Here they will start at
address zero where "0x0000” indicates “hexadecimal zero”.
The next statement is the “NOP” instruction. It is a true instruction that gets
translated into machine code. “NOP” stands for “No Operation” (it does nothing). It is
coded as 14 zero bits, or, 0x0000 in hexadecimal. Remember that all of the instructions
are each 14 bits long.
The next statement is “BANKSEL PORTC”. This is not a true instruction, per
se, but it does get translated into two machine instructions which select the current
register file bank to be used. Here, “PORTC” is interpreted as “Bank Zero”. Later
“BANKSEL TRISC” will be interpreted as “Bank One”. In general, the operand of
BANKSEL will select the lowest bank number where the operand name is found in
Appendix D.
The next statement is the “MOVLW” instruction which means “move the literal
value that follows into the W register”. Remember that the W register is not a part of the
RAM but is an “accumulator” or “working register” within the PIC. The instruction:
MOVLW B’00000000’
Says, “move the binary value of all zeros to W”. The general machine coding of the
“MOVLW” instruction is:
11 0000 kkkk kkkk
where the “k”s are the single binary bits of the literal data (the data byte). The
instruction:
MOVLW B’00000000’

24
Would be coded as:
11 0000 0000 0000 or 0x3000.
Later in the program is the instruction
MOVLW B’11111111’
Which is coded as:
11 0000 1111 1111 or 0x30FF.
The next instruction is, “MOVWF PORTC”, which means, “move the value in
the W register to the register-file byte at the address given (address given here is
PORTC)”. That is, “Move W to RAM”. In general, this is coded as:
00 0000 1fff ffff
where the “f”s are the binary bits of the address for the desired register-file byte. Since
Port C has the address 0x07 the coding of “MOVWF PORTC” would be:
00 0000 1000 0111 or 0x0087.
It would be perfectly legal in the assembly language program to say, “MOVWF 7” to
mean, “MOVWF PORTC” and it would be translated in exactly the same way.
However, it would be very confusing to anyone who tried to read the program and, for
this reason, it is better to let the assembler keep track of the register-file names and let the
assembler translate them to the proper addresses. In general, it is best to let the assembler
translate all of the label names (RAM or ROM) in the program to their values. Never use
translated numbers directly in the ASCII text of the program. Always use label names.
This is a very good programming practice.
At the bottom of the program there is the label, “MAIN”, and the instruction,
“MOVF PORTB,W”. The label “MAIN” is there to mark the address of the “MOVF”
instruction so that we can come back to that address. The “MOVF” instruction stands
for, “move the register-file byte value to the W register”. Specifically, it moves the value
in Port B to the W register. (It is also possible to say, “MOVF PORTB,F”, which would
mean, “move the value in Port B to Port B”. This sounds absurd but there is a reason that
this is allowed.)
The next new instruction is the, “GOTO MAIN”, which means, “put the address-
value at MAIN into the program-counter to transfer to that address”. This causes a
“loop” to occur in the program. The instruction at the address “MAIN” is the “MOVF”.

25
The last statement is the, “END”, which is an assembler directive that means,
“end the program”.
Now that the program directives and instructions are understood, we can come to
a better understanding of how the program works.
The first step in the program is to send zeros to Port C in preparation for setting
up Port C to be an output port. This may seem backwards (sending data to a port before
it is declared to be an output port) but there is a good reason for it: When the declaration
is made as “output”, whatever data that is in the port output buffer is immediately sent
out. If there is garbage in the port output buffer, that garbage will be the first output from
the port. This is why zeros were sent to the port first.
The next steps are to declare Port C and Port B to be as “output” and “input”,
respectively, by using the TRISC and TRISB registers. If a “1” is sent to a TRISC bit,
the corresponding Port C bit will be an “input”. If a “0” is sent, the Port C bit will be an
“output”. This rule applies to all of the ports (A, B, C, D, and E) and their corresponding
TRIS registers. (Ports “A” and “E” need additional configurations, however.)
In the “MAIN” loop, the data from the DIP switches on Port B is moved to the W
register as an intermediary and is then moved to Port C to display the data on the LEDs.
Then the “GOTO” closes the loop to form an infinite loop.
The thing for you to do now is to enter this program in an ASCII text editor and
run a simulation of it in the MPLAB. See if you can enter and run this program.
3.2 Summary of Instructions and Concepts
1) When you are using MPLAB, your programs must start with:
LIST P=16F877
2) The register-file map in Appendix D shows where each special function
register-file is and which RAM banks restrict its use.
3) RAM addresses from 0x70 through 0x7F may be freely used by the user and
are accessible as the “same” data in any bank.
4) The “ORG” assembler directive tells the program where the current program
memory address is. Use, “ORG 0xhex-number”.
5) The “NOP” instruction means “No Operation”. It does take time for the PIC
to run it, however.
6) The “BANKSEL” directive selects the RAM bank for the current use.
Consult Appendix D to get the right RAM bank.
7) The “MOVLW” instruction means “Move the Literal value that follows into
the W register”.
8) The “MOVWF” instruction means “Move the W register’s data into RAM”.
The data in the W register remains unchanged after doing this instruction.

26
9) The “MOVF” instruction means “Move data from the RAM into the W
register”.
10) The “GOTO” instruction “Loads the Program Counter with the address (label)
that follows” to “Jump” to a new, non-sequential address in program memory.
11) Labels are used to reference addresses in program memory so that “GOTO”
instructions can use them.
12) The “END” directive is the last line of any program and is used to signal the
end of the assembly process.
13) Always use label-names to reference RAM or program memory.
14) Send zeros to a port before you declare the port as “output” so that garbage
will not initially corrupt the port’s output.
15) Use the appropriate “TRISx” register to declare a port as “input” or “output”.

27
Chapter 4: The PIC Instruction Set (Part I)
4.0 Chapter Summary
Section 4.1 covers the PIC instruction set by building a simple security system. It
covers the RAM and ROM in more detail, looks at instruction timing, and introduces the
use of subroutines. Section 4.2 summarizes the instructions and concepts.
4.1 The PIC16F877 Instruction Set
The PIC16F877 only has 35 instructions and it would be easy to list them all at
once. It would, then, be harder to explain them and give examples of how they are used.
A much more natural way is to pick a specific application for the PIC then illustrate the
instruction set, show how the instructions are used, and show how a program uses
sequences of instructions to accomplish the desired tasks.
Let’s consider the design of a home security system or a burglar alarm using the
PIC. A simple security system circuit is shown in Figure 4-1. The main alarm sensor
that is used in this system is the magnetic reed switch. Figure 4-1 shows five of these in
series, tied to ground, and connected with a pull-up to the PIC input port pin. Suppose
that Port B is configured as an input port and that this line connects to RB0 (bit zero).
The magnetic reed switch has two parts: The switch, itself, which is mounted to a door
frame or a window frame, and a magnet, which is mounted to the door or the window.
When the door or the window is closed, the magnet holds the reed switch “ON” or
“shorted”. When the door or the window is moved, the magnetic field is broken and the
reed switch turns “OFF” or “open”. This is the mechanism that trips the alarm.
The software segment that would sense an alarm is:
ALARM_TEST:
BTFSS PORTB,0
GOTO ALARM_TEST
Where “BTFSS” means, “Bit-Test-File-Skip-if-Set”. That is, if Port B, bit zero (RB0) is
sensed as a “one” (“set”), the next instruction will be skipped. When each reed switch is
closed, the Port B, bit zero line will read as “zero” since it is tied to ground. When any
reed switch is opened, the Port B, bit zero line will be pulled high (“one”) and this will be
the alarm condition.

28
Figure 4-1 A Simple Security System
Magnetic Reed Switches
Each Magnet Holds Switch "On"
Power
Oscillator
4 MHz
PIC16F877
+5 Volts
+5 Volts
+12 Volts
Switch
Disarm
Arm /
Relay Coil
2N3904
10K Ohm
10K Ohm
470 Ohm
1 K Ohm
LED
"Ready To Arm"
RB1 (Input)
RC0 (output)
RB0 (Input)
RC1 (Output)
0.01 uF
To Alarm
Circuit
The general format for the “BTFSS” instruction is:
BTFSS Register-File,Bit-Number
Where the bit-number selects one of eight (8) bits as 0-through-7. There is also a
“BTFSC” instruction that is identical to “BTFSS” except that it skips when the tested bit
is “Clear” (“zero”).
When the alarm condition is sensed, the alarm will be activated by setting Port C,
bit one (RC1) to a “one” (“1”) as:
BSF PORTC,1
Which means, “Bit-Set-File”. This will turn the transistor “ON”, turn the relay “ON”,
and the relay contacts will switch on the alarm. The alarm can be reset with:
BCF PORTC,1
Which means, “Bit-Clear-File”. Both “BSF” and “BCF” specify one of eight (8) bits as
0-through-7 for any RAM byte or register-file.

29
The alarm circuit has two other hardware elements that need to be controlled.
One is an “Arm/Disarm” switch input on Port B, bit one (RB1) and the other is a “Ready-
To-Arm” LED on Port C, bit zero (RC0).
The software that would control this system is as follows:
LIST P=16F877
REED: EQU 0 ; Reed Switch is PORTB, Bit Zero
ARM: EQU 1 ; Arm/DisArm Switch is PORTB, Bit One
LED: EQU 0 ; “Ready-To-Arm” LED is PORTC, Bit Zero
ALARM: EQU 1 ; Alarm is PORTC, Bit One
ORG 0X0000
NOP
BANKSEL PORTC ; Send Zeros to Output Port
MOVLW 0X00 ; Before Set-Up
MOVWF PORTC
BANKSEL TRISB
MOVLW 0XFF ; Port B is to be Input
MOVWF TRISB
MOVLW 0X00
MOVWF TRISC ; Port C is to be Output
BANKSEL PORTB
MAIN:
BTFSS PORTB,REED; Test Reed Switches, Skip if “Open”
GOTO READY
RESET:
BCF PORTC,LED ; Reset “Ready-To-Arm” LED
BCF PORTC,ALARM ; Reset Alarm
GOTO MAIN
READY:
BSF PORTC,LED ; Set “Ready-To-Arm” LED
BTFSS PORTB,ARM; Test “Arm/DisArm” Switch, Skip if Set
GOTO MAIN
ALARM_SENSE:
BTFSS PORTB,REED; Test Reed Switches, Skip if “Set”
; (Alarm)
GOTO ALARM_SENSE
BTFSS PORTB,ARM; Test “Arm/DisArm” Switch, Skip if Set

30
GOTO MAIN
SET_ALARM:
BSF PORTC,ALARM ; Activate Alarm
BTFSC PORTB,ARM; Test “Arm/DisArm” Switch….
GOTO SET_ALARM; ---- Still Set, Stay in Alarm
GOTO RESET ; ---- Not Set, Reset Alarm
END
This program uses a new assembler directive called, “EQU”, or “Equate” to
define constants, such as bit positions and data, as label names. Its usage is:
Label: EQU “data”
It is good programming practice to use EQUs to make programs more readable.
This alarm system program is fine if the “Arm/DisArm” switch and the “Ready-
To-Arm” LED are located on the outside of the house. If they are located on the inside of
the house, this program has a serious problem: It is impossible to leave the house after
the system is armed since the alarm will be tripped when the door is opened!
This problem is usually solved by adding a delay of thirty (30) seconds before
activating the alarm so that the user can leave the house and not trip the alarm. An
external timer could be attached to the PIC to provide this delay but the better solution is
to let the PIC generate its own delays in software.
The PIC runs with a clock frequency of 4 MHz and this controls instruction
execution speed. The instruction execution speed is one-fourth of the clock frequency.
Each instruction executes in one or two instruction cycles, or, in one or two
microseconds. Each instruction that “skips”, such as, “BTFSS”, executes in two
instruction cycles when it “skips” and one instruction cycle when it doesn’t. The
“GOTO” instruction always executes in two instruction cycles.
Let’s see how to build a 30-second software delay. The first step is to build a
one-millisecond delay as follows:
MOVLW D’250’ ; Load W with Decimal 250
MOVWF TIME ; Initialize RAM “TIME”
LOOP_ONE_MS:
NOP ; (1) Cycle
DECFSZ TIME,F ; (1) Cycle
GOTO LOOP_ONE_MS ; (2) Cycles
This program segment uses the “DECFSZ” instruction, which means, “Decrement-File-
Skip-if-Zero”. When the Byte’s data is decremented, it may be stored back in the RAM

31
or it may be stored in the W register. Specify the RAM as the destination with the “F” or
specify the W register with the “W”. This instruction decrements the RAM byte
(“TIME”, in this case) and skips the next instruction if the result was zero. The initial
value of “TIME” is 250 and the loop has a length of four instruction cycles. As the loop
counts down from 250 each count adds four microseconds of delay which, when
multiplied by 250, gives a total time of about one millisecond.
A loop can then be formed around this loop that counts down from 250 to give a
delay of 250 milliseconds. Then another loop can be placed around it that counts down
from 120 to give a total delay of 30 seconds!
(There is also an “INCFSZ” instruction which increments a register-file and skips
the next instruction if the result is zero. Either “INCFSZ file,F” or “INCFSZ file,W”.)
The total nested-loop structure for a 30-second delay is:
MOVLW D’120’ ; Count 120 of 250 millisecond delays
MOVWF TIME2
LOOP_30_SEC:
MOVLW D’250’ ; Count 250 of one millisecond delays
MOVWF TIME1
LOOP_250_MS:
MOVLW D’250’ ; Count of 250 Loops of four cycles
MOVWF TIME
LOOP_ONE_MS:
NOP
DECFSZ TIME,F
GOTO LOOP_ONE_MS
DECFSZ TIME1,F
GOTO LOOP_250_MS
DECFSZ TIME2,F
GOTO LOOP_30_SEC
This code must be put into two places in the alarm software: Once just before the
“ALARM_SENSE” loop and once just after it. This will allow the user 30 seconds to
enter and leave before the alarm goes off.
This is a lot of code to make two copies of and insert into a relatively simple
software loop. The resulting code would be much more complicated, much more
difficult to follow, and the probability of making a mistake would be much higher. Are
there any easier alternatives?
There are. The solution is to use subroutines with the “CALL” and “RETURN”
instructions. The “CALL” instruction works just like a “GOTO” instruction except that

32
the PIC automatically saves the address of the next instruction after the “CALL” in a
special memory. This action is automatic and invisible to the user. When you are
finished with the subroutine, you use the “RETURN” instruction to tell the PIC to go
back to the part of the program where you left off. That is, to get the address out of the
special memory and “GOTO” it automatically. The “CALL” and “RETURN”
instructions each take two instruction cycles to execute.
For example, the one-millisecond delay can be accomplished with the following
subroutine:
DELAY_ONE_MS:
MOVLW D’250’ ; Count 250 Loops
MOVWF TIME
LOOP_ONE_MS:
NOP ; Four Cycle Loop
DECFSZ TIME,F
GOTO LOOP_ONE_MS
RETURN
This would give a delay of one millisecond each time it was called, as follows:
CALL DELAY_ONE_MS
We could also make a 250-millisecond delay subroutine as follows:
DELAY_250_MS:
MOVLW D’250’ ; Count 250 Milliseconds
MOVWF TIME1
LOOP_250_MS:
CALL DELAY_ONE_MS
DECFSZ TIME1,F
GOTO LOOP_250_MS
RETURN
Notice that this subroutine calls the “one-millisecond” subroutine 250 times. There is no
law that says you can’t have one subroutine that calls another subroutine – you can – it’s
called “nesting” subroutines. The PIC allows you to nest subroutines up to only eight (8)
levels, however.
The 30-second delay can then be formed as the following subroutine:
DELAY_30_SEC:
MOVLW D’120’ ; Count 120 of 250-Millisecods
MOVWF TIME2
LOOP_30_SEC:

33
CALL DELAY_250_MS
DECFSZ TIME2,F
GOTO LOOP_30_SEC
RETURN
1) Subroutines may be nested up to eight (8) levels deep.
2) Each instruction takes either one or two instruction-cycles to run.
3) Each instruction-cycle time runs in the period of the PIC’s oscillator
frequency, Fosc, divided by four (4). If Fosc = 4 MHz, each instruction-cycle
will run in one microsecond.
4) The “EQU” directive is used to make user-defined labels for RAM addresses,
constants, and bit-positions.
5) The “GOTO”, “CALL”, and “RETURN” instructions each take two
instruction-cycles to run. Also, anytime an instruction “skips” it takes two
instruction-cycles to run.
6) New Instructions Covered:
BTFSS file,bit
BTFSC file,bit
BSF file,bit
BCF file,bit
DECFSZ file,destination
INCFSZ file,destination
CALL address
RETURN

34
Chapter 5: The PIC Instruction Set (Part II)
5.0 Chapter Summary
Section 5.1 discusses the need for a more advanced security system. Section 5.2
introduces the keypad and display interfaces. Section 5.3 introduces the processor status
flags. Section 5.4 details the keypad software. Section 5.5 details the LED digit display
software. Section 5.6 explains indirect RAM addressing. Section 5.7 covers more
advanced general features of the PIC processor. Section 5.8 shows how the keypad and
the display software work in harmony with each other. Section 5.9 is a last look at the
security system. Section 5.10 summarizes the instructions and concepts of Chapter 5.
5.1 Introduction
The basic PIC instructions were introduced in Chapter 4 with an example of a
simple security system. In a similar way, the remaining instructions will be introduced
with an example of a more complex security system.
A more complex system is needed for a large house or an industrial user. If there
are many rooms, doors, and windows, a single “Ready-To-Arm” LED is not enough to
give the location of the “insecure” site when the user desires to set the alarm. Also, a
single “Arm/DisArm” switch presents no difficulty to a thief who knows where the
switch is and wants to disable the alarm. The security system should also offer the user a,
“Home/Away” setting so that a remnant of the system can still work while the user is at
home (i.e., when he or she is asleep or wants to use the bathroom without triggering the
alarm).
These problems could be solved by having many “Ready-To-Arm” LEDs and
many “Arm/DisArm” switches but this is not practical when flexible and compact
software is available. The solution is to use a twelve-key keypad and a set of seven-
segment LED digits to get and display more complex information to run the security
system. For example, the LED digits could display a message like, “Zone 3 Not Ready
To Arm”, or, “Office 407 Not Ready”. These messages could be static, or constantly
displayed, or they could repeatedly scroll across the display over a few seconds. Code
numbers could be entered on the keypad to set-up and run the system. An
“Arm/DisArm” switch is not a problem for a would-be thief but entering a four-digit
security-code that could be changed daily is much more formidable!
Before we can introduce more instructions and show the software that could do
these things, we need to show how the keypad and the LED digits are interfaced to the
PIC.

35
5.2 Keypad And Display Interface
A twelve-key keypad usually consists of a matrix of twelve switches, as shown in
Figure 5-1. Three lines connect the columns while four lines connect the rows. When a
key is pressed, one column and one row are connected. The PIC must use its software to
scan each column-line and look for a row-line that is connected, if, and when, a switch is
pressed. The PIC keypad interface circuit is shown in Figure 5-2.
Column 1 Column 2 Column 3
Row 1
Row 2
Row 3
Row 4
"1" "2" "3"
"4" "5" "6"
"7" "8" "9"
"*" "0" "#"
Figure 5-1 Twelve-Key Matrix Keypad

36
PIC16F877
Key Pad
12- Key
Matrix
Row 4
Row 3
Row 2
Row 1
Column 1
Column 2
Column 3
10 K Ohms
+5 Volts
PORT D
RD0 through RD3
Are Inputs
RD4 through RD6
Are Outputs
RD3
RD2
RD1
RD0
RD4
RD5
RD6
Figure 5-2 PIC Matrix Keypad Interface Circuit
A typical seven-segment LED digit is shown in Figure 5-3. The digit consists of
eight (8) LEDs: Seven for the segments and one for the decimal point. The LEDs are
connected together in a “common-anode” or a “common-cathode” form. If only one digit
is to be used, the common-cathode is tied to ground or the common-anode is tied to five
volts, and the PIC would drive each segment LED through a 470 Ohm resistor, as shown
in Figure 5-4. If more than one digit is to be used, the digits are usually multiplexed, as
shown in Figure 5-5. Multiplexing digits is a way of saving PIC output port lines by
time-sharing them. Let’s see how this works.

37
A
B
C
D
E
F
G
DP
A
A
B
B
C
C
D
D
E
E
F
F
G
G
DP
DP
Common Cathode
Common Anode
OR
Figure 5-3 Seven Segment LED Digit Display
In Common Cathode and Common Anode Forms
LEDs
LEDs
For digit multiplexing, the corresponding segments of each digit are wired
together in parallel, as shown in Figure 5-5. Each common-anode or common-cathode of
a single digit is driven by a transistor. The idea is to turn on one transistor at a time and
supply that digit’s segments for a few milliseconds at a time and then turn it off. Then
repeat this for each transistor and digit’s segments for the whole display, over and over.
The overall effect on the human eye is that all of the digits appear to be “on” at the same
time! Each digit gets the right information and there is no blurring or garbage.

38
PIC16F877
470 Ohms
A
B
C
D
E
F
G
DP
RC0
RC1
RC2
RC3
RC4
RC5
RC6
RC7
Seven
Segment
LED
Digit
Display PIC16F877
470 Ohms
A
B
C
D
E
F
G
DP
RC0
RC1
RC2
RC3
RC4
RC5
RC6
RC7
Seven
Segment
LED
Digit
Display
(Common
Cathode)
(Common
Anode)
+5 Volts
Figure 5-4 Single LED Digit Drives for
Common Cathode and Common Anode Forms
(The PIC's Output to the
Common Anode is Inverted)
PIC16F877
PORT B, 8-Lines To Segments (A,B,C,D,E,F,G,DP)
(8) (8) (8) (8)
LEFT LEFT 2 LEFT 3 RIGHT
DIGIT DIGIT DIGIT DIGIT
PORT C
RC3
RC2
RC1
RC0
1 K Ohm 1 K Ohm 1 K Ohm 1 K Ohm
Figure 5-5a Multiplexed LED Digit Drives for Common Cathode Form

39
PIC16F877
PORT B, 8-Lines To Segments (A,B,C,D,E,F,G,DP)
(8) (8) (8) (8)
LEFT LEFT 2 LEFT 3 RIGHT
DIGIT DIGIT DIGIT DIGIT
PORT C
RC3
RC2
RC1
RC0
1 K Ohm 1 K Ohm 1 K Ohm 1 K Ohm
+5 Volts +5 Volts +5 Volts +5 Volts
Figure 5-5b Multiplexed LED Digit Drives for Common Anode Form
5.3 The STATUS Register and Flag Bits
We are almost ready to begin a full discussion of how the keypad and the LED
digits are run in software. First, is should be noted that the instructions that are needed to
do this affect flag bits in the status register. Three flag bits are affected automatically as a
result of doing each of these instructions.
The STATUS register is shown in Appendix C. The bits RP1 and RP0 are set by
the user to select one of the four banks of the RAM registers. This can be done with the
BCF and BSF instructions or with the BANKSEL directive. The BANKSEL directive
produces the right combinations of BCF and BSF instructions for the STATUS register to
select the appropriate RAM bank.
The three flag bits that interest us most are the “Z”, the “DC”, and the “C” bits.
The “Z” bit is set if the instruction produces a result of zero. The “C” bit is the “carry”
bit and shows the carry-out of the seventh bit of the result (The result bits range from
zero-through-seven). The “DC” bit shows the “carry-out” of the third bit of the result
(this is used for binary-coded-decimal arithmetic). A complete list of which instructions
affect the flag bits, and how, can be found in Appendix A. The “MOVF f,d” instruction
is the only instruction that we have seen so far (Chapter 3) that affects a flag bit.

40
Specifically, it only affects the “Z” bit, showing if the value or data that is moved is zero
(Z = 1) or not (Z = 0).
We are now ready to write the software to control the keypad and the LED digits.
5.4 The Keypad Software
Look again at Figure 5-2. The row lines are pulled-up to five volts (Logic “1”)
and are connected through the key-switches to the column lines. The idea is to set the
column lines low (Logic “0”), one-at-a-time, and search for a low (Logic “0”) on the row
lines. If a row line is low, a key has been pressed, and the scanning software subroutine
returns a number code corresponding to the key. If no key is pressed, the routine returns
a code of zero.
The scanning subroutine is as follows:
KEY_SCAN:
BCF PORTD,4 ; Column 1 = LOW
BSF PORTD,5 ; Others = HIGH
BSF PORTD,6
BTFSS PORTD,0 ; Row 1
RETLW 1 ; Key = “1”
RETLW 10 ; Key = “*”
BSF PORTD,4
BSF PORTD,5

41
RETLW 12 ; Key = “#”
RETLW 0 ; No Key Pressed
The new instruction used in this subroutine is “RETLW k”, which means,
“Return from subroutine with the W register containing the eight-bit value, k”. Thus,
when KEY_SCAN returns, the key-code will be in the W register. The “RETLW”
instruction runs in two instruction cycles.
The KEY_SCAN routine can be improved by eliminating the repeated code to
scan the rows and replacing them with subroutine calls to a “ROW_SCAN” routine.
Notice that each of the row scans in KEY_SCAN return values as:
(1,4,7,10) (2,5,8,11) and (3,6,9,12).
These are additions of zero, one, and two on the base values of (1,4,7,10). Assume that
the ROW_SCAN routine will also return a zero if no keys are pressed. The improved
KEY_SCAN routine is as follows:
KEY_SCAN:
BSF PORTD,5 ; Others = HIGH
BSF PORTD,6
CALL ROW_SCAN ; (W) = Code
ADDLW 0 ; Add Zero to Set Zero Flag
BTFSS STATUS,Z
RETURN ; (W) = 1,4,7,10
BSF PORTD,4
BTFSC STATUS,Z
GOTO KEY_SCAN2
ADDLW 1 ; Adjust (W)
RETURN ; (W) = 2,5,8,11

42
KEY_SCAN2:
BSF PORTD,5
BTFSS STATUS,Z
ADDLW 2 ; Adjust (W)
RETURN ; (W) = 3,6,9,12 OR (W) = 0
ROW_SCAN:
BTFSS PORTD,0
RETLW 1 ; Key “1”
BTFSS PORTD,1
BTFSS PORTD,2
BTFSS PORTD,3
RETLW 10 ; Key “*”
RETLW 0 ; No Key Found
The new instruction used in the new KEY_SCAN routine is, “ADDLW k”,
which means, “Add the W register and the eight-bit value, k, and put the result in the W
register”. The “ADDLW” instruction affects the “Z”, “DC”, and “C” status flag bits.
Addition can also be done between a RAM byte (register-file) and the W register
with the “ADDWF f,d” instruction. This does, “W = (file) + W” or “(file) = (file) + W”
and it also affects the “Z”, “DC”, and “C” status flag bits.
Using the “ADDLW” instruction it is possible to make more improvements in the
KEY_SCAN and ROW_SCAN routines. In the original KEY_SCAN each set of row
value codes had a difference of three (i.e., From “1”, “4” = “1” + 3, “7” = “4” + 3, …).
What if the ROW_SCAN routine were called with the first row code in W and each time
the key test failed, we add three to W? Also, the “Z” flag can be set within the
ROW_SCAN routine so we don’t have to check it in the KEY_SCAN routine. Can you
write an improved version of the KEY_SCAN and ROW_SCAN routines using these
techniques? Can you think of even more improvements?

43
5.5 The LED Display Software
Let’s now look at the software controls for the four-digit, multiplexed, seven-
segment display shown in Figure 5-5a. Port B of the PIC controls the segments while
Port C, bits (0,1,2,3), control the digit-drives.
The display subroutine software is as follows:
DISPLAY:
CLRF PORTC ; Turn Off All Segments and Digits
CLRF PORTB
MOVF LEFT_DIGIT,W ; Get Left Digit’s Segment Codes
MOVWF PORTB ; Set-Up Segments for Activation
BSF PORTC,3 ; Activate Digit-Drive
CALL DELAY_5_MILLISECONDS
CLRF PORTB
MOVF LEFT_DIGIT2,W ; Get Left2’s Segments
MOVWF PORTB
BSF PORTC,2
CLRF PORTB
MOVF LEFT_DIGIT3,W ; Get Left3’s Segments
MOVWF PORTB
BSF PORTC,1
CLRF PORTC ;Turn Off All Segments and Digits
CLRF PORTB
MOVF RIGHT_DIGIT,W ; Get Right-Digit’s Segments
MOVWF PORTB
BSF PORTC,0
CLRF PORTC
CLRF PORTB
RETURN
One new instruction used is, “CLRF f”, which means, “Clear Register-File f”.
There is also, “CLRW”, which means, “Clear W”. Both of these set the “Z” (Zero) flag
so that Z = 1. The “C” and “DC” flags are not affected.

44
The “DISPLAY” routine must not have very much delay between calls, otherwise
the digits will appear dim or flicker.
The use of the “DISPLAY” routine presupposes that the proper display codes are
already in the digit RAM spaces (LEFT_DIGIT, …, RIGHT_DIGIT). What are these
display codes and how do we get them? Each display segment shown in Figure 5-3
corresponds to a bit position on Port B as:
Segment Port B Bit
“a” 0
“b” 1
“c” 2
“d” 3
“e” 4
“f” 5
“g” 6
“d.p.” 7
If we were to make a table of base-ten digits to display and their display codes, it would
look like:
Digit Display Code
0 0x3F
1 0x06
2 0x5B
3 0x4F
4 0x66
5 0x6D
6 0x7D
7 0x07
8 0x7F
9 0x6F
This table would be used, directly, by the following subroutine:
GET_DISPLAY_CODE:
ADDWF PCL,F ; Add (W) to PCL, With Result in PCL
RETLW 0x3F ; “0”
RETLW 0x06 ; “1”
RETLW 0x5B ; “2”
RETLW 0x66 ; “4”
RETLW 0x6D ; “5”
RETLW 0x7D ; “6”
RETLW 0x07 ; “7”

45
When this routine is called, the user puts the digit to find in the W register. The first
instruction, “ADDWF PCL,F”, adds the digit in W to the low-half of the program-
counter, PCL, and puts the sum in PCL. In effect, this does a “GOTO”-like instruction
and goes to the corresponding RETLW instruction where W is filled with the proper
display code upon returning.
For example, suppose that W = 1 when “GET_DISPLAY_CODE” is called.
When control is transferred to the “ADDWF PCL,F” instruction, the whole program-
counter is incremented, automatically, to get to the next instruction, and then W is added
to the low-half of the program-counter to point to the proper RETLW instruction. This is
the, “RETLW 0x06”, which has the display code for “1”, and this code is in the W
register when the subroutine returns.
Remember that the program counter is the register where the PIC keeps track of
the address of the current instruction being executed. It is actually composed of two
registers: A “high” half and a “low” half. This is because the PIC can address 8 K of
program memory and it needs 13 bits to do this. These must be split into two 8-bit bytes.
When the program counter is incremented, the low half is incremented first, and, if a
carry is generated, it is automatically added to the high half. The low half is the register
“PCL”. The high half is not directly accessible but it can be set-up through the register
“PCLATH” (We will see more of this later.)
Any table look-up routine, like “GET_DISPLAY_CODE”, can be used with a
number of entries up to or equal to 255. Some care must be taken, however, for any
table, large or small. The “ADDWF PCL,F” instruction only adds to the low-half of the
program-counter. There is no carry propagation to the high-half. If, for example, the
“ADDWF PCL,F” instruction is located at the address 0x01FC and the table extends for
ten entries, a value in W equal to four (4) will transfer control to the address 0x0100 and
whatever is in that location will be treated as if it were an instruction (it may be) and the
program will continue there! You must make sure, by using the “ORG” directive, to set
an address that allows a large enough space to fit the table.
If the CALL to a table look-up routine is not within the same 256-byte address
block as the table itself, the table’s high-half address must be loaded into the PCLATH
register before the CALL is made. For example,
ORG 0x0100
------
------
MOVLW HIGH GET_DISPLAY_CODE
MOVWF PCLATH
MOVF DIGIT,W
CALL GET_DISPLAY_CODE

46
-----
-----
ORG 0x0380
GET_DISPLAY_CODE:
ADDWF PCL,F
RETLW 0x3F
-----
-----
This code gets the high-address of the GET_DISPLAY_CODE routine into the PCLATH
register before the CALL. The CALL is in the 0x01 block while the routine is in the
0x03 block.
5.6 Improved DISPLAY and Indirect Addressing
In actual practice, the “DISPLAY” routine would be done differently to improve
its usage of delay timings. The five-millisecond delay routine was called four times –
once for each digit. This eats up too much time. It would be better to call the
“DISPLAY” routine once every five milliseconds and let the program keep track of
which digit is to be displayed. This requires an addressing technique called, “indirect
addressing”.
Indirect addressing allows you to use a RAM location to index the address of
another RAM location. The index RAM location is called, “FSR”, and the RAM address
to be indexed is put into the “FSR” register-file. With the address of the RAM to use, in
place, in the “FSR” register, the user works with a RAM location called, “INDF”. See
Figure 5-6 for more information and explanation of the concept of indirect addressing.

47
Suppose we are in Bank Zero and we want to index 16 Bytes in a row in RAM
starting from Address = 0x30 to Address = 0x3F. This will be a RAM Data Array.
1) Move the Address Value = 0x30 into the FSR register.
2) Work with any instruction that references a RAM Byte, where “INDF” is
used in place of the desired RAM Byte. For example:
INCFSZ INDF,F will increment the Byte at Address = 0x30
And skip if zero.
MOVF INDF,W will move the data in Address = 0x30 into W.
BSF INDF,2 will set bit two of the Byte at Address = 0x30.
3) If we increment the “FSR” register (by doing INCF FSR,F) and do the
examples above, we are now working with the Data at Address = 0x31.
Suppose we want to Clear this Array. Do:
MOVLW 0x30 ; Set-Up the Start of the Array
MOVWF FSR ;--- Start in FSR
MOVLW D’16’ ; Count Out 16 Bytes
MOVWF COUNTER
LOOP:
CLRF INDF ; Clear the Byte at FSR’s Address
INCF FSR,F ; Increment FSR’s Address
DECFSZ COUNTER,F
GOTO LOOP
Figure 5-6 Illustration of Indirect RAM Addressing

48
For this to work in the example program, the RAM locations for the digits (the
segment codes) must be in increasing sequential order as: LEFT_DIGIT, LEFT_DIGIT2,
LEFT_DIGIT3, and RIGHT_DIGIT.
Let’s see how this works in an improved version of the “DISPLAY routine as
follows:
DISPLAY2: ; Call this subroutine once every 5 milliseconds.
CLRF PORTB ; Blank The Display
CLRF PORTC
MOVF DIGIT_INDEX,W ; Get The Current Address
MOVWF FSR ; For Display
MOVF INDF,W ; Get The Digit’s Segment Codes
MOVWF PORTB ; Display the Segments
MOVF DIGIT_POSITION,W ; Get The Digit to Drive
MOVWF PORTC ; Activate the Drive
INCF DIGIT_INDEX,F ; For Next Time, Increment
; Address
BCF STATUS,C ; Shift to Next Digit to Drive
RRF DIGIT_POSITION,F
BTFSS STATUS,C ; Check Limits
RETURN
INITIALIZE_DISPLAY:
MOVLW LEFT_DIGIT ; Put the ADDRESS of LEFT_
; DIGIT into W (Not the DATA)
MOVWF DIGIT_INDEX ; Set-up with LEFT_DIGIT
MOVLW 0x08 ; Set Digit Drive for Left Digit
MOVWF DIGIT_POSITION
RETURN
At the start, the segments and digits were blanked (cleared) to prevent flickering
when a new digit is set-up. Then we get the value in “DIGIT_INDEX”, which is the
RAM address of the digit’s segment codes to be displayed, and put this into the “FSR”
register. Using this RAM address in “FSR”, we get the data at this address (the segment
codes) using the “MOVF INDF,W” instruction and send it to Port B. Next, we get the
“DIGIT_POSITION” and send it to Port C to drive the digit transistor.
This completes the part of the “DISPLAY2” routine which uses indirect
addressing. The rest of the routine must set up the next digit to be displayed when the
routine is called the next time. This introduces two new instructions. The first is,

49
“INCF f,d”, which increments a register-file (RAM byte) and can store the result in the
RAM or in the W register. There is also a, “DECF f,d”, instruction which decrements a
register-file. Both the “INCF” and the “DECF” have the same operation format and both
affect the “Z” (Zero) flag. The “C” and “DC” flags are not affected.
The next new instruction is the, “RRF f,d”, which does a bit-wise rotate-right on
a register-file. There is also a, “RLF f,d”, instruction which does a bit-wise rotate-left.
The actions of these instructions are best seen in Figure 5-7. The only flag bit affected in
both of these instructions is the “C” (Carry) flag.
7
7
6
6
5
5
4
4
3
3
2
2
1
1
0
0
C
C
Flag
Flag
Register-File Byte
Register-File Byte
RLF file,d
RRF file,d
Figure 5-7 Diagram of RLF and RRF Instructions
To set-up the next digit for display, the “DIGIT_INDEX” is incremented to point
to the next sequential RAM address. If that was “LEFT_DIGIT”, its increment would be
“LEFT_DIGIT2”. The “DIGIT_POSITION” is then rotated right, after clearing the carry
bit, to set the next digit-drive bit. The sequence of digit-drive bits on Port C looks like:

50
Bit 3 Bit 2 Bit 1 Bit 0
1 0 0 0 ---- LEFT_DIGIT
0 1 0 0 ---- LEFT_DIGIT2
0 0 1 0 ---- LEFT_DIGIT3
0 0 0 1 ---- RIGHT_DIGIT
If the current digit-drive bit is as (0 0 0 1) and this gets rotated right, the “C” (Carry) flag
will be set (=1). This indicates that the data must be re-initialized and set-up for the
LEFT_DIGIT. The instruction, “MOVLW LEFT_DIGIT”, moves the RAM address
into the W register and this is placed in the “DIGIT_INDEX” RAM byte.
5.7 Odds & Ends
One remaining point in using indirect addressing of RAM is the STATUS register
bit, “IRP” (Bit 7), which selects which banks of RAM to use. Direct addressing of RAM
used the STATUS bits, “RP1” and “RP0”, to select among four banks of seven-bit-
addressable RAM locations. Since indirect addressing uses 8-bit addresses in the “FSR”
register, only one extra bit, “IRP”, must be set-up to select pairs of RAM banks. It is as
follows:
IRP = 0 for Bank 0 and Bank 1
IRP = 1 for Bank 2 and Bank 3.
In the security system, the KEY_SCAN routine is good for when only one key-
press is needed. If the system requires multiple key-presses, such as when entering a
multi-digit number, the software must check for the key to be released before looking for
the next key to be pressed. This will avoid a “string” or “run” of a single key being
interpreted as “several keys”.
Once a key is pressed, released, and recognized as “valid”, it must be decoded
into a number or digit. This may require another look-up table. Suppose that a two-digit,
decimal number is to be entered. The most-significant digit must be multiplied by ten
and added to the least-significant digit to form the complete binary number. Or, suppose,
that a three-digit, decimal number (less than 256) is to be displayed on the LED digits.
How would a binary number be converted to a set of decimal digits?
To multiply a four-bit number by ten you could use a look-up table or you could
multiply it by four, add it to itself to get a “multiply by five”, and then multiply the result
by two, to get a “multiply by ten”. This is as follows:
MOVWF TEMP1 ; Save W
MOVWF TEMP2 ; Save W
BCF STATUS,C ; Prepare to Rotate
RLF TEMP2,F ; Multiply By two

51
RLF TEMP2,W ; Multiply By Two, Again (4*W)
ADDWF TEMP1,F ; Add W to (4*W) to Get (5*W)
BCF STATUS,C ; Prepare to Rotate
RLF TEMP1,W ; Multiply By Two, To Get (10*W) in W
Converting from a binary byte to a three-digit, decimal number can be done with
division or by repeated subtraction. For example, use a counting-loop and count the
number of times that ten or one hundred can be subtracted from the number without
going over. There are two subtraction instructions in the PIC. The first is,
“SUBWF f,d”, which means, “subtract W from the register-file (RAM)” or:
W = (file) - W
Or (file) = (file) - W.
The second is, “SUBLW k”, which means, “Subtract W from the eight-bit value, k, and
put the result in W”. That is:
W = k - W.
In both of these, the STATUS flag bits “Z” (Zero), “C” (Carry), and “DC” (Digit Carry)
are all affected.
In addition to the arithmetic instructions there are six logic instructions as follows:
ANDWF f,d Logical “AND” with register-file
ANDLW k Logical “AND” with eight-bit data
IORWF f,d Logical “OR” (“Inclusive OR”) with register-file
IORLW k Logical “OR” (“Inclusive OR”) with eight-bit data
XORWF f,d Logical “XOR” (“Exclusive OR”) with register-file
XORLW k Logical “XOR” (“Exclusive OR”) with eight-bit data.
Each of these affect only the “Z” (Zero) flag.
The logical “AND” function is used to force zeros (zero-bits) into data bytes.
That is, it is used to mask off unwanted data bits. For example, suppose you wanted the
lowest three bits of the W register and you wished to mask off the rest of the bits (make
them zero). You could say, “ANDLW 0x07”, to retain only the three lowest bits. If W
contained 0xC6, doing the “AND” with 0x07 would give W = 0x06.
The logical “OR” function is used to force ones (one-bits) into data bytes. For
example, suppose you wanted to put ones in the upper four bits of the W register. You
could say, “IORLW 0xF0”. If W contained 0xC3, doing this “OR” would give W =
0xF3.

52
The logical “XOR” function is useful for toggling bits “on” and “off”. This could
be used for making LEDs blink or flash. It is also useful for selectively complementing
bits in memory or in the W register. For example, if you had W = 0x01 and you
“XOR”ed it with 0x03, the result would be W = 0x02. If you “XOR”ed it again with
0x03, you would get W = 0x01. Two successive XORs cancel each other out.
Another logic instruction is, “COMF f,d”, which forms the one’s complement of
a register-file (RAM). The “COMF” instruction only affects the “Z” (Zero) STATUS
flag. To form the one’s complement of the W register, you could use, “XORLW 0xFF”.
The last instruction to study in this chapter is, “SWAPF f,d”, which does a
swapping of the upper and lower four-bits of a register-file (RAM). For example, if the
location “TEMP” contained 0xA9, doing a “SWAPF TEMP,F” would put 0x9A in
“TEMP”. “SWAPF” does not affect any flags.
The PIC has three more instructions which we will cover in Chapter 8. These
instructions will not be covered, now, since they are special and more background is
needed to understand them.
Adding or subtracting multi-byte numbers is more difficult in the PIC since there
are no “add” or “subtract” instructions which include the “C” (Carry) STATUS flag at the
start of the add/subtract. Other processors have an “add-with-carry” and a “subtract-
with-borrow” instructions to make multi-byte arithmetic easier.
Suppose there are two, two-byte numbers to be added to get a two-byte result
(sum). Let these RAM locations be defined as follows:
(IN_1_HIGH IN_1_LOW) (IN_2_HIGH IN_2_LOW)
(SUM_HIGH SUM_LOW) (CARRY_HOLD).
The program to add these two-byte numbers is as follows:
MOVF IN_1_LOW,W ; Load first number (low)
ADDWF IN_2_LOW,W ; Add second number (low)
MOVWF SUM_LOW ; Store Sum (low)
CLRF CARRY_HOLD ; Reset Carry-Hold Byte
BTFSC STATUS,C ; Test If Carry Set
BSF CARRY_HOLD,0 ; Set Carry-Hold if so
MOVF IN_1_HIGH,W ; Load first number (high)
ADDWF IN_2_HIGH,W ; Add second number (high)
ADDWF CARRY_HOLD,W ; Add Carry to High Sum
MOVWF SUM_HIGH ; Store Sum (high)
A double-byte subtraction is similar, but if the “C” (Carry) is set, a “one” is subtracted
from the high-byte of the difference.

53
Another “odds & ends” point that needs to be covered is memory paging in the
program memory. The “CALL” and “GOTO” instructions only have eleven (11) address
bits, which gives a range of destinations only in a 2048 (2K) block of program memory.
The PIC16F877 has a full program memory space of 8192 (8K). How can we get beyond
the 2K limit?
The answer is to directly set or clear the upper-most bits in the high-half of the
program-counter, “PCLATH”, using the “BCF” and “BSF” instructions, prior to doing a
“CALL” or “GOTO” instruction. These bits do not take effect immediately upon doing
the “BCF” or “BSF” but are delayed or postponed until after the PIC gets the “CALL” or
“GOTO” instruction to go to a higher memory beyond the 2K limit. In the case of a
“CALL” instruction, the corresponding “RETURN” does not need to adjust the
“PCLATH” bits since the full return-address is saved (Just do a “RETURN”).
For example, consider the following program segment:
ORG 0x0100
BSF PCLATH,4 ; Set Most Significant ADDR Bit
BSF PCLATH,3 ; Set 2nd
Most Significant ADDR Bit
CALL HIGH_SUB
ORG 0x1827
HIGH_SUB:
RETURN
This would call the “HIGH_SUB” subroutine in the upper 2K block of the 8K program
memory.
Just as there is the “BANKSEL” directive, there is also the “PAGESEL”
directive. The “PAGESEL” directive simplifies the set-up of the PCLATH register prior
to doing a GOTO or a CALL. Its syntax is, “PAGESEL program-address”. For
example, the above CALL to “HIGH_SUB” could be done more easily as:
PAGESEL HIGH_SUB
CALL HIGH_SUB

54
5.8 Using KEY_SCAN and DISPLAY Together
Let’s see how KEY_SCAN and DISPLAY can be used together so that the user’s
keystrokes will appear in the LED digit display.
First, assume that the KEY_SCAN routine has been modified to de-bounce the
keys and allow only one key at a time to be recognized. This will be shown in principle
with the techniques of Chapter 6. Also, assume that a RAM location called
“SKIP_CODE” has been filled with either 0x01 or 0xFF when the KEY_SCAN routine
returns. As before, assume that the KEY_SCAN routine returns with the key-code
number in the W register.
Let the following instructions be executed every five milliseconds and in this
order:
CALL KEY_SCAN
DECFSZ SKIP_CODE,F
CALL PROCESS_THE_KEY
CALL DISPLAY
If the “SKIP_CODE” RAM byte contained 0xFF, then the “DECFSZ” would not skip
and the “PROCESS_THE_KEY” routine would be called. The KEY_SCAN routine
must do this only when there is a valid key-press. Otherwise, a value of 0x01 must be
used in “SKIP_CODE” to skip over the “PROCESS_THE_KEY” routine when no key is
pressed.
Since the DISPLAY routine displays only whatever is in the display-RAM bytes,
the job of the “PROCESS_THE_KEY” routine would save the raw key-code, shift the
display-RAM bytes to the left by one digit, convert the raw key-code to a display-code,
and then fill in the new display-code on the right-hand digit.
This would be done as follows:
PROCESS_THE_KEY:
MOVWF SAVE_KEY ; Save the raw key-code
BSF USER_FLAGS,KEY_IS_READY ; Tell Program “Key”
MOVF LEFT_DIGIT2,W ; Shift Over One Digit to the Left
MOVWF LEFT_DIGIT
MOVF LEFT_DIGIT3,W
MOVWF LEFT_DIGIT2
MOVF RIGHT_DIGIT,W
MOVWF LEFT_DIGIT3

55
MOVF SAVE_KEY,W
SUBLW D’10’ ; Check if = “*”
BTFSS STATUS,Z
GOTO NOT_STAR
CLRF RIGHT_DIGIT ; Put Blanks in Right-Digit on “*”
RETURN
NOT_STAR:
MOVF SAVE_KEY,W
SUBLW D’12’ ; Check if = “#”
BTFSS STATUS,Z
GOTO NOT_POUND
CLRF RIGHT_DIGIT ; Put Blanks in Right-Digit on “#”
RETURN
NOT_POUND:
MOVF SAVE_KEY,W
SUBLW D’11’ ; Check if = “0”
BTFSS STATUS,Z
GOTO NOT_ZERO
CLRW
MOVWF RIGHT_DIGIT
RETURN
NOT_ZERO:
MOVF SAVE_KEY,W
MOVWF RIGHT_DIGIT
RETURN
(This program will do the trick to set-up the right display codes but it is not the
best way to do it. Can you think of a better way?).

56
5.9 A Last Look at the Advanced Security System
The main body of the security system code would look like this:
LIST P=16F877
ORG 0x0000
CALL INITIALIZE
MAIN_LOOP:
CALL DELAY_FIVE_MILLISECONDS
CALL KEY_SCAN
DECFSZ SKIP_CODE,F
CALL PROCESS_THE_KEY
CALL DISPLAY
CALL SCROLL_MENU
CALL MENU_MODE
GOTO MAIN_LOOP
MENU_MODE:
MOVF MODE_CODE,W
ANDLW 0x07 ; Restrict to eight values
ADDWF PCL,F ; Start of “JUMP” Table
GOTO RESET
GOTO IDLE
GOTO SET_ZONES
GOTO ARM_HOME_AWAY
GOTO SET_ENTRY_CODES
GOTO CHECK_FOR_ACTIVE_ALARMS
GOTO SET_ALARM_CALL_POLICE
GOTO READY_TO_ARM
------- Body of Code ------------------------
END
Most of the main-loop code runs the keypad and the display. The major modes of
operation of the security system are controlled by the “MENU_MODE” routine. The
configuration of a table look-up that contains only “GOTO” instructions is very common
in assembly language programs. It is called a “Jump-Table”. The RAM byte
“MODE_CODE” selects what part of the program is active at any one time. Since there
are only eight (8) codes that are used, as (0,1,2,…,7), the “ANDLW 0x07” instruction
will restrict any code to this range. This is done for safety reasons so as not to over-index
the table look-up.

57
Each section of code in the various “modes” would have its own messages and
prompts for the user’s inputs. There must be very little delay in each of the modes and
they must all return to the main-loop. The techniques in Chapter 6 will show you how to
do this.
This method of design keeps the body of the code short and simple. It is easy to
understand both conceptually and practically.
1) The STATUS register contains the RP0 and RP1 bits which select the RAM
banks in the direct addressing mode and the IRP bit which selects the RAM
banks in the indirect addressing mode. It also contains the “Z”, “DC”, and
“C” flags, which are set-up as the results of arithmetic/logic instructions and
the “MOVF” instruction.
2) The BANKSEL directive manipulates the RP0 and RP1 bits of the STATUS
register.
3) The “ADDWF PCL,F” instruction is a “GOTO”-like instruction and is the
basis of all table look-ups, including “Jump-Tables”. The table-data must fit
squarely within a 256-word block of program memory.
4) Indirect addressing of RAM allows for the manipulation of data arrays. The
RAM address is placed in the “FSR” register and the instruction which is to
use indirect addressing must reference the memory pointed-to by the “FSR”
by using a reference to the “INDF” register.
5) The “CALL” and “GOTO” instructions are limited to a 2 K block of program
memory, in themselves. Bits “4” and “3” of the “PCLATH” register must be
manipulated before doing a “CALL” or “GOTO” whose target address is
beyond the current 2 K block. These bits do not take immediate effect when
they are set-up but are postponed until after the “CALL” or “GOTO” is
executed. The “PAGESEL” directive simplifies this process.
6) A “RETURN” or a “RETLW” instruction does not need to manipulate the
“PCLATH” register bits prior to returning from the subroutine.
7) An “Add” or “Subtract” instruction will generate a carry-bit (“C”) but there is
no way to include the carry-bit at the start of the “Add/Subtract” instruction.
8) If a CALL is made to an “ADDWF PCL,F” table look-up subroutine which is
not in the same 256-word block of memory as the “CALL”, the high-half of
the subroutine’s address must be loaded into the PCLATH register before the
“CALL” is made. (See pages 45 and 46.)
9) The following instructions were introduced in Chapter 5:
a) RETLW data
b) ADDLW data
c) ADDWF file,destination
d) CLRF file

58
e) CLRW
f) INCF file,destination
g) DECF file,destination
h) RRF file,destination
i) RLF file,destination
j) SUBLW data
k) SUBWF file,destination
l) ANDWF file,destination
m) ANDLW data
n) IORLW data
o) IORWF file,destination
p) XORLW data
q) XORWF file,destination
r) COMF file,destination
s) SWAPF file,destination
10) Three remaining instructions have not yet been introduced.
This concludes Chapter 5. A complete list of all of the PIC16F877’s instructions
and how to use them is contained in Appendix A.

59
Chapter 6: Fundamental ESP Techniques
6.0 Chapter Summary
Section 6.2 covers software readability. Section 6.3 covers software
maintainability. Section 6.4 discusses the most general fundamentals of embedded
systems software. Section 6.5 discusses the background routine. Section 6.6 covers the
theory of the watch-dog timer. Section 6.7 discusses event driven software. Section 6.8
covers the theory of processor interrupts. Section 6.9 discusses slow inputs and outputs.
Section 6.10 covers software time measurement techniques. Section 6.11 discusses
hashing techniques. Section 6.12 covers the software methods of waveform encoding.
Section 6.13 covers waveform decoding techniques. Section 6.14 discusses the
fundamental tradeoffs of time, program memory, and RAM and how these affect program
size and execution speed. Section 6.15 discusses “ROM states” which are useful for high
speed operation. Section 6.16 discusses the limitations of C/C++ in low end embedded
systems programming.
6.1 Introduction
Programming for embedded applications involves more than just programming in
the traditional sense. Embedded systems programming embodies total systems design.
The programming part requires an active understanding of the electronics and mechanics
of the system.
Even so, programming for embedded systems has a style, technique, art, and
science of its own. It is always true that each individual engineer and programmer has a
style of programming that is uniquely his or hers, but there are sets of common practices
in the industry which unify the goals of embedded systems programming in general. This
chapter will give guidelines and the common practices for embedded systems
programming in a general sense.
6.2 Software Readability
It is generally true that all programs need sound documentation but this is
especially true for embedded systems programming since there is such a strong
connection to the hardware. Often the documentation should explain the hardware
connections that parallel the software constructs. Such things as gates, memories, LEDs,
switches, motors, and other devices have wiring conventions of their own which become
important when they are interfaced to the PIC. The documentation should clearly state
how these devices are to be used and which parts of the PIC will control them.

60
Other aspects of documentation are similar to traditional programming in that it
should clearly define the software operations that are used and fully explain the
programmer’s intent in using combinations of instructions to accomplish a given task.
Software readability should give another programmer a solid idea of what the
program is about, without ego getting in the way. This is another aspect of writing with
the intent of being understood. This is also important when you, as the programmer,
leave the program alone for a few years and then come back to it when you need it. Your
documentation should give you a clear picture of your program even after you are away
from it for years. You should state all of the “hidden” secrets in the documentation
without trying to carry them in your head. The documentation should be complete in
every way.
6.3 Software Maintainability
Maintainability is the art of building complex software systems out of simple
components that are easy to modify and make the software easily adaptable to new, but
similar, applications. Products and systems rarely “stand alone” after they are created
and are dynamic. New markets and new applications will come into being and the
software systems must meet these future uses without having to be completely
redesigned. Maintainability is the art of designing for the future as well as the present.
6.4 Software Fundamentals
Embedded software should be modular and should have a simple “main-loop”.
The main-loop should be “small” or at most two pages long. It should call subroutines as
the functional, modular blocks to perform each major task. Each task should be free of
overhead and its subroutine should do its tasks in as simple a way as possible. The
security system software example at the end of Chapter 5 is a good example of a simple
main-loop and a simple and maintainable software structure.
Subroutines should also be “small” if possible. They should have multiple entry-
points and serve multiple purposes. This is contrary to high-level programming practices
which emphasize “one entry-point, one function” subroutines. It may be necessary, due
to memory constraints, to squeeze as many functions as possible into the smallest space.
Subroutines that work as modules must be careful when the subroutines they call
take too long to run. There must be no or, very little, time delays or bottlenecks that
would interfere with the timely operation of the system. Generally, it is wise to avoid
recursive subroutine calls.
Another good programming practice is “information-hiding”. Subroutines should
be classed into layers of low-level, medium-level, and high-level functions. This does
not mean that they should be nested this way. It means that information should be

61
processed in separate layers. For example, in a security system, a low-level function
would read the raw data, such as, switch closures and voltages, set LEDs and other bit-
level outputs and put the results in RAM. A medium-level function would set-up alarm
conditions from the low-level data and put its results in RAM. The high-level functions
would, then, perform the “command-and-control” tasks such as setting and reporting an
alarm. Since each level puts its intermediate data into RAM, there is no need to nest
these routines. Information and processes can be “hidden” from the upper levels and
make the overall program more compartmentalized and make it more maintainable.
6.5 The Background Routine
One large subroutine should be called from the main-loop to do housekeeping
tasks. This is called the “background” routine. It consists of common tasks which can
remain invisible to the rest of the program, such as keeping a time-of-day clock and
calendar, refreshing port settings, servicing slow inputs and outputs, sampling a keypad,
running an LED digit display, and, in the case of a robot, running stepper motors.
6.6 The Watch-Dog Timer
Microprocessor and microcontroller systems are not as stable as other digital
hardware due to the fact that they are more complex and they work with flexible
software. It is easy for such a system to get “locked-up” in infinite software loops due to
a noise spike or unanticipated data. As a programmer, you try to account for all of the
possible input data that can come into your system and all of the possible results of the
calculations on that data, but in reality, you can’t. An unexpected situation may arise and
put the processor into an infinite software loop.
The watch-dog timer is typically a hardware device external to the processor that
has the power to reset the processor over-and over if the watch-dog timer itself is not kept
in reset by the processor’s software. Its job is to break the software out of any infinite
software loops if they should occur. The software must be active all of the time and it
must not have any substantial delays. It must reset the watch-dog timer on a regular basis
and, if it doesn’t, the watch-dog will keep resetting the processor over-and-over.
The PIC16F877 has a built-in watch-dog timer which is built into the chip, but it
operates or can operate independently from the other PIC hardware. It can be enabled
and disabled from software with an adjustable time-out time, and the PIC has a special
instruction, “CLRWDT”, which resets the watch-dog timer. The specific details of its
use will be covered in Chapter 8 (PIC Peripherals).
Other processors may not have a built-in watch-dog timer and it may be necessary
to get a watch-dog timer chip to use in the system. An industrial product cannot do
without one! It is a MUST! One watch-dog timer chip available from Maxim is the
MAX690CPA. It is also possible to make your own watch-dog timer from all-digital

62
gates as shown in Figure 6-1. The microprocessor’s port-pin drives the clock input of a
“one-and-only-one” pulse circuit which in turn drives the counter’s reset line. If there are
no watch-dog resets, the counter will reset the microprocessor. The clock input of the
pulse circuit is an edge-triggered input. This is critical. The watch-dog timer must not be
reset with a level signal so that it cannot be accidentally left in “reset”. The software
must always toggle the port-pin line.
D Q
Q
CK
Set
Reset
1
D Q
Q
CK
Set
Reset
1
1
8-Bit Counter (74HC393)
ResetClock
NAND
1
To uP/uC Reset
74HC04
Schmidt-Trigger
NAND Gate
74HC14
R C
250 Hz
Clock
Q0 Q1 Q7
From uP/uC Port
(Send Edge To
Reset the Watch-Dog Timer)
Dual D-FF
74HC74
Figure 6-1 "All Digital" Watch-Dog Timer Circuit
The command or command sequence to reset the watch-dog timer should be
placed once-and-only-once in a program. It should be placed in the main-loop since the
software must always come back to the main-loop. The time-out time of the watch-dog
timer should be about one second for most applications. A good rule of thumb for the
watch-dog’s period is about ten times the longest software delay in the main-loop.
6.7 Event-Driven Software
Event-driven software is the central unifying concept in embedded systems
programming. It is the main idea which separates ESP from traditional programming.
Event-driven software uses flags and counters in RAM to mark the progress of an
input. When there are changes to the input, the flags and counters are updated to reflect

63
these changes. Event-driven software can control many delays and many arbitrary
processes at once without waiting in loops. There are no bottlenecks in event-driven
code.
An example program will show the structure and function of event-driven code.
This example is for de-bouncing a push-button switch. This software routine looks for
the contact and the release of the switch before declaring that the data is valid. The
flowchart for this process is shown in Figure 6-2.
Push
Button
Return
Return
Return
Return
Return
Reset
Counter
Reset
Counter
Reset
Counter Reset
Counter
Counter
Increment
Counter
Increment
Is
Button
Sensed as
"DOWN"
?
Is
Button
Sensed as
"DOWN"
?
Is
Counter
> or = Limit
?
Is
Counter
> or = Limit
?
Is
"Pressed"
Condition
Recognized
?
Set Flag:
Set Flag:
"Pressed" "Pressed"
Reset Flag:
"Valid Button"
Figure 6-2 Event-Driven Push-Button Switch DeBouncing
Y Y
Y
Y
Y
NN
N
N
N
Assume that the following routine is called every five milliseconds at a time:
PUSH_BUTTON:
BTFSC FLAGS,PRESSED ; Test if the button is de-bounced
; AND “ON” (Pushed)
GOTO WAIT_FOR_RELEASE
BTFSC PORTB,PUSH_BTN ; Test the raw button state
GOTO WAIT_FOR_SET ; = Pushed (“ON”)
GOTO RESET_COUNTER ; = “OFF”, Reset De-Bounce
; Timer

64
WAIT_FOR_SET:
DECFSZ WAIT_COUNTER,F; Delay for De-Bounce
RETURN
BSF FLAGS,PRESSED ; Set “De-Bounced, ON”
GOTO RESET_COUNTER
WAIT_FOR_RELEASE:
BTFSC PORTB,PUSH_BTN ; Test raw button state
GOTO RESET_COUNTER ; = Pushed (“ON”)
DECFSZ WAIT_COUNTER,F; = “OFF”, wait for De-Bounce
RETURN
BCF FLAGS,PRESSED ; De-Bounce Cycle Done!
BSF FLAGS,DATA_READY ; Done! Data is Ready!
RESET_COUNTER:
MOVLW WAIT_TIME ; Fill Initial Value of Counter
MOVWF WAIT_COUNTER
RETURN
At the start of the routine, the software checks if the push-button has been pressed
and has passed the de-bouncing in the “ON” state. If it has, the software goes to look for
the release of the push-button. If not, the raw state of the switch is checked. If it is
pressed (“ON” ), the counter loop is run to satisfy the de-bouncing condition. If it is not
pressed, the counter is reset. If the “ON” state de-bouncing is satisfied, the “PRESSED”
flag is set, the counter is reset, and the routine will then look for the de-bouncing in the
released state. Once the release is complete, the “DATA_READY” flag is set to register
a complete “press and release” of the push-button.
At no time does this routine do any waiting. It is called, it checks for conditions,
it performs its actions, and in each case, it exits.
The time the routine takes to run is negligible and many similar routines can be
run with the illusion of being simultaneous.
Event-driven software can also be used with internal events, such as for flashing
LEDs and for generating waveforms.
An application that is event-driven will spend 99 percent of its time doing
nothing! It acts only on the change of internal and external events and has no waiting
loops. Event-driven techniques make working with the watch-dog timer easy and the
resulting program is very maintainable!

65
6.8 Interrupts
Interrupts are a hardware process whereby a piece of hardware can cause the
processor to execute a subroutine at a special address. The process makes the CPU drop
whatever it is doing at the time and run the subroutine. The PIC16F877 has a total of 14
interrupts related to its peripheral functions. The mechanics of how to use them will be
shown in Chapter 8 (PIC Peripherals).
Many processors (including the PIC) allow the interrupts to be “enabled” and
“disabled” under software control. In some processors, there are interrupts which cannot
be disabled and will respond “knee-jerk” fashion at all times. Interrupts must be used
with extreme care and can cause many severe software/hardware problems if they are
abused.
Why are interrupts used? An interrupt is the fastest way to get the processor’s
attention to work on an urgent problem. It should be used only in this case. If a timer or
counter is set to run as a time-base for the system (the PIC can also do this), an interrupt
is an ideal way to respond to the timer/counter and re-initialize it. If there is a peripheral
that must be serviced very quickly and there is no other way to do it, then an interrupt
must be used.
I regard interrupts as “the method of last resort” in any programming situation.
The potential for abuse when using interrupts is very severe. The problems they can
cause can be extremely difficult to track-down and solve. If you can avoid using
interrupts, I strongly advise that you do not use them. Event-driven techniques are far
easier to work with in every way.
One particularly nasty abuse of interrupts is to use them for keypads and switch
de-bouncing. This is the most inappropriate use of interrupts. There is no way that a key
needs an interrupt to process it, since it works at millisecond speeds and a human key
operator is much slower.
Interrupts and their service-subroutines should NEVER be used to reset the watch-
dog timer. The interrupt is a “knee-jerk” response and a subroutine. If the service-
subroutine is called from an infinity loop, it will return to that infinity loop upon its return
from the interrupt, thereby defeating the whole purpose of having the watch-dog timer in
the first place!
6.9 Slow Inputs and Outputs
As a general rule of thumb, an input or an output is “slow” if changes to or from it
occur in one millisecond or more time. Such things as DIP switches, push-button
switches, telephone-ringing sensors, LEDs, multiplexed LED digits, relays, motors,
temperature sensors, and heating elements are considered “slow” I/O.

66
Slow inputs and outputs should be sampled in one, and only one, place in the
software, and, preferably, should be done in the background routine.
All slow inputs should be buffered, processed, and sorted by special routines
before their information is made available to the rest of the program. All slow outputs
should do the same before being presented to the outside world.
There are several reasons for doing this. If there were design changes to the
hardware where the input or output port-pins are swapped, only these special routines
would have to be changed instead of the whole program! These routines could de-
scramble the inputs and outputs so that they are in uniform order (e.g., Input-0 = Bit-0,
Input-1 = Bit-1, …). Some inputs and outputs may be active-low or active-high. These
routines can convert every input and output to an “active-high” state. This is another
example of layered software and information hiding.
After the slow inputs are buffered, processed, sorted, and made uniform, the
results should be put into RAM. Any time that the program needs the information, it can
check these RAM bytes. The program should never check the bit directly at the port.
The slow outputs, in a similar way, should be buffered, processed, sorted, and made
uniform from a RAM location when it is to be sent to the output port. Changing a bit in
these RAM locations is equivalent to sending it out to a port. Only the background
routine and the special input/output routines should interface with the port directly.
6.10 Software Time Measurement
Although the PIC and other processors have hardware for measuring time-delays
and pulse-widths, it is still useful to have software techniques for doing these things, in
that they can be made more fault-tolerant than the hardware versions.
A simple software time-measurement loop is as follows:
TIME_MEASURE:
CLRW ; Reset Time-Measure Counter
TIME_LOOP:
BTFSC PORTB,EVENT_SENSE ; Look For The Event
RETURN
MOVWF TEMP
INCF TEMP,W ; Increment (W) Time-Measure Counter
BTFSS STATUS,Z ; Check for End of Loop
GOTO TIME_LOOP
RETLW 0xFF
The time-measurement is returned in the W register. The total loop-time in this routine is
seven instruction cycles.
A faster and higher resolution software time-measurement process is as follows:

67
TIME_MEASURE:
BTFSC PORTB,EVENT_SENSE ; Look for Event
RETLW 0
BTFSC PORTB,EVENT_SENSE
RETLW 1
BTFSC PORTB,EVENT_SENSE
RETLW 2
---- And So On For The Desired Time Length -----
This “non-loop” has a resolution of only two instruction cycles instead of seven in the
first loop. It takes up much more space, but it runs very quickly and has very high time
resolution.
6.11 Hashing
Hashing is a table-search technique that uses the input data directly as the address
or index of the table item to find. Actually, we have already seen hashing in action
through the look-up tables in Chapter 5. It is not only fast, but it also runs in uniform-
time (i.e., the time it takes to find a table entry is the same for each table entry). The
security system at the end of Chapter 5 also introduced us to “jump tables”. This is a
very important concept since it gives you the ability to “GOTO” a variable address in
program memory.
Conversion tables for converting from ASCII to seven-segment codes are very
common in all programs. Also, rapid scrambling or de-scrambling of data can be done in
these tables. The key is to use the data as the index of the table. Whole waveforms can
be stored in these tables for the production of music and other sounds.
An improved jump table is as follows:
DO_JUMPS:
MOVWF TEMP ; Multiply (W) by three
BCF STATUS,C
RLF TEMP,F
ADDWF TEMP,W
ADDWF PCL,F ; Index the Jump Table
BCF PCLATH,4 ; (We could use PAGESEL here)
BSF PCLATH,3
GOTO SOME_HIGH_JUMP1
BSF PCLATH,4
BSF PCLATH,3
GOTO SOME_HIGH_JUMP2
----- And So On For More Long Jumps -----

68
This allows you to do “GOTO”s to any memory location.
6.12 Waveform Encoding
In some applications a microprocessor or a microcontroller may need to be used
as a modem and must generate a coded waveform. This is an excellent example of where
software can be used to mimic hardware. If the waveform is a “burst”, or, short packet of
a bit-stream, the routine to do it can be a short loop. If the waveform is to be a
continuous, running stream of data, the routine must use event-driven techniques.
For example, consider a Frequency-Shift-Keying (FSK) burst encoder that uses
two frequencies as 2400 Hz and 2250Hz for logic “1” and logic “0”, respectively.
Assume that the data rate is 150 Baud and that the data structure is: A preamble of ten
“1”s, a “0”, twenty-five (25) data bits, and an end bit, for a total of 37 bits to send. In this
timing a “1” will be sent as 16 cycles of 2400 Hz and a “0” will be sent as 15 cycles of
2250 Hz. This time delay between bits is not critical but the frequency stability is critical
and is set as a half-cycle as a “delay then complement output line state”. This gives 32
half-cycles of 2400 Hz and 30 half-cycles of 2250 Hz.
The following routine will do the above FSK:
FSK:
MOVLW DATA_BIT_ARRAY ; Get Address of Data
MOVWF FSR ; Set-Up Indirect ADDR
SEND_2400:
MOVLW 32 ; Do 32 Half-cycles of 2400 Hz
MOVWF COUNT
LOOP_2400:
CALL DELAY_SEND_2400 ; Do Half-Cycle
DECFSZ COUNT,F
GOTO LOOP_2400
MOVF INDF,W ; Get Next Bit To Send
INCF FSR,F ; Point To Next Bit To Send
ADDWF PCL,F ; Look-Up Table: Do Next Bit
GOTO SEND_2250 ; 0 = Send Zero Bit
GOTO SEND_2400 ; 1 = Send One Bit
RETURN ; 2 = DONE, Return
SEND_2250:
MOVLW 30 ; Do 30 Half-Cycles of 2250 Hz
MOVWF COUNT
LOOP_2250:
CALL DELAY_SEND_2250 ; Do Half-Cycle
DECFSZ COUNT,F

69
GOTO LOOP_2250
MOVF INDF,W ; Get Next Bit to Send
INCF FSR,F ; Point To Next Bit To Send
ADDWF PCL,F ; Look-Up Table: Do Next Bit
GOTO SEND_2250 ; 0 = Send Zero Bit
GOTO SEND_2400 ; 1 = Send One Bit
The Data Array would contain the sequence of bits starting from the preamble,
and going to the end-bit. After the end-bit would be a code of “2” or “3” to stop and
return.
Another burst-type data transmission technique is the Manchester code. This is
best seen in Figure 6-3. This works as a kind of “differential phase-shift keying” (DPSK)
on one cycle of a square wave. It is a polar waveform and it can only be used where the
transmission medium can support edges and pulse-widths without distortion. Its timing
in all of its parts is very critical and only a strict ratio of pulse widths as 2:1 are allowed.
Assume for the following example that the data structure is: A preamble of ten “1”s, a
“0”, and sixteen (16) data bits. The Manchester Code waveform routine is as follows:
TEMP: EQU 0x20
FLAGS: EQU 0x21
DATA_ARRAY: EQU 0x30
OUTPUT_BIT: EQU 0 ; PORTC, Bit 0
SAMPLE_FLAG: EQU 1 ; Mark Half-Cycle Where To Sample
; Data
INHIBIT_FLAG: EQU 0 ; If Set, Inhibit State-Complement
MANCHESTER_WAVE:
BCF PORTC,OUTPUT_BIT ; Reset Output Bit
BSF FLAGS,SAMPLE_FLAG ; Sample on Odd Cycles
BCF FLAGS,INHIBIT_FLAG ; Assume For Start,
; Preamble
MOVLW DATA_ARRAY
MOVWF FSR ; Indirect ADDR
MANCHESTER_LOOP:
CALL DELAY ; Delay For Half-Cycle
CLRF TEMP ; Set-Up TEMP with
BCF TEMP,OUTPUT_BIT ; Complement of INHIBIT
BTFSS FLAGS,INHIBIT_FLAG ; In Position of
BSF TEMP,OUTPUT_BIT ; Output Bit of PORTC
BCF FLAGS,INHIBIT_FLAG

70
MOVF TEMP,W ; Then, XOR This Bit
XORWF PORTC,F ; With Output Bit
BTFSS FLAGS,SAMPLE_FLAG ; See If Sample
GOTO DELAY_COMPENSATION; No, Do Delay Comp
BCF FLAGS,SAMPLE_FLAG ; Do Sample Now
MOVF INDF,W ; Get Data To Send
BCF STATUS,C
MOVWF TEMP ; Multiply By Two
ADDWF TEMP,W
ADDWF PCL,F ; Look-Up Table
BCF FLAGS,INHIBIT_FLAG ; 0 = Send “Same” Data
GOTO MANCHESTER_LOOP
BSF FLAGS,INHIBIT_FLAG ; 1 = Change Phase, Data
RETURN ; 2 = 3 = DONE, Return
RETURN
RETURN
RETURN
DELAY_COMPENSATION:
BCF FLAGS,INHIBIT_FLAG ; No Phase Change
BSF FLAGS,SAMPLE_FLAG ; Sample Next Time
NOP
NOP
NOP
NOP
Whenever there is a change of data, the INHIBIT flag performs the phase-change
by inhibiting the state-change on Port C, Bit zero.
Also notice that both the FSK and the Manchester encoders use one RAM byte
per bit to send which is wasteful of RAM. Usually single bits in a small number of RAM
bytes are used for the bits to send. Can you think of some schemes to use indirect
addressing and use single bits in each byte? It could be “destructive” in that the byte got
rotated and filled with zeros or it could be “non-destructive” and just sense the bits as
they are. Try to think of several ways to do this.

71
1 1 1 10 0 0
Pulse-Widths are of a Uniform Square Wave (Ratio 2:1)
Rising-Edges are Logic One (=1)
Falling-Edges are Logic Zero (=0)
Only Where Shown
Transitions Occur as a Double-Pulse-Width
Figure 6-3 Manchester Code Waveform
6.13 Waveform Decoding
Waveform decoding is the process of recovering data from an encoded,
modulated waveform. The FSK signal in the previous section can be decoded with a
phase-locked loop (PLL), where the control-voltage feeds the PIC’s analog-to-digital
converter input. Samples on the ADC are taken several times over the space of one bit to
eliminate noise.
The Manchester Coded waveform can be decoded by following the highs and
lows of the digital waveform. This requires getting into synchronization with the wave,
measuring the pulse-widths, counting the pulses, and recovering the data by judging the
phase-reversals. This process is shown in detail in Figure 6-4.

72
1 1
1 1
1
1
1
1
10 0
0
0 0
Proper Waveform
Improper Waveform
Improper Waveform
Illegal
Illegal
Transition
Transition
Must Recognize Only Two Valid Pulse Widths in 2:1 Ratio
Must Also Recognize Any Illegal Transitions
Figure 6-4 Decoding the Manchester Waveform
Both the FSK and Manchester data recovery processes must use hashing to place
the data bits that are decoded.
A detailed example program of the Manchester decoding process is as follows:
MANCHESTER_DECODE:
CALL TIME_LOW ; Start of Synchronization
ADDLW 0 ; See If W = Zero
BTFSS STATUS,Z
GOTO DO_HIGH_MEAS ; Was In LOW, Look At Full HIGH
CALL TIME_HIGH ; Was In HIGH, Complete HIGH
CALL TIME_LOW ; Do Full LOW
DO_HIGH_MEAS:
CALL TIME_HIGH ; Measure HIGH Pulse W = Count
MOVWF TIME
CALL TIME_LOW ; Measure LOW Pulse W = Count
ADDWF TIME,F ; Average These Counts
RRF TIME,W
MOVWF SHORT_BASE ; Save SHORT Pulse Width

73
ADDWF SHORT_BASE,W ; Mult By Two
MOVWF LONG_BASE ; Save as LONG Pulse Width
MOVF SHORT_BASE,W ; Form Threshold =
ADDWF LONG_BASE,W ; Average(SHORT,LONG)
MOVWF THRESH
RRF THRESH,F
MOVLW NINE ; Track The Preamble Pulses (9)
MOVWF PREAMBLE_CNT
PREAMBLE_LOOP:
CALL SORT_TIMES ; Classify Count
ADDWF PCL,F
GOTO STILL_PREAMBLE ; 0 = SHORT, Still in PREAMBLE
GOTO GET_THE_DATA ; 1 = LONG, Start of Data
RETLW 2 ; 2 = ERROR in Timing
STILL_PREAMBLE:
ADDWF PCL,F
GOTO DEC_PREAMBLE ; 0 = SHORT, Do Loop DEC
RETLW 1 ; 1 = LONG, Polarity ERROR
DEC_PREAMBLE:
DECFSZ PREAMBLE_CNT,F; Decrement Counter
GOTO PREAMBLE_LOOP
RETLW 3 ; ERROR Time-Out
GET_THE_DATA:
MOVLW N_DATA_BITS ; Count Number Of Data Bits
MOVWF BIT_COUNT
IN_A_ZERO:
ADDWF PCL,F
GOTO STILL_IN_ZERO ; 0 = SHORT, Still In a ZERO
GOTO TRANSIT_TO_ONE; 1 = LONG, Go To ONE LOOP
STILL_IN_ZERO:
CALL PLACE_ZERO_BIT ; Hash Store Data Bit = ZERO

74
ADDWF PCL,F
GOTO DO_DEC_ZERO ; 0 = SHORT, Dec Zero Loop
RETLW 1 ; 1 = LONG, ERROR Pulse
TRANSIT_TO_ZERO:
CALL PLACE_ZERO_BIT ; Hash Store Data Bit = ZERO
DO_DEC_ZERO:
DECFSZ BIT_COUNT,F ; Decrement Bit Counter
GOTO IN_A_ZERO
RETLW 0 ; DONE, OK
IN_A_ONE:
ADDWF PCL,F
GOTO STILL_IN_ONE ; 0 = SHORT, Still in a ONE
GOTO TRANSIT_TO_ZERO ; 1 = LONG, Data is a ZERO
STILL_IN_ONE:
CALL PLACE_ONE_BIT ; Hash Store Data Bit = ONE
ADDWF PCL,F
GOTO DO_DEC_ONE ; 0 = SHORT, Do Dec Loop
RETLW 1 ; 1 = LONG, ERROR Pulse
TRANSIT_TO_ONE:
CALL PLACE_ONE_BIT ; Hash Store Data Bit = ONE
DO_DEC_ONE:
DECFSZ BIT_COUNT,F ; Decrement Bit Counter
GOTO IN_A_ONE
RETLW 0 ; DONE, OK
This routine calls several subroutines to measure time, classify the times, and
place the data bits into RAM. The time measuring routines, TIME_LOW and
TIME_HIGH, work just like the seven-cycle loop of the “Time Measurement” section.
The “TIME_LOW” routine returns when the input is sensed “high”, while the
“TIME_HIGH” routine returns when the input is sensed “low”. The data placement
routines work by hashing the bit counter. That is, the bit counter is multiplied by two and
looked-up in a table with “BCF”s or “BSF”s and a “RETURN”. The time classifier
compares the time against the threshold and then tests the time as “Short” or “Long” to

75
within an absolute value difference of two using the “SORT_TIMES” routine. Then it
returns with codes for, “Short”, “Long”, and “Error”.
6.14 RAM, ROM, and Time Tradeoffs
There are three major tradeoffs of memory and time in embedded systems
programming. These are:
1) “RAM vs. ROM”
2) “ROM vs. Time”
3) “Run-In-ROM vs. Run-In-RAM”
In the “RAM vs. ROM” tradeoff, the usage of RAM and ROM tends to vary inversely.
For example, a calculation subroutine may use more RAM to figure intermediate values
or it may use more ROM in the form of a look-up table. Event-driven programs tend to
be RAM-intensive since they use flags and counters to track input and output signals.
In the “ROM vs. Time” tradeoff, the size of the ROM code varies inversely with
the code-speed execution. For example, the short-loop time measurement scheme used
only seven cycles of time and eight ROM words. The “non-loop” scheme could use up to
512 ROM words but had a resolution of only two cycles. The short-loop could measure
times as long as 1.75 milliseconds but the “non-loop” could measure only half a
millisecond.
In the “Run-In-ROM vs. Run-In-RAM” tradeoff, a processor that has the
capability of running its programs from RAM has a speed and versatility advantage over
processors that only run their programs in ROM. For example, if a high-speed waveform
is to be sent out, the processor that only runs in ROM may not be fast enough to run it. If
a processor that can run its programs in RAM uses the ROM program to synthesize a
short subroutine in RAM and call it from ROM, it can run at the highest possible speeds!
The PIC cannot do this exactly, but it does have field-programmable data and program
memories that can exploit some of these tricks. (These will be shown in Chapter 8.)
6.15 ROM States
The idea of a “ROM State” is to make a program that “remembers” previous data
values by being in different parts of the program in the program memory or ROM. That
is, where you are in the program is a reflection of what the previous data was. This
technique is used only for “high-speed” or “high-efficiency” applications since “ROM
State” subroutines get very large for complex tasks.
An example application which uses ROM States for its most efficient coding is a
“median filter”, which will be discussed in full in Chapter 10 (DSP Fundamentals). The
idea of a median filter is to keep five samples of data in the order they were received,

76
transfer them to a second data array, sort the second data array into order, and send out
the middle point (the median) as the “output”.
The time-optimal way to do this using ROM States is to keep part of the previous
order in the second array and, when a new sample comes in, fit that sample into the
mostly-sorted array so that the whole array is sorted again. First, find the oldest sample
(which will be discarded) in the sorted array and mark its position in the array by doing
one of five “GOTO” instructions. At the target addresses of each of these “GOTO”s the
new sample is tested against the other four sorted positions and, depending on where the
new sample will fit in, five more “GOTO” instructions are used to mark the new position.
In each of these program parts, in turn, only the smallest number of shifts in the data
array are used to insert the new sample into the new sorted order. Twenty-five “GOTO”
instructions, as above, are used altogether but the code is not redundant. The code is
specific to the tasks to be done.
If the median filter is done without ROM States, in the traditional way, the
running-time is 113 instruction-cycles and the subroutine takes up about 80 program
memory words. If the median filter does use ROM States, the running-time is 57
instruction-cycles and the subroutine uses 320 program memory words. Using ROM
States cuts the running-time in half but quadruples the program size.
6.16 Limitations of C/C++
The major selling-point of using a C/C++ compiler is that it is said to produce
object code that is very nearly like that of an assembly language program while allowing
the user to write in machine-independent code. While this is true for traditional
programs, high-end ESP programs, and medium-end ESP programs, it is not true, in
general, for low-end ESP programs.
The C/C++ language can be used successfully in low-end embedded programs
which are not timing-critical. Low-end systems that need to work at high speeds or high
efficiencies cannot use C/C++ because the compiler produces code which is far inferior
to assembly language code. Optimum-time low-end code is necessarily a function of the
code geometry and is by no means “just an algorithm”.
For example, how would a C/C++ compiler for the PIC16F877 fill a twenty-
element RAM array with zeros when it is given the statement:
“for(k=0;k < 20;k++) Array[k] = 0;” ?
It would probably produce a loop that counts to twenty and stores zeros in the array by
using indirect addressing. This is fine for an application which is not timing-critical but it
is useless for one that is. What would be the fastest way to fill a twenty-element array in
RAM with zeros using the PIC assembly language? Use twenty “CLRF ARRAYn”

77
statements in a row! How many standard C/C++ compilers will produce twenty
“CLRF”s in a row when given the “for”-loop above? None! In general, how many
standard C/C++ compilers for the PIC can make the distinction between making “slow-
code” for one part of the program and “fast-code” for another part? None!
If special instructions, keywords, or classifiers were added to the C/C++ language
specifically for use in low-end systems to tell the compiler what “speed” of code to use,
the resulting compiler would still need some advanced artificial intelligence software to
get the right code. The blanket statement that, “C/C++ produces object code just like
assembly code”, when it is applied to low-end systems, is absurd!
Someone will say to me, “All your examples use the 4 MHz PIC when there is a
20 MHz PIC available. Why not go with the 20 MHz PIC and stop complaining about
C/C++?” That reply is OK for “all-digital” systems or ones that have shielding and filters
(at some extra cost). But what if the product has sensitive analog circuits or it is to be
used in an RF-sensitive environment? The RF noise produced by a 20 MHz PIC may be
prohibitive! A 4 MHz system may work marginally, but a 20 MHz system may not work
at all!
Another idea is to use a 20 MHz PIC and under-clock it to run at, say, 5 MHz, to
increase the speed of the system if the maximum speed of 20 MHz cannot be used. This
too may have problems.
When I did embedded systems programming and system-design in the security
systems industry, our company tried to use our radio receiver with another company’s
security panel. When we hooked the receiver up to the panel we got a transmitting range
of about two inches! The panel used a microprocessor with what was then a fast speed of
12 MHz and its busses were multiplexed. It turned out that the chip they used to
demultiplex the busses had signal rise-times of two-nanoseconds! This alone produced
enough RFI to block our signal. We substituted a slower demultiplexer chip and we got
ranges of about 25 feet! The 12 MHz clock speed never changed. This is why fast chips
may be just as bad as high clock speeds.
Embedded systems design, especially at the low-end, is inherently a multi-
disciplinary field. Electromagnetic interference and electromagnetic compatibility
(EMI/EMC) problems must be considered at the start of the design process. Often the
easiest way to get rid of noise and EMI/EMC problems is to recognize the potential for
them at the start of the design process and prevent them from happening in the first place!
Never assume that a system that works at a “low” speed will also always work at
a “high” speed.
The special techniques in this chapter are used for “high-speed” operations. In
cases where a low clock-speed is used, these are the software techniques that are time-
optimal. They make the most efficient use of the processor’s time. Using fast chips and
C/C++ in a low-end system may not be practical.

78
Chapter 7: Advanced ESP
7.0 Chapter Summary
Section 7.2 discusses sine wave generation by the direct digital synthesis method.
Section 7.3 uses section 7.2’s results to generate Touch Tone/DTMF signals. Section 7.4
covers software generation of pulse width modulation. Section 7.5 covers ADPCM data
compression method. Section 7.6 discusses ideas for testing and practical embedded
systems.
7.1 Introduction
This chapter will look at some useful techniques, tools, testing methods, and
system ideas for advanced ESP. A technique for generating sine waves is developed and
is an essential part of an embedded systems programmer’s tool-kit. A data compression
technique for speech signals is also developed. Testing methods and system ideas are
discussed in full.
7.2 Sine Wave Generation
One way to generate high-quality sinusoidal signals is to use the Direct Digital
Synthesis method (DDS). This process, also called the “phase-addition method”, can
generate high frequencies with high resolution and high spectral purity. It requires a
digital-to-analog converter (DAC) and an analog low-pass filter to shape the sine wave.
The DDS process can be done in hardware or in software.
To develop this technique, consider a look-up table with 256 entries containing a
complete sine wave. If one value were taken and sent out to a DAC with a low-pass filter
at the DAC output, and this was repeated with a table-increment of one (1) at a rate of ten
kilohertz (10 kHz), there would be a sine wave at the filter output of about 39 Hz or
roughly 10000 Hz / 256. If the table-increment were increased to two (2), instead of one
(1), the resulting output sine wave frequency would be 78.125 Hz or roughly
10000 Hz / 128. We could also say Freq = 10000 Hz / (256 / 2).
What would happen if the table-increment were fifteen (15)? What frequency
would be produced and would it still be a sine wave? First, it would still be a sine wave.
Though the sequence of values in the look-up table would be “choppy”, the low-pass

79
filter would smooth-out the “chop” and produce a clean sine wave. Its frequency would
be as follows:
Freq = 10000 / (256 / Table-Increment)
= 10000 / (256 / 15)
= 585.9375 Hz
= 15 * (39 Hz), roughly.
What would be the largest possible frequency that could be produced in this
system? It turns out that the largest frequency is exactly one-half of the ten-kilohertz data
rate. The minimum number of points needed to produce a sine wave is only two (2).
(These points must be at the positive and negative peaks of the sine curve to get the
maximum amplitude at the output.) The corresponding table-increment for this case is
128 and gives a frequency of 10000 / (256 / 128) = 5 kHz. (The formal name for the
statement that “the maximum output frequency is half of the data rate” is the Nyquist
Sampling Theorem. We will see this again in Chapter 10 (DSP Fundamentals).)
Is it possible to produce a frequency less than the 39 Hz? Yes! What if we did a
table-increment of one (1) as before but at every other time step? This would effectively
give a table-increment of “one-half”. This would produce a frequency of half of the 39
Hz or 19.53 Hz. If the table-increment of one (1) were done at every fourth time step, the
output frequency would be one-fourth of the 39 Hz or 9.766 Hz.
Is it possible to produce a frequency that is half way in between the 39 Hz and its
double, 78 Hz? Yes! Repeat the following sequence of table-increments over-and-over:
Do a table-increment of one (1) at one time step then a table-increment of two (2) at the
next time step. This would give an effective table-increment of “one-and-a-half” and an
output frequency of 10000 / (256 / 1.5) = 58.594 Hz.
Let’s generalize this idea.
Suppose that the sine-table, the DAC, the filter, and the ten-kilohertz data rate are
all the same as before. Suppose now that there is a 16-bit register that holds the result of
each “sum-of-table-increments” (an “Accumulator”). Suppose that only the upper eight
bits of this accumulator will be used to index the sine wave look-up table. Suppose that
there is also a 16-bit register that holds the table-increment value which may now be a
16-bit number. The process is now to repeatedly add the 16-bit table-increment value to
the 16-bit accumulator over-and-over but let only the upper eight bits of the accumulator
do the indexing of the sine wave look-up table. The effect of the lower eight bits of the
16-bit table-increment and the 16-bit accumulator is to provide what would be
“fractional” table-increments, relative to the previous “whole-value/alternate time-step”
scheme of before.
For example, if the 16-bit table-increment is 0x0100, this would only increment
the upper eight bits of the accumulator, and the output frequency would be 39 Hz. Also
if the 16-bit table-increment is 0x0180, the following sequence of values would be

80
produced in the 16-bit accumulator: 0x0000, 0x0180, 0x0300, 0x0480, 0x0600, … and
so on. The series of table-increments in the upper eight bits of the accumulator is 0, 1, 3,
4, 6, … and so on. This corresponds to the alternate steps of one and two in the previous
examples and the output frequency would be 58.594 Hz.
In general, we can get the ratio of “Frequency-to-Increments” as:
F-to-I = Data Rate / Maximum-Number-of-Register-Values
= 10000 Hz / (2 ** 16)
= 10000 Hz / 65536
= 0.152588 Hz/Inc.
This value is also the lowest frequency that can be produced by this system. (This
corresponds to a Table-Increment of 0x0001.)
We can use this ratio to figure out how much of a table-increment we need to
produce any frequency. For example, suppose we want a frequency of 777 Hz:
Increment = 777 Hz / F-to-I
= 5092.1
= 5092 (Rounded Down).
Therefore, the frequency for this table-increment is Freq = 5092 * F-to-I = 776.98 Hz.
Here is one example of how this algorithm could be coded on the PIC:
SINE_DDS:
MOVLW DELAY_LOW
MOVWF TIME_LOW ; Set Fixed Duration
MOVLW DELAY_HIGH ; For Sine Wave
MOVWF TIME_HIGH
SINE_LOOP:
CLRF CARRY ; Reset Store for Carry Bit
MOVF INC_LOW,W ; Get Low-Half of INC
ADDWF ACCUM_LOW,F ; Add to Low ACCUM
BTFSC STATUS,C ; Get Carry Bit
INCF CARRY,F
MOVF INC_HIGH,W ; Get High-Half INC
ADDWF CARRY,W ; Add Carry Bit
ADDWF ACCUM_HIGH,F ; Add to High ACCUM
MOVF ACCUM_HIGH,W
CALL SINE_LOOK_UP
MOVWF DIGITAL_TO_ANALOG
CALL DELAY
DECFSZ TIME_LOW,F ; Decrement Duration
GOTO SKIP ; In Constant Time

81
DECFSZ TIME_HIGH,F
SKIP:
GOTO SINE_LOOP
RETURN
The total running-time of this loop must be 100 microseconds to get the ten
kilohertz data rate of the example (10 kHz).
7.3 Dual-Tone-Multi-Frequency (DTMF) Signaling
Telephones in the USA use DTMF dialing signals or “Touch Tone” signals. As
its name implies, this scheme uses two tones at a time to represent a symbol to dial. The
DTMF tones are split into a “high” frequency group and a “low” frequency group and the
two tones that are sent are as “one tone from each group”. The symbols and their
frequencies are as follows:
Symbol Low Frequency High Frequency
0 941 Hz 1336 Hz
1 697 Hz 1209 Hz
2 697 Hz 1336 Hz
3 697 Hz 1477 Hz
4 770 Hz 1209 Hz
5 770 Hz 1336 Hz
6 770 Hz 1477 Hz
7 852 Hz 1209 Hz
8 852 Hz 1336 Hz
9 852 Hz 1477 Hz
* (Star) 941 Hz 1209 Hz
# (Pound) 941 Hz 1477 Hz
Each of these tones must fit in a bandwidth of two-percent (2%) in order to be recognized
by the telephone company.
The DTMF tones can easily be generated by using the DDS technique. Two DDS
processes are run in parallel and the two sine values they produce are added together and
sent to the DAC. The same sine wave table is used but with half the maximum amplitude
as before so that the adding of the two sine values will not need a “divide-by-two” to fit
the DAC. The code for the DTMF DDS process is as follows:

82
DTMF_BY_DDS:
MOVLW DELAY_LOW
MOVWF TIME_LOW ; Set Duration of DTMF
MOVLW DELAY_HIGH
MOVWF TIME_HIGH
DTMF_LOOP:
CALL DO_HIGH_TONE_DDS
CALL SINE_LOOK_UP
MOVWF SINE_HOLD
CALL DO_LOW_TONE_DDS
CALL SINE_LOOK_UP
ADDWF SINE_HOLD,W
MOVWF DIGITAL_TO_ANALOG
CALL DELAY
DECFSZ TIME_LOW,F
GOTO SKIP
DECFSZ TIME_HIGH,F
SKIP:
GOTO DTMF_LOOP
RETURN
The DDS process is abbreviated here but is the same as before in every way.
7.4 Pulse-Width Modulation
A simple way to get around the need for a DAC in many applications is to use
pulse-width modulation (PWM). A PWM signal represents an analog value by varying
the duration of digital pulses. Just as a DAC uses variable voltages over constant time to
express analog values, PWM uses digital pulses over variable time. Integrating the DAC
signal and the PWM signal over time will give the same results. A PWM output needs
only a single digital output pin. An analog low-pass filter is still required, however.
The PIC16F877 already has two built-in hardware PWM units. Even so, it is
useful to know how to do PWM in software in case the processor you are using does not
have a hardware PWM unit. The basic PWM program is as follows:
PWM_ROUTINE:
MOVWF SAMPLE1 ; Call with W = PWM Output
MOVWF SAMPLE2
BSF PORTC,OUTPUT_BIT ; Assume This Output
PWM_LOOP1:
DECFSZ SAMPLE1,F ; Send “High” for Duration W
GOTO PWM_LOOP1

83
BCF PORTC,OUTPUT_BIT
PWM_LOOP2:
INCFSZ SAMPLE2,F ; Send “Low” for
; Duration 256 – W
GOTO PWM_LOOP2
RETURN
The total time this routine takes to run is about 770 microseconds. If each PWM DAC
value took this long, the data rate would be 1300 Hz. Accordingly, only frequencies up
to 650 Hz could be represented. If the total time were only 100 microseconds, a signal of
five kilohertz (5 kHz) could be represented, but the total range and resolution in the
above software loop would be only 32 steps. That is a maximum value of 32 could be
sent as an output value.
This is the main problem and tradeoff in using PWM. It does not matter if the
PWM is done in hardware or in software. The same problem and tradeoff exists for both
cases. It is a fundamental problem.
The software PWM process can be improved by using different software
techniques such as the “non-loop” time-delay process demonstrated in Chapter 6. This
will allow for a 100 microsecond data rate and a range and resolution of one hundred
(100). The full range of values is from zero to 100. This PWM program is as follows:
PWM_PROCESS:
MOVWF LOW_DURATION
SUBLW D’100’ ; Form W = 100 – W
ADDWF PCL,F ; For High Duration
BSF PORTC,OUTPUT_BIT
---- Continue “BSF”s for a total of 100 “BSF”s ------
MOVF LOW_DURATION,W
ADDWF PCL,F
---- Continue “BCF”s for a total of 100 “BCF”s -----
RETURN
For example, if W = 90 when this routine was called, the “ADDWF PCL,F” after the
“SUBLW” would do a “GOTO” to the point only ten steps away. This would allow 90 of
the “BSF”s to run. When the “Low Duration” value of 90 is given to the “ADDWF”, this
would do a “GOTO” to a point of 90 steps away, and allow only ten “BCF”s to run. This
solution is another example of the “ROM vs. Time” tradeoff in embedded systems

84
programming. It is much longer and uses more than 200 ROM words, but it is faster and
allows the greatest resolution.
There is an assembly directive that can simplify the fast PWM process called
“FILL” which fills in repeated instructions in program memory. Its form looks like:
FILL (assembly instruction),number-of-times
For example, “FILL (NOP),20” will fill in twenty “NOP”s into program memory.
The fast PWM loop can be re-coded as follows:
PWM_PROCESS:
MOVWF LOW_DURATION
SUBLW D’100’
ADDWF PCL,F
FILL (BSF PORTC,OUTPUT_BIT),100
MOVF LOW_DURATION,W
ADDWF PCL,F
FILL (BCF PORTC,OUTPUT_BIT),100
RETURN
Before the sound cards came out for the IBM-PC, PWM was used by some
companies to send speech to the PC’s built-in speaker attached to a digital output port
pin. The data rate was high enough that the speaker could not respond at that frequency
and it did not emit a high-pitched squeal. Clear speech came out of the PC’s speaker.
7.5 ADPCM Data Compression
ADPCM is a coding technique that allows an easy way to compress speech
signals in real-time by a factor of 2:1 or better. The compression allows some loss of the
exact information, however. ADPCM stands for “Adaptive Differential Pulse Code
Modulation”.
ADPCM encodes the difference between successive signal samples and a running
approximation to the signal. It is also adaptive in that if the signal is too large or too
small relative to the approximation of the signal, the range or scale is changed for the
next sample. Changes in volume are automatically compensated for and are reflected in
the coding. The encoding does introduce some noise, but there are filtering techniques
that can remove the noise and give very clear speech in the output (these filtering
techniques will be shown in detail in Chapter 10 --- DSP Fundamentals).

85
Suppose that speech is sampled at an appropriate data rate and is read as 8-bit
samples. After the ADPCM process, the data can be reduced to four-bits and then two of
these four-bit samples can be packed into one byte. The approximation to the signal is
stored in an 8-bit accumulator initially set to zero. When a signal value comes in, the
difference between it and the accumulator is found and is tested against a range of
allowed values. The best-fit of the difference to those values is found and a “code” or an
“index” is used to represent the best-fit value. If the best-fit value is at the maximum or
the minimum of the range of values, the range of values is doubled or halved,
respectively, for the next input sample when it is read the next time. The value of the
accumulator is updated with the best-fit value in the table by adding it to the accumulator.
For example, if the initial range is of 15 values from the set {-7, -6, -5, …, -1, 0,
1, …, 5, 6, 7}, doubling this range will give a range of values as {-14, -12, -10, …, -2, 0,
2, …, 10, 12, 14}. However, the range will not be doubled beyond a maximum range nor
will it be halved below a minimum range.
It should be noted that although the ADPCM process will reproduce the input
from the output, it does introduce noise and is a nonlinear process. Changes in volume
are accounted for but the process of doing so is not instantaneous. The “codes” or
“index” values mentioned above are four-bits wide and are the compressed data values.
The ADPCM compression-phase program is as follows:
ADPCM_COMPRESS:
SUBWF ACCUM,W ; W = ACCUM – W
SUBLW 0 ; W = W – ACCUM
MOVWF DIFF ; Store the Difference
CALL GET_STEP_CODE ; Double-Index Table
; Look-Up
; on Scale & Difference
MOVWF CODE
CALL GET_STEP ; Double-Index Table Look-Up
; on Scale & Code: Get Best-Fit
ADDWF ACCUM,F ; Update ACCUM, Form Sum
CALL GET_SCALE; Double-Index Table Look-Up
; Judge Max/Min Range, Get
; New Scale Index
MOVWF SCALE
BTFSS FLAGS,DATA_READY
GOTO ADPCM_COMBINE ; Pack the Code, Set
; “Ready”
NOP
MOVF CODE,W ; Set New Half-Byte
ANDLW 0x0F
MOVWF HALF_OUT ; Store New Half Code
BCF FLAGS,DATA_READY ; Data NOT Ready Yet

86
RETLW 0xFF ; Return an Invalid Code
ADPCM_COMBINE:
SWAPF CODE,W ; Put 2nd
Code in Upper Half Byte
ANDLW 0xF0
IORWF HALF_OUT,W ; Combine two Codes
BSF FLAGS,DATA_READY
RETURN ; W = Packed Byte Codes
The “GET_STEP_CODE” routine is a double-indexed table look-up with “DIFF”
and “SCALE” as the indices. First, do the “SCALE” index as:
GET_STEP_CODE:
MOVF SCALE,W
ADDWF PCL,F
GOTO SCALE0
GOTO SCALE1
GOTO SCALE2
GOTO SCALE3
GOTO SCALE4
The “SCALEn”s are for 256-entry tables indexed with the “DIFF” value. The object of
each one is to return the “CODE” value for the closest approach (best-fit) values in the
following table:
SCALEn Max/Min Values Delta Steps
0 +-7 1
1 +-14 2
2 +-28 4
3 +-56 8
4 +-112 16
There are 15 codes for each of the SCALEn tables. These codes run as zero for
the most negative value and 14 for the most positive value. The reason why the double
hashing is used and why the tables are allowed to take up so much memory is so that this
routine can run at the highest possible speeds. If this were not so, the time it would take
to find the codes, the steps, and the scales would be prohibitive and the ADPCM process
would be too slow to run in a 4 MHz PIC. A 20 MHz PIC might not have this problem
and a smaller ADPCM routine could be used.
Once the “CODE” is found and the “SCALE” is given, these are used to get the
“STEP” or “Best-Fit Value”, update the accumulator, and adjust the “SCALE” for the
next time. The “CODE”s are then packed in groups of two to a byte and returned in W to
the calling program. The value of “SCALE” must not be disturbed by the rest of the
program between subroutine calls. The “SCALE” and the accumulator must be
initialized to zero and the “Data-Ready” flag must be “set” (=1) initially.

87
The ADPCM expansion process is similar to the compression process except that
the “CODE” is given as the compressed data to be expanded. This process looks like:
ADPCM_EXPAND_NEW_CODE:
MOVWF HOLD_CODE ; W = New Packed Code
ADPCM_EXPAND_CODE2:
ANDLW 0x0F ; Isolate First Code
MOVWF CODE
CALL GET_STEP ; Get Corresponding
; Best-Fit Value
ADDWF ACCUM,F ; Update Accumulator
CALL GET_SCALE ; Get Next Scale
MOVWF SCALE
MOVF ACCUM,W ; Current Accum = Output
RETURN ; W = Expanded Data
ADPCM_EXPAND_SECOND_OLD_CODE:
SWAPF HOLD_CODE,W ; Get 2nd
Code
GOTO ADPCM_EXPAND_CODE2
The expansion is done in two parts. Once for a new code byte to get the first expansion
and then once for that same byte to get the second expansion.
7.6 Test Functions and System Ideas
Embedded systems should make liberal use of test pins and test functions. They
are useful in all phases of development, production, testing, and field-testing. Special
DIP-switches or header pins should be added to the system for testing purposes alone.
Fifteen or twenty percent of the available program code space should be reserved, if
possible, for testing and de-bugging aids.
The simplest, most common, and most general testing and de-bugging aids use
LEDs. An LED can be made to flash at a given rate to show a system’s activity. A non-
flashing LED would indicate a system failure. PWM can be used on an LED to make it
bright or dim, which can also serve as a system indicator. An LED can flash when the
watch dog timer gets its reset pulse. In an encoder or a transmitter, an LED can show
when a data-word is transmitted or, in a decoder or receiver, an LED can flash when a
data-word is decoded.
If the system uses seven-segment LED digits, the system can display messages for
testing and de-bugging. Detailed messages for showing the data and the system status
can be displayed.

88
An oscilloscope with a variable-delay sweep can be used to display the system’s
data and status using a serial stream of pulses. A single digital output pin can be used to
send this information. A simple sequence of instructions can be used to send out pulses.
For example:
BCF PORTC,TEST_PIN
BTFSC RAM_WORD,DATA_BIT
BSF PORTC,TEST_PIN
CALL DELAY1
BCF PORTC,TEST_PIN
CALL DELAY2
Also, for events which occur too quickly for the human eye to see, a pulse like
that above can be generated, held active for a few hundred milliseconds, and then
removed. This aids in the detection of transient events like a receiver’s “valid data
decode” signal.
A security system could be made to report its status information or it could be set
up with pseudo-random time delays for activating its alarms. If it can dial telephones, it
could call the factory or office and send a complete set of systems information for that
day. A special alarm mode could be reserved for units that get frequent or repeated
resets.
A system can often incorporate utility functions like a voltmeter, a frequency
counter, a timer, an alarm clock, a signal generator, a waveform generator, a pulse
generator, a noise generator, or a logic analyzer. Any of these things may be useful in a
field-testing situation.
If your embedded system interfaces with another (larger?) system, primitive
“look-alike” signals can be generated in your system to test it “as if” you had the other
system. That is, send diminutive mock-ups of the other system’s signals to mimic its
actions so that you don’t need to carry that system around with you to test your system.
Systems which are highly interactive with user menus can be built with a series of
interconnected jump-tables. A time-out feature should be included to partially reset the
command sequence if the user does not enter a command, does not finish responding to
the system in the way that the system “expects”, or if the user makes an error. Such a
system should be user friendly. The time-out can take the user back to a previous menu
or later to the main menu but give a warning before doing so.
Embedded systems can be made to be adaptive and have self-testing and self-
diagnostic features. For example, a decoder or a receiver could include the code for the
encoder and produce its own mock-up of a “transmitted signal with noise”. The
decoder/receiver could then try to decode and get calibrated with its own test signal.

89
EEPROMs can be used to store calibration information and other system settings
for its normal operation. The system can be built with its own calibration routines to set-
up a new or updated system.
Sometimes one processor isn’t powerful enough to perform all of the required
actions that the system needs. There may be severe time bottlenecks. The solution may
be to use a second or even, a third processor. Always watch out for the possibility that
your system may be over burdened and that the best solution might be to include more
processors.
Some kinds of system failures may be statistical in nature. It may be necessary to
measure probability distributions and statistical averages in order to diagnose the
problem.
Gaussian white noise can be used to improve a stable, DC-valued ADC reading by
plotting a histogram of the samples and comparing the histogram to a known Gaussian
distribution. The histogram is just a count of the number of times the data occurs. This is
accomplished by finding the upper and lower limits of the noise signal and setting up a
RAM array for the values within these limits (start with the RAM array reset to zeros).
The data is sampled and used to hash the RAM array then the value of the RAM at the
hashed address is incremented. After this process is repeated several hundred times, the
discreet histogram is formed. The discreet histogram values can be used to interpolate
between the ADC steps. It is possible to use a ten-bit ADC with this technique and
produce the equivalent of a 16-bit ADC! This technique of using noise to improve a
signal’s quality or measurement is called “dithering”. Figure 7-1 gives an illustration of
this process.
For processors that do not have ADCs built-in, it is possible to measure DC-
valued voltages using dithering on a one-bit ADC. The one-bit ADC is just a
comparator. The voltage to measure is fed in on one input and Gaussian white noise is
fed in on the other input. This produces a set of random, square wave pulses that the
processor can measure. The time duration, frequency, and overall time are measured and
are then compared to a known Gaussian distribution. The position on the Gaussian curve
can be found and the voltage can be determined to an almost arbitrary accuracy. It is not
uncommon to get measurements and an accuracy of up to 24-bits by using this method.
One lesson to be learned from the concept of dithering is that noise itself can be
used as a valuable signal in its own right. There are of course many times that noise is
bad but it can sometimes be used to a great advantage.

90
X
f(X)
1
0
f(X) =
1
2Pi s
e
-
X
s( )
2
X0.0 0.5 1.0 1.5-0.5-1.0-1.5
20 87 234 378 370 220 79
Sampled
Data
Points
Notice the Asymmetry About the Peak of 378
Fitting a Gaussian Curve Gives X-Offset = 0.23 and S = 1.0
Figure 7-1 Gaussian Probability Density Function and a
Set of Sampled Values

91
Chapter 8: PIC Peripherals and Interrupts
8.0 Chapter Summary
Section 8.1 gives an overview of the PIC’s peripherals. Section 8.2 covers the
input/output ports. Section 8.3 discusses the PIC’s interrupt system. Section 8.4
discusses the analog to digital converter and the analog multiplexer. Section 8.5 covers
the PIC’s built-in watch-dog timer. Sections 8.6, 8.7, and 8.8 cover the counters/timers.
Section 8.9 covers the capture mode operations. Section 8.10 covers the compare mode
operations. Section 8.11 discusses the PIC’s hardware pulse width modulators.
Section 8.12 covers the parallel slave port. Section 8.13 discusses reading and writing
the data EEPROM. Section 8.14 discusses reading and writing the program memory.
Section 8.15 discusses the data protection modes. Section 8.16 covers the
CONFIGURATION word and its settings. Section 8.17 discusses the PIC’s sleep and
reset modes.
The PIC16F877 is more than just a CPU with some RAM and ROM. There are
also several useful peripheral hardware modules built-in to make it a complete computer
control system. Many products can be made that use the PIC as a complete single-chip
solution to its system design.
This chapter will focus on the peripherals and special functions of the PIC. The
details and instructions on how to use them will be given in full with examples.
8.1 Overview of the PIC Peripherals
1) Input/Output Ports
There are five port-sets as Ports (A,B,C,D,E) with bits and pins that may be set in
software as inputs or outputs. These pins are also shared with other peripheral
functions and to use these other functions requires setting up the input/output
ports for compatibility. Thirty-three (33) input/output port pins are available.
2) Interrupts
As briefly noted in Chapter 6, an interrupt is a way for peripherals and other
hardware to capture the attention of the CPU and have it call a subroutine to
service the hardware with the user’s software. The subroutine address is at a
fixed location in program memory and is built-in to the CPU. There are a total of
14 possible interrupts and each of them can be enabled or disabled in software.
3) Analog-to-Digital Converter & The Analog Multiplexer
The PIC contains a ten-bit ADC and has as many as eight available analog input
channels. These analog inputs are shared with the port pins of Port A and Port E.

92
It is also possible to select among these pins a place to attach an external voltage
reference in the case where the user does not want to use the internal voltage
reference of five volts. This feature is also software selectable.
4) The Watch-Dog Timer
As noted in Chapter 6, the PIC contains its own independent watch-dog timer
which can be disabled not by software but by a setting in the downloading
process. The watch-dog timer has the power to reset the PIC when it overflows.
The user’s job in the software is to provide the watch-dog timer with regular
commands to reset the watch-dog timer so that it will not reset the PIC.
The purpose of the watch-dog timer is to make sure the software does not get
trapped in infinity loops and thus the software is made much more reliable by
using the watch-dog timer. In addition to the watch-dog timer is a prescaler
which can extend the watch-dog’s time-out period.
5) Timer0, Timer1, and Timer2
These three timers can count clock pulses or count external pulses on the PIC’s
port pins. They are programmable and have prescalers and sometimes postscalers
to modify their counts and counting processes. They often serve as time-bases for
other peripherals or for providing regular interrupts to the user’s software.
6) Capture Mode (Two of them)
The capture mode modules are hardware-controlled ways to measure pulse-widths
with reference to the system clock.
7) Compare Mode (Two of them)
The compare mode modules compare Timer1’s count to a fixed, but user-
programmable, register value. When that value is reached, the hardware can send
out a pulse, trigger an interrupt, or do some kind of “special event” within the
PIC such as resetting and reloading Timer1 and/or starting the ADC. This is also
useful for setting up a time-base for the software.
8) Pulse-Width Modulation (Two of them)
The PWM modules use Timer2 to generate signals on a PIC output pin and serve
as DACs. Both PWM modules have a maximum of ten bits of resolution.
9) Parallel Slave Port
Port D can be used as a bi-directional, microprocessor-type, data-bus port. The
PIC is controlled as a slave to three external control signals which are
manipulated by the microprocessor or other device.
10) EEPROM Data Memory
The PIC has 256 bytes of non-volatile EEPROM and can be programmed in the
PIC software independent of a device programmer.

93
11) FLASH Program Memory
In a similar way to the EEPROM Data Memory, the FLASH Program Memory
can also be programmed in software independent of a device programmer. New
program features and updates can be downloaded to the PIC’s software while the
unit is in the field.
12) Code and Data Protection
The PIC can be configured to lock-out attempts to read or write its FLASH and
EEPROM Data Memories. This can be set only during the downloading process.
13) Resets and Sleep Mode
The PIC has several modes of resets and status conditions to identify which of
them has occurred. Also, there is a power-saving mode called “sleep” which can
be initiated in software to power-down the CPU. This is useful in battery
powered applications.
In all of the sections to follow, the special function registers and the bits that
control the peripherals are shown in detail in Appendix C.
8.2 Input/Output Ports
There are five input/output ports as Port (A, B, C, D, E). Most of these act alike
but some are different and use different features and options.
8.2.1 Port A
There are six (6) Port A pins and each can be set as an input or an output. All of
the Port A pins except RA4 have a shared use with the analog multiplexer (MUX). The
RA4 pin can be a Schmitt Trigger input (hysteresis) or an open-drain output.
The first step in configuring Port A is to select which pins are to be set as digital
input/output and which are to be analog inputs. This is done with the lower four bits of
the ADCON1 register file (RAM). These settings are shown in Figure 8-1. If we want
all of Port A to be digital, the setting is 0x06.
Of the pins which are digital input/output the selection of “input” vs. “output” is
made with the TRISA register file. If a Port A pin is to be an “input”, the TRISA bit
corresponding to the Port A pin/bit must be set to one (=1). If it is to be an “output”, the
TRISA bit must be set to zero (=0).

94
PCFG3:
PCFG0
AN7 AN6 AN5 AN4 AN3 AN2 AN1 AN0
RE2 RE1 RE0 RA5 RA3 RA2 RA1 RA0
0000
0001
0010
0011
0100
0101
0110
0111
1000
1001
1010
1011
1100
1101
1110
1111
A
A
A
D
D
D
D
D
D
D
D
D
D
D
D
D
A
A
A
D
D
D
D
D
D
D
D
D
D
D
D
D
A
A
A
A
A
A
D
D
D
D
D
D
D
D
D
D
A
A
A
A
A
A
A
A
A
D
D
D
D
D
D
D
A
A
A
A
Vref+
Vref+
Vref+
Vref+
Vref+
Vref+
Vref+
Vref+
Vref+
D
D
D
A
A
A
A
A
A
Vref-
Vref-
Vref-
Vref-
Vref-
D
D
D
D
D
D
D
D
D
A
A
A
A
A
A
A
A
A
A
A
A
D
D
A
A
A
A
A
A
A
A
A
A
A
A
A
A
PCFG3:PCFG0 are bits 3,2,1,0 in ADCON1
D = Digital I/O A = Analog Input Channel
Figure 8-1 ADCON1 "Analog vs Digital" Selection Codes
For example, if we want all of the Port A bits as “digital” and RA0, RA1, RA2 are
to be outputs and RA3, RA4, and RA5 are to be inputs, the code would be as follows:
BANKSEL PORTA ; Bank 0
CLRF PORTA ; Reset Port A Before Configure
BANKSEL ADCON1 ; Bank 1
MOVLW 0x06 ; Select “All Digital” on Port A
MOVWF ADCON1
MOVLW 0x38 ; RA(0,1,2) = Out
MOVWF TRISA ; RA(3,4,5) = In
BANKSEL PORTA ; Bank 0

95
8.2.2 Port B, Port C, Port D
None of Port B, Port C, or Port D is shared with the analog multiplexer and none
need be configured with the ADCON1 register file. There are TRISB, TRISC, and
TRISD register files which control the “input” vs. “output” selections of each of these
port pins. These work in the same way as Port A. The Port B pins also have a user-
selectable “weak pull-up” option that can be enabled by clearing the “/RBPU” bit of the
“OPTION_REG” register file. ( Do “BCF OPTION_REG,RBPU”.) This feature is
automatically disabled when a Port B pin is configured as an output.
8.2.3 Port E
The Port E pins also share their pins with the analog multiplexer as in Port A and
its use is identical with Port A. (Use the ADCON1 and TRISE register files.)
8.3 Interrupts
The PIC16F877 has a total of 14 sources of interrupts. Each of these has an
“Interrupt Enable Bit” which must be set (=1) to enable or use the interrupt and an
“Interrupt Flag”, which is set (=1) automatically when the interrupt is activated. There is
also a “Global Interrupt Enable Bit” and a “Peripheral Interrupt Enable Bit”. The
“Global” must be set (=1) before any of the other interrupts will be enabled. The
“Peripheral” must be set (=1) before any peripheral interrupt will be enabled. Figure 8-2
shows the schematic for the PIC’s interrupt logic.
When an interrupt occurs the CPU treats it like a subroutine “CALL” and the
program-counter is set for the program address, 0x0004. The user must place the
interrupt-handling subroutine at this address (this subroutine is called, the “Interrupt
Service Routine” or “ISR”). A special instruction is used to return from an ISR: It is the
“RETFIE” instruction.
Before returning from an interrupt, the user must reset the “Interrupt Flag” which
was set (=1) when the interrupt occurred. If this is not done, the interrupt hardware will
automatically and immediately cause another interrupt to occur when the CPU executes
the “RETFIE” instruction! It is also possible to inadvertently cause an interrupt to occur
when the user sets (=1) an “Interrupt Enable Bit”. If the corresponding “Interrupt Flag”
is set (=1) when the user sets the “Interrupt Enable Bit”, an inadvertent or accidental
interrupt will occur!

96
AND
OR
AND
AND
AND
AND
AND
AND
AND
AND
AND
AND
AND
OR
EEIF
EEIE
PSPIF
PSPIE
ADIF
ADIE
RCIF
RCIE
TXIF
TXIE
SSPIF
SSPIE
CCP1IF
CCP1IE
TMR2IF
TMR2IE
CCP2IF
CCP2IE
TMR1IF
TMR1IE
BCLIF
BCLIE
AND
AND
AND
AND
PEIE
T0IF
T0IE
INTF
INTE
RBIF
RBIE
GIE
Wake-Up (If in SLEEP)
CPU INTERRUPT
Figure 8-2 PIC16F877 Interrupt Logic Tree
The user’s response to an interrupt in the ISR must do two things:
1) The contents of the W register and the “STATUS” register must be saved in
RAM.
2) The set of “Interrupt Flags” being used must be searched to determine which
“Interrupt Flag” caused the interrupt to occur.
Before returning from an interrupt, four things must be done in the following order:
1) When the peripheral which caused the interrupt has been serviced, the
peripheral’s “Interrupt Flag” must be reset by the software.
2) All of the other “Interrupt Flags” of the ones being used, must be checked to
see if they are reset (=0). If any are set, another interrupt by a different
peripheral has been called and it must be serviced. When it has been serviced,
go back to Step 1.
3) Restore the W and “STATUS” registers.
4) Do the “RETFIE” instruction.
Note that the worst-case response-time for any interrupt, when it has been initiated, is
four instruction cycles in duration.

97
The code to save the W and “STATUS” registers has some tricks to it. The code
is as follows:
MOVWF W_TEMP ; Save W
SWAPF STATUS,W ; Get STATUS into W
MOVWF STATUS_TEMP ; Save STATUS
The code to restore these registers is as follows:
SWAPF STATUS_TEMP,W ; Restore STATUS
MOVWF STATUS
SWAPF W_TEMP,F ; Restore W
SWAPF W_TEMP,W
The trick is that we can’t use “MOVF” anytime we want to move data from a register file
to the W register since the “MOVF” instruction affects the “Z” (Zero) STATUS Flag!
Doing so would destroy the “Z”-Flag state we are trying to preserve! This is why the
“SWAPF” instructions are used. (This code is repeated in Appendix F.)
Now let’s look at two specific interrupts and their options. The first is the
external interrupt (“INT”) which is located on pin 33 and is shared with Port B, Bit Zero
(RB0). To use the “INT”, Port B, Bit Zero must be set-up with TRISB to be an “input”
port pin (TRISB, Bit 0 = 1). If this were not so, and if Port B, Bit Zero were an “output”
pin, whatever value was on the pin would jam the “INT” input. The “INT” interrupt is
edge-sensitive and the user may select one of either the positive or the negative edges
using the “INTEDG” bit in the “OPTIONS_REG” register file. If “INTEDG” = 1, this
selects the positive or rising edge. If “INTEDG” = 0, this selects the negative or falling
edge.
The code sequence to activate the “INT” interrupt is as follows:
BANKSEL TRISB ; Bank 1
BSF TRISB,0 ; Set Port B, Bit Zero as an “Input”
BSF OPTIONS_REG,INTEDG ; Set “Rising” Edge
BCF INTCON,INTF ; Reset “Interrupt Flag” of INT
BSF INTCON,INTE ; Set “Interrupt Enable Bit” of INT
BSF INTCON,GIE ; Set “Global” Int Enable
To reset the “INT”s “Interrupt Flag” when the ISR is done, do, “BCF INTCON,INTF”.
Another interrupt is the “RB Port-Change Interrupt”. When this is enabled, any
change on Port B (RB7, RB6, RB5, RB4) will cause an interrupt. These bits must be
selected as inputs with TRISB. The code to enable this interrupt is:

98
BCF INTCON,RBIF ; Reset “Interrupt Flag”
BSF INTCON,RBIE ; Set “Interrupt Enable Bit”
BSF INTCON,GIE ; Enable “Global” Interrupt Enable
To reset the “Interrupt Flag”, do, “BCF INTCON,RBIF”.
In the ISR for this interrupt, Port B must be read to prevent a latch-up of the PortB
bit-change that caused the interrupt to occur. Reading Port B between interrupts may
cause a mistake in the interrupt process.
There are interrupts available for each peripheral device. These will be discussed
when the peripherals are covered.
8.4 ADC and Analog MUX
The PIC’s ADC can convert an analog voltage to a ten-bit number. The analog
multiplexer (MUX) will allow up to 8 analog input channels to be converted by the ADC.
The voltage range limits may be taken internally as “+5 Volts and Ground”, or the user
can supply an external voltage reference on the analog input channel pins.
The ADC module uses two control registers in the register file map to set-up the
ADC and the analog MUX. These are “ADCON0” and “ADCON1”. The ADC may be
used with or without interrupts.
The first step in using the ADC is to set-up the analog MUX inputs which are
shared with Port A and Port E. This was discussed under using Port A by selecting a
code from Figure 8-1 and putting that four-bit code into the lower four bits of
“ADCON1”. For the selected analog MUX inputs, the TRISA and TRISE registers must
be set-up so that those Port A and Port E pins are configured as “inputs”. If they are
configured as “outputs”, a voltage of either five volts or ground will jam the ADC.
The next step is to select the ADC conversion clock rate using the “ADCS1” and
“ADCS0” bits of the “ADCON0” register. The selections are as follows:
ADCS 1:0 Clock Rate
(0,0) Fosc / 2
(0,1) Fosc / 8
(1,0) Fosc / 32
(1,1) Internal RC Clock (6 microseconds)
The rule is that the Clock Rate period must not be less than 1.6 microseconds. For Fosc
= 4 MHz this gives Fosc / 8 = 500 kHz and a period of 2.0 microseconds. Therefore,
select ADCS 1:0 as (0,1).
If the ADC is to be used with interrupts, the following code sequence will set this:

99
BANKSEL PIR1 ; Bank 0
BCF PIR1,ADIF ; Reset “Interrupt Flag” for ADC
BANKSEL PIE1 ; Bank 1
BSF PIE1,ADIE ; Set “Interrupt Enable Bit” for ADC
BSF INTCON,PEIE ; Set Peripheral Interrupt Enable
BSF INTCON,GIE ; Set Global Interrupt Enable
The results of the ADC conversion are ten bits long and are split into two, eight-
bit register file bytes: “ADRESH” for the high bits and “ADRESL” for the low bits. The
storage of the ten-bit result can be formatted as “Left-Justified” or “Right-Justified”. In
the “Left-Justified” format, the 8 most significant bits are placed in “ADRESH” and the
two least significant bits are placed in bits 6 and 7 of the “ADRESL” register. In the
“Right-Justified” format, the two most significant bits are placed in bits 1 and 0 of the
“ADRESH” register and the 8 least significant bits are placed in “ADRESL”. The user
selects the format by setting or clearing the “ADFM” bit of the “ADCON1” register as
follows:
ADFM = 1 = Right Justified
ADFM = 0 = Left Justified.
The ADC must also be turned “on” with the command, “BSF ADCON0,ADON”.
It can be turned “off” by using “BCF”, if needed, to conserve power.
When the ADC is used, the analog input channel must be selected. There are
eight channels and only one may be selected at a time with the “CHS2:CHS0” bits of
“ADCON0” as follows:
CHS2, CHS1, CHS0 Channel Name
(0,0,0) Channel 0 AN0
When the single channel to be used is selected, as above, the ADC can be started
as:
BSF ADCON0,GO ; Start the ADC
The “ADIF” interrupt flag is set when the ADC conversion process is finished. If the
“ADIE” interrupt enable bit is set (=1), setting the “ADIF” will cause an interrupt to

100
occur. Even if the “ADIE” interrupt enable bit is not set, the “ADIF” interrupt flag will
be set (=1) when the ADC is finished and it can be used independently of the interrupt
system. The “ADIF” can be tested by the software to see if the ADC conversion is
finished.
Let’s look at an example program that uses the ADC. Suppose that we want to
use only Channel Zero (AN0) and display the Left Justified, most significant results on
Port B.
First, here is an ADC example using interrupts:
LIST P=16F877
FLAGS: EQU 0x20 ; User’s Flags, Bit 0 = “Data Ready”
ADC_DATA: EQU 0x21 ; User’s ADC Data Hold
DATA_READY: EQU 0 ; Data Ready = Bit 0
ORG 0x0000
GOTO INIT
ORG 0x0004
GOTO INT_SERVICE
ORG 0x0006
INIT:
BANKSEL PORTB ; Bank 0
CLRF PORTB ; Reset PORTB
CLRF TRISB ; Port B = All Outputs
MOVLW 0x0E ; Chan 0, AN0, Left Justify
MOVWF ADCON1
BSF TRISA,0 ; RA0 = Input
MOVLW 0x41 ; ADC = “on”, Chan 0 Select,
MOVWF ADCON0 ; Clock = Fosc / 8
BANKSEL PIE1 ; Bank 1
BSF PIE1,ADIE ; Set ADC Int Enable Bit
BSF INTCON,PEIE ; Set Peripheral Int Enable
BSF INTCON,GIE ; Set Global Int Enable
BCF FLAGS,DATA_READY ; Reset Data Ready
; Flag
BSF ADCON0,GO ; Start the ADC

101
MAIN:
CLRWDT ; Reset Watch-Dog Timer
BTFSS FLAGS,DATA_READY ; Check if ready
GOTO MAIN
BCF FLAGS,DATA_READY ; Reset Data Ready
MOVF ADC_DATA,W ; Get ADC Data
MOVWF PORTB ; Send to Port B
GOTO MAIN
INT_SERVICE:
----- Save Registers ----- (See Appendix F)---
MOVF ADRESH,W ; Get Raw ADC Results
MOVWF ADC_DATA ; Save in User’s Hold
BCF PIR1,ADIF ; Reset “Int Flag”
BSF FLAGS,DATA_READY ; Set Data Ready
BSF ADCON0,GO ; Start ADC Again
----- Restore Registers -----
RETFIE ; Return From Interrupt
END
Now, here is a non-interrupt ADC example:
LIST P=16F877
ORG 0x0000
CLRF PORTB ; Reset PORTB
CLRF TRISB ; Port B = All Outputs
MOVLW 0x0E ; Chan 0, AN0, Left Justify
MOVWF ADCON1
BSF TRISA,0 ; RA0 = Input
MOVLW 0x41 ; ADC = “on”, Chan 0 Select,
MOVWF ADCON0 ; Clock = Fosc / 8
MAIN:
BCF PIR1,ADIF ; Reset “ADC Done”
BSF ADCON0,GO ; Start ADC
TEST:
BTFSS PIR1,ADIF ; Is ADC Finished?

102
GOTO TEST
MOVF ADRESH,W ; Get Raw ADC Data
MOVWF PORTB ; Send to Port B
GOTO MAIN
END
The ADC has its own sample-and-hold amplifier with a built-in capacitor and the
total conversion-time is a combination of the amplifier settling-time, the sample-and-hold
capacitor charging-time, the temperature, and the successive approximation convergence-
time of the ADC clocked logic. The detailed conversion-time breakdown is as follows:
Time of Amplifier Settling = 2 microseconds.
Time of S & H Charging = (Chold)*(Ric + Rss + Rs)*(ln(1/2047))
= 16.5 microseconds.
Time of Temperature Coef = (T – 25)*(0.05 microseconds/ Celsius Degree)
= 2.5 microseconds (worst case)
Time of ADC Logic = 12*(ADC Clock Period)
= 12*(2 microseconds) = 24 microseconds.
Total ADC Conversion-Time = 45 microseconds.
Therefore, the ADC can be cycled at a rate of 22.2 kHz.
The voltage step-resolution of a ten-bit ADC over a voltage range of “+5 Volts to
Ground” is figured as:
Voltage Resolution = ((+5 Volts) – (0 Volts)) / ((2 ** 10) – 1) = 4.89 millivolts
per step.
In general, the voltage resolution is the difference of the voltage ranges divided by the
number of steps minus one.
8.5 Watch-Dog Timer
The watch-dog timer built-in to the PIC runs with its own RC oscillator and has a
typical minimum time-out period of 7 milliseconds which in general is too short an
amount of time. There is a programmable prescaler available that can multiply this
period by 128 to give a total time-out period of 850 milliseconds, which is very good for
most applications.
The software must reset the watch-dog timer before it overflows and resets the
PIC. This is done with the “CLRWDT” instruction. If the watch-dog timer does reset the

103
PIC, there is a “STATUS” bit, “/TO”, which indicates that this has occurred. If “/TO” =
Zero (=0), the watch-dog timer has reset the PIC.
The watch-dog timer can be enabled or disabled using the “CONFIGURATION”
word or in the down-loading process. The watch-dog timer should NEVER be disabled!
Also, NEVER use an Interrupt Service Routine to reset the watch-dog timer!
The user can select the watch-dog timer prescaler by using the “PSA” bit in the
“OPTION_REG” register. This is done by setting the “PSA” (=1) and selecting the
amount of scaling desired as:
PS2, PS1, PS0 Scale Ratio (WDT)
(0,0,0) 1:1
(0,0,1) 1:2
(0,1,0) 1:4
(0,1,1) 1:8
(1,0,0) 1:16
(1,0,1) 1:32
(1,1,0) 1:64
(1,1,1) 1:128
The prescaler may also be configured to work with the Timer0 module. However, you
cannot use it for both the watch-dog timer and the Timer0 module at the same time. The
better choice, in my opinion, is to keep the prescaler tied to the watch-dog timer since this
greatly improves the watch-dog timer.
8.6 Timer0
Timer0 is an 8-bit counter/timer which can be driven from the Fosc/4 clock, a
prescaled Fosc/4 clock, or an external pin source called, “T0CKI”, which is shared with
Port A, RA4. Timer0 is readable and write-able and its output can generate an interrupt,
“T0IF”. If Timer0 is used as a counter (with “T0CKI” input), the user can select if it
should count on the rising edge or the falling edge.
The programmable prescaler may be used on either the timer mode or the counter
mode. This prescaler is the same one that is used by the watch-dog timer and it cannot be
used by both Timer0 and the watch-dog timer at the same time.
Since the “T0CKI” input is shared with the Port A, RA4 pin, the user must set-up
Port A, RA4 as an “input” with TRISA, if the counter mode is to be used.
Whenever the Timer0 register, “TMR0”, is written to, there is an “inhibit” on its
incrementing process for the next two instruction cycles. For example, if the timer

104
overflows and we want to re-initialize it so that it will reach an effective count of 250, we
must write a value of “8” instead of “6” to allow for the extra delay. (This does not
apply if the prescaler is set to a ratio larger than two since its effects will not be seen.)
The prescaler is selected by using the “PSA” bit in the “OPTION_REG” register.
The “PSA” bit is cleared (=0) and the scale ratio is set as:
PS2, PS1, PS0 Scale Ratio (Timer0)
(0,0,0) 1:2
(0,0,1) 1:4
(0,1,0) 1:8
(0,1,1) 1:16
(1,0,0) 1:32
(1,0,1) 1:64
(1,1,0) 1:128
(1,1,1) 1:256
To set Timer0 as a “timer” do “BCF OPTION_REG,T0CS”. To set Timer0 as a
“counter” use “BSF”. In the “counter” mode do “BCF OPTION_REG,T0SE” to select
counting on the falling edge and do “BSF” to select counting on the rising edge.
Let’s look at an example program that uses the Timer0 module as a “timer” that
serves as a time-base and generates an interrupt every four milliseconds. The prescaler is
used at a ratio of 1:32.
LIST P=16F877
FLAGS: EQU 0x20 ; User flags, use bit 0
ORG 0x0000
GOTO INIT
ORG 0x0004
GOTO TIMER0_ISR
ORG 0x0006
INIT:
----- Do Other Inits Prior to Doing Timer0 -----
BANKSEL OPTION_REG ; Bank 1
MOVLW 0x04 ; T0 Prescale, Timer, 1:32
MOVWF OPTION_REG
MOVLW D’131’ ; Set count for 125
MOVWF TMR0
MOVLW 0xA0 ; Enable T0 & Global INTs
MOVWF INTCON

105
CLRF FLAGS
MAIN:
------ Do Program Steps -----
------ Test For “FLAGS, Bit 0” ------
ORG 0x0200
TIMER0_ISR:
------ Save Registers ------(See Appendix F)----
MOVLW D’131’ ; Reset Count for 125
MOVWF TMR0
BCF INTCON,T0IF ; Reset T0 INT Flag
BSF FLAGS,0 ; Set User Flag: INT Occur
------ Restore Registers -------
RETFIE ; Return from INT
END
8.7 Timer1
Timer1 is a 16-bit counter/timer with its own dedicated prescaler and an option to
use an external oscillator in place of the Fosc/4 clock. There is also an option to
synchronize the external clock to the internal clock, if it is so desired. Timer1 can
generate interrupts, if they are enabled. The two Timer1 counter registers, “TMR1H” and
“TMR1L”, are the most significant and least significant bytes, respectively, and are both
readable and write-able.
To use Timer1, it must first be turned “on” by doing, “BSF T1CON,TMR1ON”.
The “counter” vs. “timer” mode may be selected as, “BCF T1CON,TMR1CS”, to select
the Fosc/4 timer input. Doing “BSF” of the same will select the external input on pin
“T1CKI” which is shared with Port C, bit RC0 which must be set-up as an “input” with
TRISC.
The external crystal oscillator is connected to the, “T1OSO” and “T1OSI”, pins
and the oscillator module can be turned “on” by doing, “BSF T1CON,T1OSCEN”. The
external clock source may be synchronized by doing, “BCF T1CON,T1SYNC”. The
“T1OSO” and “T1OSI” pins are shared with Port C, pins RC0 and RC1. These must be
selected as “inputs” with TRISC.
The Timer1 prescale selection is done as:
T1CKPS1:T1CKPS0 Scale Ratio
(0,0) 1:1
(0,1) 1:2
(1,0) 1:4
(1,1) 1:8

106
Since Timer1 counts in two halves, the timer should be turned “off” when new values are
written to the timer registers. Do this as “BCF T1CON,TMR1ON”.
Examples of using Timer1 will be shown in the “Capture” and “Compare” modes
of operation.
8.8 Timer2
Timer2 is an 8-bit timer that has a user programmable prescaler and postscaler.
There is also a period register, “PR2”, which can be set-up and matched to the Timer2
register “TMR2”. When a match occurs, “TMR2” is reset and an interrupt can be
generated. Timer2 is used mostly for generating precise and exotic clock frequencies.
The “TMR2” and “PR2” registers may be read and written to at any time.
The Timer2 prescale selection is done as:
T2CKPS1:T2CKPS0 Scale Ratio
(0,0) 1:1
(0,1) 1:4
(1,0) 1:16
(1,1) 1:16
The Timer2 postscaler selection is done as:
TOUTPS3:TOUTPS0 Post-Scale Ratio
(0,0,0,0) 1:1
(0,0,0,1) 1:2
(0,0,1,0) 1:3
(0,0,1,1) 1:4
(0,1,0,0) 1:5
(0,1,0,1) 1:6
(0,1,1,0) 1:7
(0,1,1,1) 1:8
(1,0,0,0) 1:9
(1,0,0,1) 1:10
(1,0,1,0) 1:11
(1,0,1,1) 1:12

107
(1,1,0,0) 1:13
(1,1,0,1) 1:14
(1,1,1,0) 1:15
(1,1,1,1) 1:16
An example program which uses Timer2 will be shown in the “Pulse-Width Modulator”
section.
8.9 Capture Mode
The capture mode is a hardware method of measuring pulse-widths and delays
between pulses. There are two capture mode modules and both of them use the 16-bit
timer, Timer1, for their time-bases. Timer1 must be set-up prior to using the capture
mode modules.
The capture mode modules use the “CCP1” and “CCP2” pins, which are shared
with Port C, pins RC2 and RC1, respectively. These pins must be set-up as “inputs” with
the TRISC register.
The capture mode inputs may look for the following “events” on the input pins:
1) Every falling edge.
2) Every rising edge.
3) Every 4th
rising edge.
4) Every 16th
rising edge.
When one of these events occurs, the Timer1 values are placed into “CCPR1H” and
“CCPR1L” or “CCPR2H” and “CCPR2L” as the high and low bytes of Timer1’s count,
respectively. Interrupts must be used since, if another event occurs before the previous
event’s data can be read, that previous data will be overwritten and destroyed.
If Timer1 is being used with an external clock, it must be set-up as
“synchronized” or else the capture process may fail.
The capture mode modules use the “CCP1CON” and “CCP2CON” control
register files to select the types of events for each module. The lower four bits of each
are used for this selection and are coded as:

108
CCP1M3:CCP1M0
CCP2M3:CCP2M0 Meaning
(0,0,0,0) CCP1 or CCP2 disabled, reset.
(0,1,0,0) Every Falling Edge
(0,1,0,1) Every Rising Edge
(0,1,1,0) Every 4th
Rising Edge
(0,1,1,1) Every 16th
Rising Edge
Switching between these modes or events while the interrupts are enabled may cause
false triggering of an interrupt.
Note that the capture, compare, and PWM modules are not independent of each
other and must not be used without considering the possible conflicts of their usage.
An example program that uses the capture mode and Timer1 is as follows.
Suppose that the “CCP1” input (Port C, RC2) is used with the “Every Rising Edge” event
to measure pulse-widths.
LIST P=16F877
LOW_HALF: EQU 0x20 ; Pulse-Width, Low
HIGH_HALF: EQU 0x21 ; Pulse-Width, High
FLAGS: EQU 0x22 ; User Flags
EDGE: EQU 0 ; Flag: First Edge Found
READY: EQU 1 ; Flag: Data Ready
ORG 0x0000
GOTO INIT
ORG 0x0004
GOTO CCP1_ISR
ORG 0x0006
INIT:
BANKSEL PORTC ; Bank 0
CLRF INTCON ; Clear All INT Flags
CLRF TMR1L ; Reset Timer1
CLRF TMR1H
MOVLW 0x01 ; Timer1 ON, Prescale = 1
MOVWF T1CON
BANKSEL TRISC ; Bank 1
BSF TRISC,2 ; RC2 = Input
BSF PIE1,CCP1IE ; Enable CCP1 INT Enable Bit
BCF PIR1,CCP1IF ; Reset CCP1 INT Flag

109
MOVLW 0x05 ; Set Event: Every Rising Edge
MOVWF CCP1CON
CLRF FLAGS ; Reset User Flags
MOVLW 0xC0 ; Enable Global & Peripheral
MOVWF INTCON ; Interrupts
MAIN:
----- Look For FLAGS,READY = 1-------
----- Use This Data When Ready ----------
CCP1_ISR:
----- Save Registers ------(See Appendix F)-----
BTFSC FLAGS,EDGE ; First Edge Found?
GOTO GET_DATA_ISR ; Yes: Copy Data, Ready
BSF FLAGS,EDGE ; No: Reset Timer1
CLRF T1CON ; Turn Off Timer1
CLRF TMR1H
BSF T1CON,TMR1ON ; Turn Timer1 On
GOTO DONE_ISR
GET_DATA_ISR:
MOVF CCPR1L,W ; Get Low Data
MOVWF LOW_HALF
MOVF CCPR1H,W ; Get High Data
MOVWF HIGH_HALF
BSF FLAGS,READY ; Set Data Ready
DONE_ISR:
BCF PIR1,CCP1IF ; Reset INT Flag
RETFIE ; Return From INT
END
8.10 Compare Mode
There are two compare mode modules that test the value of the 16-bit timer,
Timer1, against a 16-bit, user specified, threshold and, when there is a match, an output
pin is set/cleared, an interrupt is generated, or a “special event” is triggered. There are
two compare mode modules, “CCP1” and “CCP2”, which each have distinct “special
events”. The special event for “CCP1” is to reset and restart Timer1. The special event
for “CP2” does the same as for “CCP1” but also starts the ADC. Interrupts may also be
generated when they are enabled.

110
The same rules, registers, interrupts, and pins are used here as in the “capture
modes” except that Port C, RC1 and RC2, must be configured as “outputs” with TRISC
rather than “inputs”.
The “CCP1CON” and “CCP2CON” register bits have different meanings and
settings in “compare mode”. They are:
CCP1M3:CCP1M0 Meaning / Action
(1,0,0,0) CCP1 = 1 and CCP1IF = 1
(1,0,0,1) CCP1 = 0 and CCP1IF = 1
(1,0,1,0) CCP1IF = 1 (Only)
(1,0,1,1) CCP1IF = 1 and “Reset/Restart Timer1”
CCP2M3:CCP2M0 Meaning / Action
(1,0,0,0) CCP2 = 1 and CCP2IF = 1
(1,0,0,1) CCP2 = 0 and CCP2IF = 1
(1,0,1,0) CCP2IF = 1 (Only)
(1,0,1,1) CCP2IF = 1 and “Reset/Restart Timer1”
and “Start ADC”
Where “CCP1IF” and “CCP2IF” are the interrupt flags of each module.
As it was stated in the “capture mode”: The capture, compare, and PWM modes
are not independent of each other and must be checked for conflicts of usage.
An example program that uses the compare mode and Timer1 is as follows.
Suppose we want to use the “CCP1” module with its special event to produce regular
interrupts at intervals of one millisecond. This is a good alternative to using Timer0 as a
time-base since the Timer0 prescaler can be freed to work with the watch-dog timer.
LIST P=16F877
LIMIT_LOW: EQU 0xE8 ; LSB of 1000
LIMIT_HI: EQU 0x03 ; MSB of 1000
READY: EQU 0 ; FLAGS, Bit 0, “INT Occurred”
ORG 0x0000
GOTO INIT
ORG 0x0004
GOTO COMPARE_ISR
ORG 0x0006
INIT:

111
CLRF INTCON ; Reset Main INT Enables
CLRF T1CON ; Turn Timer1 “Off”
CLRF TMR1H
BSF T1CON,TMR1ON ; Turn Timer1 “on”, Scale = 1
MOVLW LIMIT_LOW ; Set Compare Register Limit
MOVWF CCPR1L
MOVLW LIMIT_HI
MOVWF CCPR1H
BCF PIR1,CCP1IF ; Reset Compare Mode INT Flag
MOVLW 0x0B ; Set CCP1 Special Event
MOVWF CCP1CON
BSF INTCON,GIE ; Set Global INT Enable
BSF INTCON,PEIE ; Set Peripheral INT Enable
BSF PIE1,CCP1IE ; Enable CCP1 Interrupt
MAIN:
BTFSS FLAGS,READY ; INT Occurred?
GOTO MAIN
BCF FLAGS,READY ; Yes: Process Loop
------ Do Rest of Program Loop -------
GOTO MAIN
COMPARE_ISR:
BSF FLAGS,READY ; Set INT “Ready” Flag
BCF PIR1,CCP1IF ; Reset CCP1 INT Flag
END
Notice that the registers did not need to be saved and restored as before since the
special event automatically resets and restarts Timer1. No manipulations of the W
register and the STATUS flags were needed.
8.11 Pulse-Width Modulation (PWM)
There are two PWM modules in the PIC and both can serve as ten-bit Digital-to-
Analog Converters (DACs). Their outputs are on the “CCP1” and “CCP2” pins, or,
PortC, RC2 and RC1, respectively. These must be set-up as “outputs” using TRISC.

112
The PWM modules use Timer2 rather than Timer1. Also, the lower four
selection-bits of “CCP1CON” and “CCP2CON” must be set to (1,1,0,0) to select the
PWM modes.
In operation, TMR2 is compared with the limit set in PR2. To work over a full
bit-range, PR2 = 0xFF. For a given value of Fosc and a desired PWM frequency, this full
bit-range may not be possible but try it first as a baseline.
PWM Period = ((PR2) + 1)*4*Tosc*(TMR Prescale).
If Fosc = 4 MHz, the PWM Period is 256 microseconds. This gives a PWM frequency of
3.9 kHz, which may be acceptable for some applications, but it is not acceptable for
speech output. Speech output needs a PWM frequency of about 8 kHz.
If PR2 = 127, the PWM frequency = 7.8 kHz, which is good enough for speech
outputs. However, the bit-range is reduced to only 9 bits, but this is still very good
quality for sound.
To send an output sample in PWM1, enter the MSB value in “CCPR1L” and the
LSBs in “CCP1CON<5,4>”. When the PWM1 is ready for the next data to send, it
copies these values into the “CCPR1H” register automatically. PWM2 works in a similar
way but with “CCPR2L” and “CCP2CON<5,4>”. The PWM period and resolution are
the same as for PWM1.
An example program which uses PWM1 and Timer2 is as follows:
LIST P=16F877
USER_DATA: EQU 0x20 ; Output Data for PWM1
READY: EQU 0 ; Flags, Bit 0, Ready to Send
ORG 0x0000
GOTO INIT
ORG 0x0004
GOTO PWM_ISR
ORG 0x0006
INIT:
CLRF INTCON ; Reset Main INT Enables
CLRF USER_DATA ; Reset Data To Send
CLRF PORTC ; Reset Port C
MOVLW 0x04 ; Turn “on” TMR2

113
MOVWF T2CON ; No Scales
BCF TRISC,2 ; RC2 = Output, Use CCP1
MOVLW D’127’ ; PR2 = 127
MOVWF PR2
MOVLW 0x0C ; Set CCP1 = PWM1
MOVWF CCP1CON
BCF PIR1,TMR2IF ; Reset TMR2 INT Flag
BSF INTCON,GIE ; Enable Global INTs
BSF INTCON,PEIE ; Enable Peripheral INTs
CALL SEND_PWM ; Send a Zero
BSF PIE1,TMR2IE ; Enable TMR2 INT
MAIN:
----- Get Data To Send (Put it in W ) ------
WAIT:
BTFSS FLAGS,READY ; INT Occurred?
GOTO WAIT ; --- No, Wait For INT
BCF FLAGS,READY ; --- Yes, Get More Data
MOVWF USER_DATA ; Set New Data to Send
GOTO MAIN
PWM_ISR:
----- Save Registers -----(See Appendix F)------
BSF FLAGS,READY ; “INT” = “Data Sent”
CALL SEND_PWM
BCF PIR1,TMR2IF ; Reset TMR2 INT Flag
SEND_PWM:
BCF CCP1CON,4 ; Reset LSBs of Data
BCF CCP1CON,5
BTFSC USER_DATA,0 ; Copy LSB to LSB, Bit 5
BSF CCP1CON,5
BCF STATUS,C ; Rotate Right
RRF USER_DATA,W ; Fit MSB to 7 Bits
MOVWF CCPR1L ; Send MSB
RETURN
END

114
8.12 Parallel Slave Port
Port D can be operated as a “parallel slave port” which is a bi-directional,
microprocessor data-bus port controlled with the “/CS”, “/RD”, and “/WR” control lines.
These lines are shared with Port E and the TRISE register must configure all of the Port E
pins as “inputs”. Since Port E also shares its lines with the analog input channels, the
“ADCON1” register must be selected per Figure 8-1 so that the Port E lines are as
“digital”. When the PSP is selected, the TRISD may be ignored – it is not needed.
The PSP selection and status bits are located in the TRISE register. To select the
PSP, do “BSF TRISE,PSPMODE”. The PSP may generate interrupts upon a read or a
write.
In the discussion that follows about PSP “reads” and PSP “writes”, the “read” and
“write” are defined relative to the microprocessor that controls the PIC’s PSP. The
microprocessor “writes” data to the PIC’s PSP so that the PIC can “read” its data. That
is, a PSP “write” is for the PIC to get information from the PSP. The microprocessor
“reads” the PIC’s PSP so that the PIC can “write” data to the microprocessor. That is, a
PSP “read” is the way the PIC sends information to the microprocessor. Relative to the
PIC, a “write” in an “input” and a “read” is an “output”.
A “write” to the PSP occurs when the “/CS” and “/WR” lines are made low by the
external device (“microprocessor”). This causes the “IBF” flag in the TRISE register to
be set (=1), where “IBF” means “Input Buffer Full”. The “write” is completed when
either the “/CS” or the “/WR” are made high. This causes the interrupt flag, “PSPIF”, to
be set, which generates an interrupt, if it is enabled. The “IBF” flag is reset automatically
when Port D is read in software. The “IBOV” flag (“Input Buffer Overflow”) in the
TRISE is set if a second “write” is attempted before the first Port D data is read. The
“IBOV” flag must be reset in software.
A “read” from the PSP is for sending data from the PIC to the external device.
The user does this by writing data to Port D and waiting for the external device to pick it
up. Writing data to Port D causes the “OBF” flag in the TRISE register to be set (=1),
where “OBF” means “Output Buffer Full”. A “read” from the PSP occurs when the
“/CS” and the “/RD” lines are made low by the external device. When this occurs, the
“OBF” flag is reset immediately so that the PIC’s software “knows” that the external
device got the data. When either the “/CS” or the “/RD” lines are made high, the
interrupt flag, “PSPIF”, is set which generates an interrupt if it is enabled.
An example program that uses the PSP is as follows:
LIST P=16F877
IN_DATA: EQU 0x20 ; Data From PSP “Write”
OUT_DATA: EQU 0x21 ; Data To PSP “Read”

115
ORG 0x0000
INIT:
BANKSEL TRISE ; Bank 1
MOVLW 0x06 ; ADCON1 – Select “All Digital”
MOVWF ADCON1
MOVLW 0x17 ; Set PSP Mode, Port E=INPUTS
MOVWF TRISE
BANKSEL PORTD ; Bank 0
MAIN:
----- Get Data to Send -----
BTFSC TRISE,OBF ; Got Data?
GOTO NOT_READY_YET1
BTFSS PIR1,PSPIF ; “Read” Complete?
GOTO NOT_READY_YET1
MOVF OUT_DATA,W ; Ready to Send Next Data
MOVWF PORTD
BCF PIR1,PSPIF ; Reset INT Flag
GOTO WHEREVER1
----- Prepare to Get New Data -----
BTFSS TRISE,IBF ; New Data In?
GOTO NOT_READY_YET2
BTFSS PIR1,PSPIF ; “Write” Complete?
GOTO NOT_READY_YET2
MOVF PORTD,W ; Get Data
MOVWF IN_DATA
BCF PIR1,PSPIF ; Reset INT Flag
GOTO WHEREVER2
END

116
8.13 EEPROM Data Memory
The PIC has 256 bytes of EEPROM that the user can read and write using
software. This is ideal for long-term data since the EEPROM is a non-volatile memory.
The data EEPROM can support up to one-hundred-thousand “erase/write” data cycles.
The reading of the EEPROM data is fast but the worst-case writing-time for one byte is
eight milliseconds. A watch-dog timer induced reset can abort the writing process so it is
important to make sure the watch-dog timer is set with a time-out period greater than its
minimum of seven milliseconds.
Both the data read and data write subroutines can be done in cookbook fashion
and are as follows:
READ_EEPROM: ; Call with Address in W
BANKSEL EEADR ; Bank 2
MOVWF EEADR ; Set Address
BANKSEL EECON1 ; Bank 3
BCF EECON1,EEPGD ; Do “Data EEPROM”
BSF EECON1,RD ; Start “Read”
BANKSEL EEDATA ; Bank 2
MOVF EEDATA,W ; Get Data in W
RETURN
The data write subroutine uses two RAM locations: “ADDR” and “VALUE” to hold the
address and the data value to write, respectively. Reset the watch-dog timer and disable
all interrupts before calling this subroutine. This routine returns a value of zero in the W
register if the “write” was successful.
WRITE_EEPROM:
BTFSC EECON1,WR ; Is Previous “Write” Done?
RETLW 0x01 ; No, Return W = Non-Zero
MOVF ADDR,W ; Get Address
MOVWF EEADR ; Set Address
MOVF VALUE,W ; Get Data to Write
MOVWF EEDATA ; Set Data to Write
BCF EECON1,EEPGD ; Data EEPROM Select
BSF EECON1,WREN ; Do “Write Enable”
MOVLW 0x55 ; “Lock & Key” Combinations
MOVWF EECON2 ;(****** See Below ******)
MOVLW 0xAA
MOVWF EECON2

117
BSF EECON1,WR ; Start “Write” Process
NOP
BCF EECON1,WREN ; Disable “Writes”
RETLW 0x00 ; OK, Return W = Zero
The code sequence labeled “Lock & Key” may seem confusing at first. It is a
sequence of actions selected by Microchip to prevent the possibility of writing data to the
EEPROM by mistake. Using this code sequence makes writing data to the EEPROM a
conscious and deliberate act. There is nothing special about the code words that are sent
to the “EECON2” register. They were chosen arbitrarily by Microchip to do the job.
What matters is that these two codes must be sent in this order and one sent immediately
after the other. This sequence “unlocks” the “Write-to-EEPROM” process.
8.14 FLASH Program Memory
In a similar way to the Data EEPROM, the FLASH Program Memory can also be
read from and written to in software. However, the FLASH can only support one-
thousand “erase/write” data cycles. The worst-case writing-time is still the same as the
Data EEPROM at eight milliseconds. While the writing process is active, the program
memory cannot be accessed for the normal running of the software – the CPU “freezes”
until the “write” is complete. Again, you must reset the watch-dog and disable all of the
interrupts before calling the “writing” subroutine.
Also, since the FLASH program memory has 13-bit addresses and 14-bit data
words, two bytes, each, are needed to hold the address and data for a FLASH read or
FLASH write. Assume that the user’s registers that hold this data are as:
ADDRH:ADDRL --- For the Address
DATAH:DATAL --- For the Data.
Both the FLASH read and the FLASH write subroutines can be done in cookbook
fashion. They are as follows:
FLASH_READ:
MOVF ADDRL,W ; Get/Set Low Address
MOVWF EEADR
MOVF ADDRH,W
MOVWF EEADRH
BSF EECON1,EEPGD ; Select FLASH Memory
BSF EECON1,RD ; Start “Read” Process
NOP ; Delay two cycles
NOP

118
BANKSEL EEDATA ; Bank 2
MOVF EEDATA,W ; Get Data, Low & High
MOVWF DATAL
MOVF EEDATH,W
MOVWF DATAH
RETURN
FLASH_WRITE:
MOVF ADDRL,W ; Get/Set Address, High & Low
MOVWF EEADR
MOVF ADDRH,W
MOVWF EEADRH
MOVF DATAL,W ; Get/Set Data to Write
MOVWF EEDATA
MOVF DATAH,W
MOVWF EEDATH
BSF EECON1,EEPGD ; Select FLASH Memory
BSF EECON1,WREN ; “Write” Enable
MOVLW 0x55 ; “Lock & Key” Combination
MOVWF EECON2 ; (**** Same Idea as EEPROM **)
MOVLW 0xAA
MOVWF EECON2
BSF EECON1,WR ; Start “Write” Process
NOP ; Delay two cycles
NOP
BCF EECON1,WREN ; Disable “Writes”
RETURN
8.15 FLASH Code & Data EEPROM Protection
The CONFIGURATION word at address 0x2007 contains bit settings that block
read/write access to the FLASH program memory and the Data EEPROM. This is a
security function to prevent software piracy. These bits can be specified in the program
or in the down-load process. Other options allow restricted areas of FLASH memory to
be set-up.

119
8.16 The CONFIGURATION Word
The CONFIGURATION word is located at the address 0x2007 and may be
specified in the assembly language program or specified during the down-load process. It
has 14 bits and is broken-down as follows:
Bit 13, Bit 5, “CP1”
Bit 12, Bit 4, “CP0”
CP1:CP0 Code Protection (FLASH)
(1,1) Code Protect OFF
(1,0) 0x1F00 to 0x1FFF Protected
(0,1) 0x1000 to 0x1FFF Protected
(0,0) 0x0000 to 0x1FFF Protected
Both sets of bits (13,5) and (12,4) must have the same values.
Bit 11, “DEBUG”
= 1 = “In-Circuit Debugger Disabled”
= 0 = “Enabled”
Bit 9, “WRT”
= 1 = “UnProtected FLASH, Can be written under software control”
= 0 = “Cannot be written under software control”
Bit 8, “CPD”
= 1 = Data EEPROM Code Protect Off
= 0 = Code Protected
Bit 7, “LVP” = “Low Voltage In-Circuit Serial Programming Enable Bit”
= 1 = “Enabled”
= 0 = “Disabled”
Bit 6, “BODEN”
= 1 = “Brown-Out Reset Enabled”
Bit 3, “/PWRTE”
= 1 = “Power-Up Timer Enabled” = “Power-On Reset Enabled”
Bit 2, “WDTE”
= 1 = “Watch-Dog Timer Enabled”

120
Bit 1, Bit 0, “FOSC1” and “FOSC0”
FOSC1:FOSC0 Oscillator Selection Bits
(1,1) RC Oscillator (No External Components Needed)
(1,0) HS Oscillator (4 MHz to 20 MHz XTAL)
(0,1) XT Oscillator (200 kHz to 4 MHz XTAL)
(0,0) LP Oscillator (32 kHz to 200 kHz XTAL)
8.17 Sleep Modes & Reset Modes
The PIC has a power-saving mode where the CPU can be deactivated only to be
re-activated later but without losing RAM data or register data. This is called the “Sleep”
mode and it is initiated by executing the “SLEEP” instruction in the user’s software.
Processor “resets” and interrupts can take the PIC out of “Sleep” mode. These will be
discussed shortly.
There are several kinds of processor “resets” in the PIC. A list of these is as
follows:
1) Power-On Reset (POR)
2) /MCLR --- Normal Operation
3) /MCLR --- From Sleep
4) Watch-Dog Timer (WDT) --- Normal Operation
5) Watch-Dog Timer (WDT) --- From Sleep
6) Brown-Out Reset (BOR)
The “power-on reset” can be enabled to wait until the DC power has come up to a safe
level before initializing the PIC. Likewise, the “brown-out reset” can be enabled to look
for drops in the DC voltage powering the PIC and cause a reset to occur. The “power-
on” reset, the “brown-out” reset, and the watch-dog timer are all enabled or disabled from
the CONFIGURATION word. The “/MCLR” reset pin (pin 1) is held high with a pull-up
in normal operation.
Each of these resets has “flag” or “status” bits that allow the user to detect which
of these reset has occurred from software. These bits are located in the “PCON” and
“STATUS” registers. A table of their states and meanings is as follows:

121
/POR /BOR /TO /PD Meaning/ Significance
(0) (x) (1) (1) Power-On Reset
(0) (x) (0) (x) Illegal
(0) (x) (x) (0) Illegal
(1) (0) (1) (1) Brown-Out Reset
(1) (1) (0) (1) Watch-Dog Timer Reset
(1) (1) (0) (0) Watch-Dog Timer, Wake From Sleep
(1) (1) (u) (u) /MCLR Reset, Normal Operation
(1) (1) (1) (0) /MCLR or INT, Wake From Sleep
Where “x” = “Don’t Care” and “u” = “unchanged”. The program-counter’s value on a
reset is 0x0000. The PC’s value on a wake-up is the increment of its value when the PIC
entered “Sleep”.
The interrupts which can awaken the PIC from “Sleep” when they are enabled are
as follows:
1) External INT pin
2) Parallel Slave Port (Read/ Write)
3) Timer1
4) CCP1/CCP2 Special Events
5) Serial Peripheral Interface (Chapter 9)
6) SPI / I2C Slave Mode (Chapter 9)
7) USART (Chapter 9)
8) ADC
9) EEPROM Writes

122
Chapter 9: PIC Peripherals, Serial
Communications Ports
9.0 Chapter Summary
Section 9.2 discusses the USART in all of its modes. Section 9.3 covers the SPI
master mode. Section 9.4 covers the SPI slave mode. Section 9.5 covers the I2C in two
examples of its master and slave modes.
9.1 Introduction
The PIC has three major serial communications modes:
1) USART
The USART is a “Universal Synchronous/Asynchronous Receiver/Transmitter”.
It performs RS-232 serial communications with the IBM-PC serial port when it is
operated in its asynchronous mode. It can also work in a master or slave
synchronous mode but only in a half-duplex form.
2) Master Synchronous Serial Port, Serial Peripheral Interface
The MSSP/SPI mode is a simple 8-bit serial input/output used for working with
shift-registers and other simple serial interfaces. It is not used for RS-232. Like
the USART, it also has a master or slave mode.
3) Master Synchronous Serial Port, Inter-Integrated Circuit
The MSSP/I2C mode is a more complex serial communications mode that is
supported by many off-the-shelf integrated circuits. It is intended for more
complex systems where there are several master mode devices and many slave
mode devices. There is a complicated communications protocol that links each
master to each slave. It is not used for RS-232.
9.2 USART (Overview)
The USART performs RS-232 serial communications. This can be done with
another PIC, a microprocessor serial port, or the IBM-PC’s serial port.
The USART can operate with 8-bit data or 9-bit data. It may be configured to
work in one of three modes:

123
1) Asynchronous (Full-Duplex)
2) Synchronous, Master (Half-Duplex)
3) Synchronous, Slave (Half-Duplex)
The “TXSTA” register governs the status and control of the transmitter, while the
“RCSTA” register governs the same for the receiver. The data to transmit is placed in the
“TXREG” and the received data is placed in the “RCREG” register. There is also the
“SPBRG” register which is used to select the baud rate (“data clock-speed”).
Interrupts may be generated by both the transmitter and the receiver. These will
be shown later as they are needed.
9.2.1 USART (Asynchronous Mode, Full-Duplex)
The first step in using the USART in the asynchronous mode is to activate the
USART by setting (=1) the “SPEN” bit in the “RCSTA” register. The Port C pins, RC6
and RC7, must be configured with the TRISC register as “output” and “input”
respectively. Setting the “SPEN” bit enables the “TX” pin (RC6) as the transmitted data
output and the “RC” pin as the received data input.
Next, the “SYNC” bit of the “TXSTA” register must be cleared (=0) to select the
“Asynchronous” mode.
Eight-bit vs. nine-bit transmission and reception is done with the “TX9” bit of the
“TXSTA” register and the “RX9” bit of the “RCSTA” register, respectively. Setting
each (=1) enables the nine-bit operation while clearing them (=0) enables the eight-bit
operation. If the nine-bit operation is enabled, the ninth transmitted data bit, “TX9D”,
must be set-up in the “TXSTA” register, and the ninth received data bit, “RX9D”, can be
found in the “RCSTA” register.
The baud rate must be selected next. There are two speeds for the baud rate
model. The user enters a value “X” in the “SPBRG” register according to the formulas:
Low-Speed Baud Rate = Fosc / (64 * (X + 1))
High-Speed Baud Rate = Fosc / (16 * (X + 1)).
The “BRGH” bit in the “TXSTA” register selects between these “low-speed” and “high-
speed” models. If BRGH = 1, the “high-speed” is selected and BRGH = 0 selects the
“low-speed”. The selection of these speeds is somewhat arbitrary in that one or the other
may give a better approximation to the desired baud rate, and this is really what counts.
For example, suppose we want a baud rate of 19200 and the Fosc = 4 MHz. For
the “low-speed”, this gives SPBRG = 2 and the approximate baud rate = 20833, which
has a percent error of 8.5%. For the “high-speed”, this gives SPBRG = 12 and the

124
approximate baud rate = 19230 which has a percent error of 0.16%. Therefore, the “high-
speed” selection is the better choice.
Last, the receiver must be enabled by setting (=1) the “CREN” bit in the
“RCSTA” register.
The overall data format for the USART in the asynchronous mode is transmission
and reception with the least-significant-bit first, in a non-return-to-zero (NRZ) format,
with the bit sequence as “start-bit, data-bits, stop-bit”, and with no internally-generated
parity bits.
The transmitter is enabled by setting (=1) the “TXEN” bit in the “TXSTA”. To
send data for output, write the data to the “TXREG” register. (If the nine-bit mode is set,
the “TX9D” data bit must be set-up first and then write the other part, the byte, to the
“TXREG” register.
There is a “transmitter” interrupt flag, “TXIF”, which is set when the “TXREG”
register is empty and waiting for the next byte to send. That is, when a data byte is
placed in the “TXREG” register, the byte is shifted out, and, when the “TXREG” is
empty, the interrupt flag is set. The “TXIF” flag cannot be reset in software. It is only
reset by writing another byte to the “TXREG” register. If there is no more data to
transmit, the interrupt process can only be disabled by disabling the “TXIE” interrupt-
enable bit.
To receive a data byte, the “receive” interrupt flag, “RCIF”, is set when a new
byte is received, and will generate an interrupt if it is enabled. The “RCIF” bit is
automatically reset when the software reads the data from the “RCREG” data register.
(If the nine-bit mode is set, the “RX9D” data bit in the “RCSTA” register must be read
before reading the “RCREG” data.)
Another feature that is available in the nine-bit asynchronous mode is “automatic
address detection”. This is used for “multi-PIC” or “multi-device” operations where each
device has a nine-bit “address” and it will respond to the transmitting PIC only when the
address matches the address sent to it.
To select this mode of operation, set-up the PIC in nine-bit asynchronous mode
and set (=1) the “ADDEN” bit in the “RCSTA” register. Also enable the receiver
interrupt with the “RCIE” bit (RCIE = 1). When the interrupt occurs, read the nine data
bits, interpret them as an address, and see if the address matches the user-defined address.
If it does, clear (=0) the “ADDEN” bit to allow the software to read the data to follow.
The address is distinguished from the data by setting the ninth-bit of the address to one
(=1). The ninth-bit of the data will be zero (=0).
It should be noted that the asynchronous USART mode is halted if the CPU enters
the “Sleep” state.

125
There are two receiver-error condition bits that need to be discussed. These are
the “OERR” bit meaning “Overrun Error” and “FERR” meaning “Framing Error”. Both
of these bits are in the “RCSTA” register.
The “RCREG” register where the data comes in is double-buffered so that it can
store two bytes in succession. If a third data byte comes in, without the first two having
been read, the “OERR” bit will be set (=1) to indicate “Overrun Error”. When this
happens, the third data byte is lost and the whole receive process is inhibited. To correct
and reset this error, the “RCREG” register must be emptied and the “OERR” can be reset
only indirectly by doing:
BCF RCSTA,CREN
BSF RCSTA,CREN.
The “Framing Error” condition, indicated by the “FERR” bit is set (=1) if a
received data byte has an illegal “Stop” bit indicating that the data is illegal. This error
condition does not inhibit the receive process and does not need to be cleared. A new
“FERR” value will appear when the “RCREG” data register is read. Therefore, check the
“FERR” bit before reading the “RCREG” to get the current “FERR” value.
The example program that follows is for an 8-bit, asynchronous USART mode
with receiver interrupts, transmitter polling, and a baud rate of 19200.
LIST P=16F877
RCV_DATA: EQU 0x20 ; Received Data
READY: EQU 0 ; Flag: RCV Data Ready
FRAME: EQU 1 ; Flag: Frame Error
ORG 0x0000
GOTO INIT
ORG 0x0004
GOTO RCV_ISR
ORG 0x0006
INIT:
CLRF PORTC
BCF PIR1,RCIF ; Reset RCV INT Flag
MOVLW 0x80 ; Enable USART, 8-Bit
MOVWF RCSTA ; --- RCV Mode

126
BCF TRISC,6 ; RC6 = Out
BSF TRISC,7 ; RC7 = In
MOVLW D’12’ ; Set Baud Rate = 19200
MOVWF TXSTA
BSF PIE1,RCIE ; Set RCV Int Enable Bit
BSF INTCON,PEIE ; Enable Peripheral Ints
BSF INTCON,GIE ; Enable Global Ints
BSF RCSTA,CREN ; Enable RCV
BSF TXSTA,TXEN ; Enable XMTR
MAIN:
------ When Ready To Send Data ---------
------ Put Data in (W) -----------------------
MOVWF TXREG ; Send Data
WAIT:
BTFSS PIR1,TXIF ; Wait for XMT to Finish
GOTO WAIT
------ If Data is Received --------------------
BTFSS FLAGS,READY ; Is Data Ready to Read?
GOTO NO_DATA_YET
MOVF RCV_DATA,W ; Get the Data
BCF FLAGS,READY
RCV_ISR:
------ Save Registers ------- (See Appendix F)
BCF FLAGS,FRAME ; Reset Previous FERR
BTFSC RCSTA,FERR ; Check for Frame Error
BSF FLAGS,FRAME ; Set “FRAME” if FERR=1
MOVF RCREG,W ; Get New RCV Data
MOVWF RCV_DATA ; Store in User’s Data
BSF FLAGS,READY ; Set User’s Data Ready
------ Restore Registers ---------------------
RETFIE ; Return From Int
END

127
9.2.2 USART (Synchronous, Master Mode)
Both the synchronous master mode and the synchronous slave mode run the
USART as “half-duplex” meaning that the data cannot be transmitted and received at the
same time. The primary difference between the master mode and the slave mode is that
the master mode generates the serial clock while the slave mode receives the external
serial clock.
Many features, bits, and conditions of the synchronous master mode are similar to
the asynchronous mode. This section will illustrate only the differences between these
modes.
The baud rate generator works with a model and formula different from the
asynchronous mode and the “BRGH” bit, which was the speed control, has no meaning.
The baud rate is:
Baud Rate = Fosc / (4 * (X + 1))
Where “X” is the value written into the “SPBRG” register. This allows for much higher
baud rates.
The “Sleep” mode halts the synchronous master mode. The “Address Detection”
mode is not available in the synchronous master mode. The Port C pins, RC6 and RC7,
are set up as before. The clock (CK) is sent on RC6 while the data is both sent and
received one way at a time, on RC7 (DT).
The “CSRC” bit and the “SYNC” bit, both of the “TXSTA” register, must be set
(=1) for the synchronous master mode.
The operation of the USART in the synchronous master mode is more
complicated due to the restriction of doing only half-duplex communications. The
“TXEN” bit of the “TXSTA” register and the “CREN” bit of the “RCSTA” register must
not both be set (=1) at the same time – only one or the other is to be set (=1) at any one
time. To transmit data, the “TXEN” bit must be set (=1) and the “CREN” bit must be
cleared (=0). To receive data, the “TXEN” bit must be cleared (=0) and the “CREN” bit
must be set (=1). All other things are equal. In the receive mode, the data is sampled on
the falling edge of the clock and, in transmit mode, the data is shifted on the rising edge
of the clock and is stable on the falling edge.
One exception is the possibility of receiving one single byte during the
transmission of (typically) a stream of transmitted data bytes. This is done by setting the
“SREN” bit (=1) of the “RCSTA” register while the “TXEN” bit of the “TXSTA”
register is set (=1). When a single byte is received, the “SREN” bit is cleared (=0)
automatically.

128
9.2.3 USART (Synchronous, Slave Mode)
This section will illustrate the differences between the synchronous slave mode
and the other two USART modes.
Since the synchronous slave mode does not generate a serial data clock, but
receives it externally, there is no need to set-up the “SPBRG” register to generate baud
rates. The Port C pins, RC6 and RC7, must both be set-up as “inputs” using the TRISC”
register. The “CSRC” bit of the “TXSTA” register must now be cleared (=0) for the
synchronous slave mode.
The “single byte receive” option using the “SREN” bit is disabled in the slave
mode. The “either-or” nature of the “TXEN” and “CREN” bits is still the same as in the
synchronous master mode.
The reception or transmission of data in the slave mode can awaken the CPU from
“Sleep”.
9.3 Serial Peripheral Interface (SPI, Master Mode)
The SPI master mode allows 8-bit data to be synchronously transmitted and
received at the same time. This data transfer happens on three pins as follows:
Serial Data Out = SDO = RC5 = Output
Serial Data In = SDI = RC4 = Input
Serial Clock = SCK = RC3 = Output.
Where the Port C pins, RC5, RC4, and RC3, are set-up with the TRISC register as above.
The “SSPSTAT” and “SSPCON” registers are used to control the SPI module,
while the “SSPBUF” register is used for the input and output data.
The first step in setting-up the SPI in master mode is to set the lower four bits of
the “SSPCON” register as follows:
SSPM3:SSPM0 Function / Meaning
(0,0,0,0) SPI, Master, SCK = Fosc / 4
(0,0,1,1) SPI, Master, SCK = Timer2 / 2

129
Next, the “CKP” bit in the “SSPCON” register must be set-up along with the “CKE” and
“SMP” bits of the “SSPSTAT” register. This is done according to the waveforms of
Figure 9-1. An example of the issues involved with this selection will be given later.
bit 7
bit 7
bit 7
bit 6 bit 5 bit 4 bit 3 bit 2 bit 1 bit 0
bit 0
bit 0
SCK(CKP = 0, CKE = 0)
SDO
SDI(SMP = 0)
SDI(SMP = 1)
SSPIF
Figure 9-1 Master Mode SPI Mode Timing
The last step is to set (=1) the “SSPEN” bit of the “SSPCON” register to enable
the MSSP/SPI module.
Sending data for output (on the SDO line) is done by writing the data to the
“SSPBUF” register. Getting data from the input (on the SDI line) is done by doing a
“dummy” write or a “real” write to the “SSPBUF” register. It is impossible to read data
for input without writing data for output! When the data transfer process is finished, the
“SSPIF” interrupt-flag is set (=1). If the corresponding interrupt-enable bit, “SSPIE”, is
set (=1), an interrupt will be generated.
Now let’s look at some hardware interfacing issues and the selection of bits per
Figure 9-1. A simple input/output circuit is shown in Figure 9-2. A “serial-in, parallel-
out” (SIPO) shift-register (74HC164) is used to capture the SPI’s serial output data. A
“parallel-in, serial-out” (PISO) shift-register (74HC165) is used to feed the SPI’s serial
input pin (SDI). Notice that the “clock” or “shift” input of the 74HC165 runs from the

130
SPI’s clock (SCK) through an inverter. Both of the shift-registers “clock” on the rising-
edge.
(8) (8)PIC16F877
74HC164 (SIPO) 74HC165 (PISO)
1 Reset CKI
(8) (8)PIC16F877
74HC164 (SIPO) 74HC165 (PISO)
1 Reset CKI
SDO
SDO
RB0
RB0
Shift/
Load
Shift/
Load
SDI
Data In
Data In
SCK
CK
CK
Data
Out
SDI
CK
CK
SCK
Data
SDI
SDI
OR
Figure 9-2 Serial-Out/Serial-In with the 74HC164 & 74HC165
The selection bits for this circuit are:
CKP = 0, CKE = 1, and SMP = 0.
This produces the second “SCK” waveform from the top of Figure 9-1. This is ideal for
the “output” part since the “SDO” data is stable when the “SCK” makes a rising edge. If
the 74HC165 is used without the inverter on its clock, a potential problem exists. The
rising edge would then be used to sample the input data and shift the shift-register at the
same time! In general this is a bad design practice. The problem can be fixed by using
the inverter as above, or by selecting “SMP = 1” and feeding the 74HC165’s serial output
line back into its serial input line. The latter would cause the data to be shifted an extra
space and the PIC would have to rotate the bit-pattern back into its proper place.
Another circuit situation is shown in Figure 9-3. This circuit uses an “AND” gate
to inhibit the clock-input of a SIPO, 74HC164 shift-register to prevent data from being
output to it while data from the 74HC165 is being read in.

131
AND
RB0
SCK
PIC16F877
74HC00 (8)
74HC164 (SIPO)
CK
1
Reset
Data
SDO
CKI
Serial
Data
Input
74HC165 (PISO)
(8)
CK
SDI
Data Out
RB1
Shift/Load
Figure 9-3 Serial-Out/Serial-In with Gated Clock
To Inhibit Serial Output
Since both the 74HC164 and the 74HC165 have serial data inputs, they may be
cascaded in series (serially) for multiple-byte inputs and outputs.
An example program that uses the SPI master mode and Figure 9-3 is as follows:
LIST P=16F877
DATA_RCV: EQU 0x20 ; Storage for RCV Data
OUT_ENABLE: EQU 0 ; Port B, RB0 – Enable Out
SHIFT_LOAD: EQU 1 ; Port B, RB1 =
; 1 = SHIFT
; 0 = LOAD
ORG 0x0000
INIT:
MOVLW 0x10 ; RC5 = RC3 = Out
MOVWF TRISC ; RC4 = In
CLRF TRISB ; Port B = Outputs
BCF SSPSTAT,SMP ; SMP = 0
BSF SSPSTAT,CKE ; CKE = 1

132
CLRF SSPCON ; SPI, Master, Fosc / 4
BSF SSPCON,SSPEN ; Enable SPI
MAIN:
------- Put Data to Send in W ---------
CALL SPI_SEND ; Do Data Read / Write
CALL SPI_READ ; Do Data Read (No Write)
SPI_READ:
BCF PORTB,OUT_ENABLE ; Disable Write
GOTO SPI_SKIP
SPI_SEND:
BSF PORTB,OUT_ENABLE ; Enable Write
SPI_SKIP:
BCF PORTB,SHIFT_LOAD ; “Load”
BSF PORTB,SHIFT_LOAD ; “Shift”
BCF PIR1,SSPIF ; Reset Int Flag
MOVWF SSPBUF ; “Send” Data
SPI_WAIT:
BTFSS PIR1,SSPIF ; Wait Until Done
GOTO SPI_WAIT
MOVF SSPBUF,W ; Get RCV Data
MOVWF DATA_RCV ; & Save It
BCF PIR1,SSPIF ; Reset Int Flag
RETURN
END
9.4 Serial Peripheral Interface (SPI, Slave Mode)
There are many similarities between the SPI master mode and the SPI slave mode.
Only the differences will be stated here. The main difference is that the “SCK” serial-
clock line does not generate the clock but receives it as an input on the “SCK” line. Also,
there is a fourth serial port line, “/SS”, meaning “slave-select” which can be used to
activate or deactivate the slave SPI unit, if the option to use it is enabled.

133
The summary of the SPI serial pins in slave mode is:
Serial Data Out = SDO = RC5 = Output
Serial Data In = SDI = RC4 = Input
Serial Clock = SCK = RC3 = Input
/Slave Select = /SS = RA5 = Input (TRISA & ADCON1)
The lower four bits of the “SSPCON” register now select as:
(0,1,0,0) SPI, Slave, /SS Enabled
(0,1,0,1) SPI, Slave, /SS Disabled
The “CKP” and “CKE” bits are selected according to the waveforms in Figure 9-4
and Figure 9-5. The “SMP” bit must always be zero (=0). Also if “CKE” is set (=1) then
the “slave select” (/SS) must be enabled.
bit 7
bit 7
bit 0
SDO
SDI(SMP = 0)
SSPIF
SS (Optional)
SCK(CKP = 0)
SCK(CKP = 1)
Figure 9-4 SPI Mode Timing (Slave Mode, CKE = 0)

134
bit 7
bit 7
bit 0
SDO
SDI(SMP = 0)
SSPIF
SS
SCK(CKP = 0)
SCK(CKP = 1)
Figure 9-5 SPI Mode Timing (Slave Mode, CKE = 1)
If the “slave select” pin is enabled, the controlling device must set this pin “low”
(= 0 = Ground) to activate the slave PIC’s SPI module. If the “slave select” line is
enabled and is held “high” (= 1 = +5 Volts), the slave PIC will ignore the “SCK” input
clock and none of its data will be transferred.
All of the other features of the SPI master mode are the same as in the SPI slave
mode.
9.5 I2C System Overview
The I2C system communicates on the “SCL” and “SDA” pins which are shared
with the Port C, pins RC3 and RC4, respectively. The “SCL” is the data clock and the
“SDA” is the data line. These pins must both be set-up as “inputs” with the TRISC
register before configuring the I2C modes. Furthermore external pull-up resistors
(minimum resistance is 1.7 K-Ohms) must be attached to each of these pins for the proper
operation of the I2C module. All of the above is true for the I2C module in all of its
modes, master or slave.

135
In the I2C system each slave device has a unique “address” code which identifies
it for the master device when the master device wants to access it. Many slave devices
may be used in the system and there may be several master devices, too. They all share
the same “SCL” and “SDA” lines which are “open-drain” outputs along with sensors for
“inputs”. This makes bi-directional information transfers possible at any time.
The slave’s addresses may be either 7-bits long or 10-bits long depending on the
devices to be used or its settings. The PIC I2C module supports both lengths in either of
the master or slave modes.
The slave devices in the I2C system cannot initiate a data transfer to the
master(s). It can only read or write data to the master when the master calls it and only
one master can do this at any one time.
Let’s look at the I2C communication process in its most general form. Suppose,
for example, for simplicity we are working with 7-bit slave addresses.
Suppose that the master device is the PIC in the I2C system operating in the
master mode and it wants to write data to a slave. The master PIC first looks for activity
on the “SDA” and “SCL” lines to see if another master device is using the system. If and
when there is no activity, the master sends out nine (9) clock pulses on the “SCL” line
and sends out eight (8) bits on the “SDA” line. The eight bits are the 7-bit slave address
and a “read/write” bit which is set (=1) for a “read” or reset (=0) for a “write”. Here it is
reset (=0). Then the master PIC holds the “SDA” line “high” (=1) and looks for an
“acknowledge” state from the slave also on the “SDA” line (this is possible since the
“SDA” line uses “open-drain” outputs). If the slave does not acknowledge the master, it
will hold the “SDA” line “high” (=1) and the master knows that the slave refuses the
master’s attempt to “write”. If the slave does acknowledge the master, it will bring the
“SDA” line to ground, or “low” (=0), and the master will proceed to send its data (more
on this later).
When the slave device (suppose it is a PIC in slave mode) receives the 7-bit
address and the “read/write” bit it checks the address contained in the “SSPADD”
register which was set-up by the slave’s user’s software. If there is a match, an
“acknowledge” is sent; if not, no “acknowledge” is sent.
If the slave gives an “acknowledge” to the master, the master sends out nine (9)
“SCL” clock pulses and the eight data bits to be sent on the “SDA” line. The ninth clock
pulse is used to sense another “acknowledge” state from the slave. If the slave does not
“acknowledge” the master, the data “write” process is stopped, and both the master and
the slave go into an “idle” state. If there is an “acknowledge” from the slave, the above
process repeats, but sending a new address is not necessary.
Suppose that the master wants to get data (“read”) from the slave. The whole
process is nearly identical, but with three (3) exceptions. The first is that the “read/write”
bit is set (=1) for doing a “read”. The second is that when the slave gives the

136
“acknowledge” when it gets its address code from the master, the slave holds the “SCL”
line “low” (=0) to inhibit the master from sending out clock pulses on the “SCL” line.
These are the clock pulses the master uses to synchronize the data “read” from the slave.
The master is inhibited until the slave is ready to send its data. When it is ready, it
releases the “SCL” line by letting it go “high” (=1). The third is that master is now the
one that sends the “acknowledge” out on the ninth master clock pulse for the slave to
receive. If there is an “acknowledge” from the master, the slave gets ready to send
another byte. Again, sending a new address is not necessary.
In both the “write” and “read” processes a “useless” or “dummy” byte is sent after
the last “valid-data” byte “acknowledge” is sent so that it can be refused and the data
transfer process will then stop.
The I2C module uses five user-accessible registers for its data, status, and control
functions:
1) SSPCON --- Control #1
2) SSPCON2 --- Control #2
3) SSPSTAT --- Status
4) SSPBUF --- Transmit and Receive Data
5) SSPADD --- Slave Address or Master Baud-Rate Setting
The first step in setting-up the I2C module is to set-up Port C pins RC3 and RC4
as “inputs” by using the TRISC register. Next, set-up the lower four bits of the
“SSPCON” register as:
(0,1,1,0) I2C Slave Mode, 7-Bit Address
(0,1,1,1) I2C Slave Mode, 10-Bit Address
(1,0,0,0) I2C Master Mode
(1,0,1,1) I2C Firmware Controlled Master Mode / Slave Idle
(1,1,1,0) I2C FCMM, 7-Bit Addr, Start/Stop Interrupts
(1,1,1,1) I2C FCMM, 10-Bit Addr, Start/Stop INTs
In the master mode, set-up the baud rate as “X” in the “SSPADD” register as:
Baud Rate = Fosc / (4 * (X + 1))
Where the baud rate is usually one of 100 kHz, 400 kHz, or 1 MHz.
In either master or slave modes, set-up the “CKE” bit in the “SSPSTAT” register
as:
0 = I2C Input Levels
1 = SMBus Input Levels.

137
Also, set-up the “SMP” bit which is also in the “SSPSTAT” register, to control the slew-
rates of the “SCL/SDA” pins as:
1 = Slew-Rate Disabled (100 kHz or 1 MHz baud rates)
0 = Slew-Rate Enabled (400 kHz baud rate).
Last, set (=1) the “SSPEN” bit in the “SSPCON” register to enable the I2C serial port
module.
9.5.1 I2C Slave Mode
The great complexity of the I2C communications module makes it difficult to
describe not only the available options but the sequences in which they are used. This is
best seen with an example program using the 7-bit I2C in slave mode.
LIST P=16F877
FLAGS: EQU 0x20 ; User’s Flags
RCV_DATA: EQU 0x21 ; Hold for User’s Rcved Data
ADDR_LOW: EQU 0x3D ; Slave’s Address (Arbitrary Here)
ACCEPT: EQU 0 ; Flags: Bit 0 = “Accept Next Data”
ORG 0x0000
GOTO INIT
ORG 0x0004
GOTO SLAVE_ISR
ORG 0x0006
INIT:
CLRF FLAGS ; Reset User’s Flags
BSF TRISC,3 ; Make RC3 & RC4 “Inputs”
BSF TRISC,4
BCF SSPCON2,GCEN ; Disable “General Call”
; (This would reserve Address = 0 as
; a cause for an interrupt, if enabled)
MOVLW 0x06 ; Select “7-Bit Slave” Mode
MOVWF SSPCON

138
BANKSEL SSPSTAT ; Bank 1
BCF SSPSTAT,SMP ; Enable Slew-Rate Control
BCF SSPSTAT,CKE ; Set I2C Signal Levels
MOVLW ADDR_LOW ; Set The Slave’s Address
MOVWF SSPADD
BCF PIR1,SSPIF ; Reset I2C Int Flag
BSF PIE1,SSPIE ; Enable I2C Interrupts
BSF INTCON,PEIE ; Enable Peripheral Interrupts
BSF SSPCON,SSPEN ; Enable I2C MSSP Module
BSF INTCON,GIE ; Enable Global Interrupts
MAIN:
--- Do Program & Wait to be Interrupted ----
SLAVE_ISR:
--- Save Registers ---(See Appendix F)---
BTFSC SSPSTAT,DA ; Is Word = Address?
GOTO DO_DATA ; --- No, Process the Data
BTFSC SSPSTAT,RW ; --- Yes, Is This a “Write” Op?
GOTO WRITE_DATA ; --- Yes, Do Write
BANKSEL PORTC ; --- No, Prepare To Get Data
BSF FLAGS,ACCEPT ; Set Flag to Accept Next
; Data
MOVF SSPBUF,W ; Discard Address Transmitted
BCF SSPCON,SSPOV ; Reset Overflow Flag
ISR_RETURN:
BCF PIR1,SSPIF ; Reset I2C Interrupt Flag
--- Restore Registers ----
RETFIE ; Return From Interrupt
WRITE_DATA:
MOVF SSPBUF,W ; Discard Address Transmitted
BCF SSPCON,CKP ; SCL = 0, Inhibit Master
CALL GET_DATA_TO_SEND ; W = Data
MOVWF SSPBUF ; Send Data
BSF SSPCON,CKP ; Release Master
GOTO ISR_RETURN

139
DO_DATA:
BTFSC FLAGS,ACCEPT ; Accept This Data?
GOTO YES_ACCEPT
MOVF SSPBUF,W ; Discard The Data
BCF SSPCON,SSPOV ; Reset Overflow (Send
; ACKN)
GOTO ISR_RETURN
YES_ACCEPT:
MOVF SSPBUF,W ; Get The Data
MOVWF RCV_DATA ; Store Data in User’s Data Hold
BCF FLAGS,ACCEPT ; Refuse Next Data
BSF SSPCON,SSPOV ; Set Overflow, No ACKN
; For Next Time
GOTO ISR_RETURN
END
9.5.2 I2C Master Mode
As in the slave mode, the operation of the master mode is best seen with an
example program. Assume that we are using the “I2C Firmware Controlled Master Mode
with 7-Bit Addresses and Start/Stop Interrupts Enabled” at a baud rate of 100 kHz.
LIST P=16F877
FLAGS: EQU 0x20 ; User’s Flags #1
FLAGS2: EQU 0x21 ; User’s Flags #2
RCV_DATA: EQU 0x22 ; User’s Rcved Data Hold
XMT_DATA: EQU 0x23 ; Hold For Data to Transmit
CALLED_SLAVE: EQU 0x24 ; Current Slave’s Address
BAUD: EQU 0x09 ; Value for Baud Rate = 100 kHz
SLAVE1: EQU 0x78 ; First Slave’s Address (Arbitrary)
SLAVE2: EQU 0x6E ; 2nd
Slave’s Address (Arbitrary)
SLAVE3: EQU 0x2C ; 3rd
Slave’s Address (Arbitrary)
; FLAGS -----------------------------------------------------------------
DUMMY_WRITE: EQU 0 ; Do a Dummy Write
START_OK: EQU 1 ; “Start” Process is OK
ADDR_ACK EQU 2 ; Address is ACKNed

140
MASTER_ACK: EQU 3 ; Master’s ACKN is Given
; FLAGS2 -----------------------------------------------------------------
BYTE_FOR_XMT: EQU 0 ; Main Program is Attempting XMT
GET_A_BYTE: EQU 1 ; Main is to Receive a Byte
ADDR_ERROR: EQU 2 ; Address Error – Slave Refused
XMT_DATA_OK: EQU 3 ; Transmitted Data Process OK
WRITE_ERROR: EQU 4 ; Slave Refuses Data-Write
RCV_OK: EQU 5 ; RCV Data Process OK
BUS_COL: EQU 6 ; Bus Collision Error Occurred
;------------------------------------------------------------------------------
ORG 0x0000
GOTO INIT
ORG 0x0004
GOTO INT_RESPONSE
ORG 0x0006
INIT:
CLRF FLAGS ; Reset User’s Flags #1
CLRF FLAGS2 ; Reset User’s Flags #2
CLRF INTCON ; Reset Main INT Flags
BSF TRISC,3 ; Make RC3 & RC4 “Inputs”
BSF TRISC,4
MOVLW 0x0E ; 7-Bit FCMM Start/Stop Ints
MOVWF SSPCON
BCF SSPSTAT,CKE ; I2C Signal Levels
BSF SSPSTAT,SMP ; Slew-Rate Control Disabled
MOVLW BAUD ; Set Baud Rate = 100 kHz
MOVWF SSPADD
BCF PIR1,SSPIF ; Reset I2C Interrupt Flag
BCF PIR2,BCLIF ; Reset Bus Collision Int Flag
BSF PIE1,SSPIE ; Enable I2C Interrupts
BSF PIE2,BCLIE ; Enable Bus Collision Interrupts
BSF INTCON,PEIE ; Enable Peripheral Interrupts
BSF SSPCON,SSPEN ; Enable I2C MSSP Module
BSF INTCON,GIE ; Enable Global Interrupts

141
MAIN:
----- Run the Main Program and Look at the FLAGS2 byte -----
----- This is the “Macro-Status” for the I2C Process --------------
To Write to a Slave Device:
1) Check if “BYTE_FOR_XMT” and “GET_A_BYTE” are
clear
2) Get the Slave’s address and put it in “CALLED_SLAVE”
3) Get the byte for to send, put it in “XMT_DATA”
4) Set the “BYTE_FOR_XMT” flag
5) Call the “SET_START” subroutine
6) Check if “XMT_DATA_OK” flag is set and no error flags
are set (Data Sent OK). Then Reset “FLAGS2”.
7) Possible Errors Are:
a) ADDR_ERROR --- Slave does not accept address
given
b) WRITE_ERROR --- Slave does not accept the data
c) BUS_COL --- A Bus Collision occurred.
To Read from a Slave Device:
1) Check if “BYTE_FOR_XMT” and “GET_A_BYTE” are
clear
2) Get the Slave’s address and put it in “CALLED_SLAVE”
3) Set the “GET_A_BYTE” flag
4) Call the “SET_START” subroutine
5) Check if “RCV_OK” flag is set and no error flags are set
(Received Data OK). Get the Data in “RCV_DATA”.
Then Reset FLAGS2.
6) Possible Errors Are:
a) ADDR_ERROR --- Slave does not accept address
given
b) BUS_COL --- A Bus Collision Occurred.
;------------------------------------------------------------------------------------------
INT_RESPONSE:
---- Save Registers -----(See Appendix F)----
CHECK_INT_FLAGS:
BTFSC PIR2,BCLIF ; Bus Collision?
GOTO BUS_COLLISION
BTFSC PIR1,SSPIF ; I2C Interrupt?
GOTO DATA_COMMUNICATIONS
---- Restore Registers ------
RETFIE ; Return from Interrupt

142
BUS_COLLISION:
MOVF SSPBUF,W ; Discard SSPBUF Data
BCF SSPCON,WCOL ; Reset Error Flags (In I2C)
BCF SSPCON,SSPOV
BANKSEL SSPCON2 ; Bank 1
BSF SSPCON2,PEN ; Send “Stop” Condition Signal
BSF FLAGS2,BUS_COL ; Set User Flag: Bus
; Collision
BCF PIR2,BCLIF ; Reset All INT Flags
BCF PIR1,SSPIF
GOTO CHECK_INT_FLAGS
;-----------------------------------------------------------------------------------------
DATA_COMMUNICATIONS:
BTFSS FLAGS2,BUS_COL ; Was Previous Bus
; Col?
GOTO CHECK_STATUS2 ; No
CLRF FLAGS ; Return, Main Software Must Clear
BCF PIR1,SSPIF ; The Previous Bus Collision
GOTO CHECK_INT_FLAGS ; Condition Before New
; Communications Will Start.
CHECK_STATUS2:
BTFSS SSPSTAT,S ; “Start” Condition Occurred?
GOTO CHECK_STATUS3
BCF SSPSTAT,S ; Reset The “Start” Status flag
BTFSS FLAGS2,BYTE_FOR_XMT ; Data Transmit?
GOTO CHECK_FOR_RECEIVE
SEND_SLAVE_ADDRESS:
MOVF CALLED_SLAVE,W ; Do XMT, Send Slave
; Addr
MOVWF SSPBUF
BSF FLAGS,START_OK ; “Start” is OK
RESET_RET:
BCF PIR1,SSPIF
CHECK_FOR_RECEIVE:

143
BTFSS FLAGS2,GET_A_BYTE ; Data Receive?
GOTO RESET_RET
BSF SSPCON2,RCEN ; Enable Receive
GOTO SEND_SLAVE_ADDR
;----------------------------------------------------------------------------------------
CHECK_STATUS3:
BTFSS SSPSTAT,P ; “Stop” Condition Occurred?
GOTO CHECK_STATUS4
DO_A_RESET:
BCF SSPSTAT,P ; Reset “Stop” Condition Flag
BCF SSPCON,WCOL ; Reset I2C Error Flags
BCF SSPCON,SSPOV
CLRF FLAGS
BCF PIR1,SSPIF
;-----------------------------------------------------------------------------------------
CHECK_STATUS4:
BTFSS FLAGS2,BYTE_FOR_XMT
GOTO CHECK_STATUS5
BTFSC SSPSTAT,DA ; Data or Address RCVed?
GOTO DO_DATA_PROCESS1
BTFSS SSPSTAT,ACKSTAT ; Get ACKN?
GOTO NO_ACKN1
BSF FLAGS,ADDR_ACK
BCF SSPCON,SSPOV ; Reset Overflow
MOVF XMT_DATA,W ; Get Data To Send
MOVWF SSPBUF ; Send Data
BCF PIR1,SSPIF

144
NO_ACK1:
BSF FLAGS2,ADDR_ERROR
BSF SSPCON2,PEN ; Send “Stop” Condition
BCF PIR1,SSPIF
;----------------------------------------------------------------------------------------
DO_DATA_PROCESS1:
BTFSS SSPSTAT,ACKSTAT ; Get ACKN?
GOTO NO_ACKN2
BSF FLAGS2,XMT_DATA_OK
BSF FLAGS,DUMMY_WRITE
BSF SSPCON,SSPOV ; No ACKN Next Time
MOVWF SSPBUF ; Dummy Write
BCF PIR1,SSPIF
NO_ACKN2:
BTFSS FLAGS,DUMMY_WRITE
GOTO NO_DUMMY
DO_DUMMY_END:
BCF SSPCON,SSPOV
CLRF FLAGS
BSF SSPCON2,PEN ; Send “Stop” Condition
BCF PIR1,SSPIF
NO_DUMMY:
BSF FLAGS2,WRITE_ERROR
GOTO DO_DUMMY_END

145
;----------------------------------------------------------------------------------------
CHECK_STATUS5:
BTFSS FLAGS2,GET_A_BYTE
GOTO DO_A_RESET
BTFSC FLAGS,MASTER_ACK
GOTO DO_DUMMY_END
BTFSC SSPSTAT,DA ; Data or Address?
GOTO DO_DATA_PROCESS2
BTFSS SSPCON2,ACKSTAT ; Get ACKN?
GOTO NO_ACKN1
MOVF SSPBUF,W ; Discard Address Received
BSF FLAGS,ADDR_ACK
BCF SSPCON,SSPOV
BSF SSPCON2,RCEN ; Enable RCV
BCF PIR1,SSPIF
;---------------------------------------------------------------------------------------
DO_DATA_PROCESS2:
MOVF SSPBUF,W ; Get Data RCVed
MOVWF RCV_DATA ; Store Data in User’s Data Hold
BSF FLAGS2,RCV_OK
BSF MASTER_ACK
BSF SSPCON2,ACKDT ; Send “No ACKN”
BSF SSPCON2,ACKEN
BCF PIR1,SSPIF
;----------------------------------------------------------------------------------------

146
SET_START:
BSF SSPCON2,SEN ; Set a “Start” Condition
RETURN
;------------------------------------------------------------------------------------------
END

147
Chapter 10: DSP Fundamentals
10.0 Chapter Summary
Section 10.2 looks at a simple low pass filter as a moving average. Section 10.3
looks at a similar high pass filter. Section 10.4 gives a short discussion of digital filters
in the most general sense. Section 10.5 discusses aliasing and the Nyquist Sampling
Theorem. Section 10.6 is a cookbook example of a practical low pass/high pass filter.
Section 10.7 is a cookbook example of a practical band pass filter. Section 10.8 covers
the concept of a median filter. Section 10.9 describes DTMF decoding using a series of
standard band pass filters. Section 10.10 describes DTMF decoding by the deliberate use
of aliasing. Section 10.11 discusses how DSP can be used for speech compression and
sound effects.
10.1 Introduction
Digital Signal Processing (DSP) is usually thought of as a modern idea due to the
availability of cheap computers, but its roots can be traced back to the 17th century. The
main idea of DSP is to use equations to manipulate signals. That is, signals such as
speech or communications waveforms, can be sampled at regular intervals of time,
converted from voltage levels to numbers (ADC), and used as a sequence of these
numbers, to feed a computer to run calculations on them with a given, user-defined
equation. The resulting numbers are then converted back into voltage levels (DAC) to
use as a new signal.
DSP techniques can be used to filter, detect, classify, encode, decode, and remove
noise from signals. Why would anyone want to use DSP when other filters, detectors,
and classifiers are already available? A DSP filter is an algorithm that runs on a cheap
computer. It is programmable and it is flexible in that its program can be changed. For a
given algorithm or program it is ultra-stable. The DSP filter does not change with
temperature, component aging, component tolerances, and it is very resistant to errors in
manufacturing. Part of its flexibility is in its adaptability. A much slower process can be
run at the same time as the main filter program to measure the performance of the filter
and then over time change the filter program to make it do a better job of filtering.
Today when computers, ADCs, and DACs are very cheap, DSP is a very attractive idea!

148
10.2 An Example: A Low-Pass Filter
Let’s see how a simple filter can be constructed with DSP techniques and show
that it is possible to do filtering by doing calculations alone. Suppose that there is an
ADC that takes 8-bit data sample-values at a regular rate of 20kHz and suppose that the
signals we want to filter are sine waves.
Let the filter calculations be done this way: Let 16 ADC samples be stored in
RAM in the order they were received. When a new sample comes in, it is also stored in
RAM but the oldest of the 16 previous samples gets discarded. That is, at any time, there
are only the 16 most recent samples, including the current sample stored in RAM. Let
the “output” of the filter be the average of these 16 samples. That is, add up all of the
samples together and divide the sum by 16 to get one filter “output” point or sample.
This process is called a “moving-average”.
Does this process do “low-pass” filtering? Let’s see. Suppose that the input test
signal is a sine wave at a frequency of 10 Hz. Since the sampling rate (or “the data rate”)
is 20 kHz, there are 2000 sample points in one cycle of the 10 Hz wave. Only 16 of these
at any one time are used in the moving-average. Most of the 16 points have roughly the
same values as each other. There are, at times, some differences between the points since
the wave is changing over time, but at 10 Hz this change is very slow relative to the
sampling rate. The filter output, or average, will be only slightly less than any one of the
16 sample values. That is, a “low” frequency will “pass” unchanged.
What happens if the input test signal is a sine wave with a frequency of 8500 Hz?
Since the sampling rate is 20 kHz, there are at most only two successive samples that are
“both positive” or “both negative” at any one time. If 16 successive samples were added
together, most of them will cancel each other out. This sum will be made even smaller
when it is divided by 16. So, for this “high” frequency, the filter output will be very
small at any one time. A low-pass filter will attenuate high frequencies.
So, we have a low-pass filter just by doing the moving-average calculation!
10.3 An Example: A High-Pass Filter
This example is crude and more difficult to see, but with a relaxed view it can be
seen by approximation.
Suppose that the moving-average is the same as in our high-pass filter except that,
now instead of adding the 16 samples together they are alternately added to and
subtracted from each other (and then the sum/difference result is divided by 16). That is,
add the first, subtract the second, add the third, … and so on.

149
If our input test signal is the 10 Hz sine wave, most of these samples were nearly
the same and consequently forming the new sum/difference will cancel these samples
out. Then dividing the result by 16 will make the output even smaller. A “low”
frequency is now attenuated!
The 8500 Hz signal input case is more difficult to see. Since this signal is mostly
alternating in sign, or nearly so, the sum/difference process will now “subtract negatives”
and “add positives” thereby reinforcing the “sum”. Dividing the result by 16 will make it
smaller, but the result, or output, will be much larger at any one time than for the “low”
frequencies. This process is now a high-pass filter!
10.4 DSP Filters in General
DSP filters can also use past values of the outputs of the filter for an even greater
filtering effect. In general, most DSP filters today use “weights” or “constant
multipliers” that multiply each of the past inputs and each of the past outputs before
forming the sum to be used as the filter output. Modern DSP filters are linear, recursive,
difference equations. Any type of filter can be made by taking enough past inputs and
past outputs and giving them each a proper multiplying weight. The weights modulate
the filter’s performance by making the response more uniform in the pass-band, giving a
sharper cut-off of the undesired frequencies and overall, produce the desired filter shape.
10.5 Aliasing and the Nyquist Sampling Theorem
There is a limit to the highest frequency that can be represented in a signal when
the sampling rate is set at some value. It turns out that for a given sampling rate, call it
“SR”, the largest frequency component of a signal to be sampled must not exceed the
frequency “SR / 2”. If that signal does have frequencies exceeding “SR / 2”, those
frequencies will be “reflected back” and will interfere with the valid frequencies that are
less than “SR / 2”. These “reflections back” are called the “alias” frequencies of a
sampled signal. Unless these frequency components which are greater than “SR / 2” are
removed by an analog filter prior to being sampled, they will be “aliased”; they will
interfere with the frequency components which are less than “SR / 2”, and there will be
no way to remove them once they are sampled!
Why is this so? Why does “aliasing” occur at all? Let’s go back to Chapter 7 to
the part about the Direct Digital Synthesis method for generating sine waves. (Take
some time to review that now.) It was said, at the time without proof, that the maximum
frequency sine wave, which could be produced with DDS, was at half the sampling rate
(data rate). Is this true? Assume that the system that was used in Chapter 7 is the same
as the one we will use here. The sampling rate is 10 kHz and the accumulator and the
table-increment are both 16-bit registers. Suppose we want to use this DDS system to
produce a sine wave at 7500 Hz. The table-increment = 7500 / F-to-I = 49152 = 0xC000.

150
Let the accumulator start from zero and use the table-increment of 0xC000. This
produces the following data:
Accumulator Sine Wave Value
ACCUM = 0000 Zero
ACCUM = C000 - Max
ACCUM = 8000 Zero
ACCUM = 4000 + Max
ACCUM = 0000 Zero
The sampling rate is 10 kHz and this output has a period of four steps so our frequency is
2500 Hz!
On the other hand, suppose we have an ADC that samples at 10 kHz and we give
it an input sine wave with a frequency of 7500 Hz. What does this look like?
See Figure 10-1. The resulting samples are the same as the DDS case: The samples are
for a 2500 Hz sine wave!
Input Sine Wave Frequency = 7.5 kHz
10 kHz Sample Points
Alias Sine Wave at 2.5 kHz
Figure 10-1 Example of Aliasing When Sampling an Analog Signal
We can’t tell the difference between a 2500 Hz sine wave and a 7500 Hz sine
wave if the sampling rate is 10 kHz by looking at the samples they produce! We know

151
that the inputs are very different, but we can’t tell the difference by looking at the
samples alone. As far as we know there is no difference at all.
Aliasing is a fundamental fact of the nature of sampled-data systems. In most
applications we wish to prevent aliasing from occurring. In some cases, however, we can
use aliasing as a design technique. The decision to do so must be based on the size, cost
and complexity of the system to be built. If aliasing is to be used in an answering
machine, this is OK. If you are using DSP techniques to radar-map the surface of Venus
in a Venus-orbiting satellite, avoid aliasing at all costs!
If the input signal waveform to a sampled-data system has a maximum frequency
of “Fmax”, the Nyquist Sampling Theorem says that the data can be exactly represented
and recovered without aliasing if the sampling rate is at least twice the frequency of
“Fmax”. In practice, we use a little more than twice the frequency (say, 2.2 * Fmax), for
insurance.
10.6 DSP Cookbook I --- A Simple LPF/HPF
A simple low-pass/high-pass filter can be formed with the equation:
Y(n) = X(n) – (B1) * Y(n-1).
Where Y(n) is the n-th output, Y(n-1) is the (n-1)-th output, X(n) is the n-th input, and B1
is a constant multiplying weight. If B1 is as:
B1 > 0, Then the filter is a high-pass filter
B1 < 0, Then the filter is a low-pass filter.
ABS(B1) < 1.0 is a MUST, or the filter will be unstable
(Its outputs will try to run to infinity).
In a modified frequency-domain, a “normalized” frequency called “A” is defined as:
A = (2 * Pi * f) / (Sampling Rate).
Since the frequency “f” ranges over the range 0 <= f < (SR / 2) the normalized frequency
“A” will range over the range 0 <= A < Pi.
The filter “magnitude vs. normalized frequency” equation is:
Magnitude = 1 / sqrt( 1 + (B1 ** 2) + ( 2 * B1 * cos(A))).
The maximum magnitude for this filter is Max Magnitude = 1 / ( 1 – B1).

152
The half-power (3 dB) frequency is given by:
Cosine(A) = 2 – (( 1 – (B1 ** 2)) / ( 2 * B1)).
Note that the maximum magnitude and the filter gains get very large as “B1” is
just less than one. Also, if the half-power (3 dB) frequency is to be “tight”, “B1” must be
slightly less than one, and therefore the gains will be large.
There are several problems that are encountered when we attempt to implement
this filter on the PIC. Since “B1” is a constant, the multiply routine can be made as a
“multiply-by-a-constant” using “RLF”, “ADDWF”, “SUBWF”, or by using a look-up
table (hashing). If the filter gains are too high, however, a single-byte multiply may not
be sufficient. The inputs may have to be reduced (attenuated) before entering the filter to
reduce the outputs that are made large by the large gains. This will make the small
amplitudes of the signal even smaller. If the input must be a large amplitude and the
gains must be large (“small bandwidth”), working with double-byte arithmetic is a
MUST. Overall, the sampling rates and the PIC’s oscillator frequency, “Fosc”, must be
used to judge how much time is allowed to do the filter calculations.
An example filter program is as follows: (Put the input into W when this is called
and the output will also be in W)
FILTER:
MOVWF TEMP ; Store input
MOVF OLD_OUTPUT,W ; Get Y(n-1)
CALL MULTIPLY_B1
ADDWF TEMP,W ; Add X(n), Get X(n) – B1 * Y(n-1)
MOVWF OLD_OUTPUT ; Set Y(n) as Next Y(n-1)
RETURN ; W = Y(n) = Output
10.7 DSP Cookbook II --- A Simple BPF
A simple band-pass filter (BPF) can be produced by using the following equation:
Y(n) = X(n) - B1*Y(n-1) - B2*Y(n-2).
Where the Y()s are for the current, previous, and second-previous outputs, respectively.
The X(n) is the current input and “B1” and “B2” are the constant multiplier weights. The
weights “B1” and “B2” can be expressed in a different form as:
B1 = - 2 * R * Cosine(G)
B2 = R ** 2.

153
Where “G” is the “Normalized Center Frequency Desired”. The “R” value must be
greater than zero and less than one. If “R” is greater than or equal to one, the filter will
be unstable and the outputs will try to run to infinity.
The filter’s “Magnitude vs. Normalized Frequency” equation is considerably
more complicated than that of the LPF/HPF:
Magnitude = 1 / sqrt( C1 + C2 + C3)
C1 = 4 * B2 * (Cosine(A)) ** 2
C2 = 2 * B1 * ( 1 + B2) * Cosine(A)
C3 = B1 ** 2 + ( 1 – B2) ** 2
Where “A” is the normalized input frequency. The maximum gain at the center
frequency is:
Max Magnitude = 1 / (( 1 - R ** 2) * Sine(G)).
The bandwidth may be calculated from:
Cosine(A+) = Cosine(G) * (( 1 + R ** 2)/( 2 * R)) + Sine(G) * (( 1 – R ** 2)/( 2 * R))
Cosine(A-) = Cosine(G) * (( 1 + R ** 2)/( 2 * R)) – Sine(G) * (( 1 – R ** 2)/( 2 * R))
Bandwidth = ABS((A+) – (A-)) and A+- = Arccos(Cosine(A+-)).
The bandwidth is in the normalized frequency range.
Note that the gains and the maximum magnitude get very large as “R” is just less
than one, but this is required for having narrow bandwidths. The exact same design and
implementation issues are applicable here as they are in the LPF/HPF.
10.8 DSP Cookbook III --- A Median Filter
The median filter is a departure from the other filters in that there are no
multiplying weights and no summation of values. The median filter is for removing
impulse noise that may corrupt a signal. It removes noise spikes. It works by taking an
odd number of samples, preserving their order in RAM, removing the oldest point to fill
in the current point as before, but the set of samples is copied to a separate section of
RAM and sorted into order from lowest to highest. Then the middle point (the median) is
taken as the current output.
It is important to note that the median filter is nonlinear. Traditional frequency-
domain analysis techniques will not work on it. The median filter looks in some ways
like a low-pass filter but it preserves sharp transitions and edges like a high-pass filter. It

154
is also important to notice that the median filter selects but does not calculate in order to
do its filtering action.
One application of the median filter that we will need is for cleaning-up the noise
from the ADPCM expansion process (ADPCM was covered in Chapter 7). The ADPCM
expansion process includes many overshoots and undershoots in the signal as it tries to
approximate the original signal. These come out as noise spikes and are present in
enough volume to make listening to the output difficult. The median filter removes these
noise spikes and there is no noticeable noise at the median filter’s output.
One implementation problem with the median filter is running the filter at high
speeds. If the median filter is coded with a brute-force sorting-algorithm, it will take 113
instruction cycles to run. This is too much time and the median filter could not be used in
this form without moving the PIC to a higher oscillator frequency. But, an improved
algorithm will take only 57 instruction cycles, and this will make the median filter
practical in this application. The details of this improved algorithm are outlined in
Chapter 6 under the example of ROM States.
10.9 DSP Example I --- Standard DTMF Decoding
The “standard” way to decode DTMF signals is to use a bank of seven (7) band-
pass filters to decode each single frequency. These frequencies are:
697 Hz, 770 Hz, 852 Hz, 941 Hz, 1209 Hz, 1336 Hz, and 1477 Hz.
More complete details of the DTMF signaling process can be found in Chapter 7. Each
BPF will have two multiplier weights giving a total of 14 weights. Since the highest
frequency to measure is 1477 Hz, the sampling rate can be set at 3 kHz to satisfy the
Nyquist Sampling Theorem. This will give us 333 microseconds to do a run of all of the
filters. Another requirement is to measure the outputs of each of the filters to see if they
are showing high volume outputs to show if that particular frequency is present. This
means that each output of each filter must be converted to an absolute-value and
accumulated or “integrated” to judge if that frequency is present. This must use a double-
byte buffer for each filter to make this accumulation. After about 50 milliseconds, we
can stop the process and look for peaks in the accumulated filter outputs and decode them
into symbols (if there were any at all – there might not be any).
If the PIC’s clock oscillator is running at 4 MHz, each instruction will take one
microsecond to run. This means that the complete process of all of the filters and their
accumulations must run with only 333 instruction cycles. Some minor amount of
overhead is also needed.
Assume that all of the weight multiplications use look-up tables with 256 entries
each. This requires a total memory space of 3.5 kilo-words in ROM.

155
Let’s look at one BPF and one double-byte add and estimate the time it takes to
run and then multiply that by seven (7) to see if the process can be done in less than 333
microseconds. Let the following routine be called with the new input-data in a RAM
location called, “INPUT_N”:
BPF_FILTER:
MOVF OLD_OUT1,W ; Get Y(n-1)
CALL WEIGHT_TABLE1 ; Get Product B1 * Y(n-1) in W
MOVWF TEMP
MOVF OLD_OUT2,W ; Get Y(n-2)
CALL WEIGHT_TABLE2 ; Get Product B2 * Y(n-2) in W
ADDWF TEMP,W ; Sum The Two Products in W
ADDWF INPUT_N,W ; Add the Input to the two products
MOVWF RESULT ; Store The Y(n) Temporarily
MOVF OLD_OUT1,W ; Update Y(n-2) for Next Time
MOVWF OLD_OUT2
MOVF RESULT,W ; Update Y(n-1) for Next Time
MOVWF OLD_OUT1
MOVF RESULT,W ; Convert to ABS Value
BTFSC RESULT,7
SUBLW 0
ADDWF ACCUM_LOW,F ; Add to Accumulator-Low
BTFSC STATUS,C ; Add Carry If Any
INCF ACCUM_HIGH,F
RETURN
The above BPF and accumulate code takes about 40 microseconds to run. Seven of these
gives 280 microseconds. If we can do all of the overhead in 53 microseconds, this
solution will work. If not, we may need a faster clock speed.
10.10 DSP Example II --- Alternative DTMF Decoding
It may be easier to do the DTMF decoding without band-pass filters and without a
lot of memory by using aliasing as a design technique. Each of the DTMF frequencies
are sinusoidal with very tight tolerances on each frequency (2%). At any time there are
only two frequencies present. All of the frequency signals have the same amplitude as
the others. The whole waveform is just the sum of these two sine waves.
Suppose that we set the sampling rate to be exactly one of the DTMF frequencies,
one at a time, and over a window of a few milliseconds. On the “exact frequency” we
would sample the same corresponding part of the wave at any time – there would be no

156
change in the voltage of that part of the sine wave. We would alias it to a DC value! The
other sine wave would also be aliased to some other sine wave frequency. If we
accumulated each sample, the DC parts of one frequency alias would add together, but
the other alias frequency would alternate in sign, and its samples would add to zero! If
neither frequency were present, they would both be aliased to other sine waves and they
would both add to zero.
Two problems may exist in this scheme. One is that the DC valued alias that we
measure may be near zero and add only a small amount of value each time it gets added
to the sum. The other problem is that the “other” alias sine wave may be aliased to a very
low frequency and when adding its samples it may “look like” DC and thereby confuse
the whole process.
The first problem can be solved by having a second sampling of data within one
period with a shift of 90 degrees and run a second accumulator for it. If the first sum
gives us zero, the second sum would give us the peak sine values! Between the two sums
we can measure some substantial amount of DC if that frequency is present.
To solve the second problem we need to know how to predict where the second
frequency would be aliased. That is, go through all of the possible combinations of
aliases and see if there are any frequencies that are too low. Let’s develop a rule for
finding the alias frequency.
The aliasing formulas look like:
0 <= f <= (SR / 2) No Aliasing
(SR / 2) <= f <= (SR) Alias = SR – f
(SR) <= f <= (3*SR/2) Alias = f – SR
(3*SR/2) <= f <= (2*SR) Alias = (2*SR) – f
(2*SR) <= f <= (5*SR/2) Alias = f – (2*SR)
(5*SR/2) <= f <= (3*SR) Alias = (3*SR) – f
Where “f” is the input frequency of a single sine wave and “SR” is the sampling rate.
A summary of the aliasing analysis of the DTMF tones will be given here. First,
an example of one of the aliasing table results will be given as follows:
If SR = 770 Hz, Then:
F = 697 Hz Alias = 73 Hz
F = 1209 Hz Alias = 439 Hz
F = 1336 Hz Alias = 566 Hz
F = 1477 Hz Alias = 707 Hz

157
This table is the worst case of all the possible aliasing tables. Therefore, the
lowest aliasing frequency of everything is 73 Hz. To get cancellation of this wave, we
would need at least one cycle at a period of 13.7 milliseconds. For seven frequencies to
measure we need seven periods of 13.7 milliseconds each (roughly) to give a total of 96
milliseconds to sample the frequencies. Therefore, we can judge if the DTMF signals are
present and decode them in about one hundred milliseconds. There is no “second”
problem.
The software to measure the DTMF frequencies is much shorter than the band-
pass filter DTMF decoding method. The main loop must measure two samples per wave
at a 90-degree difference. The delays that make up the period of the sampling rates can
be adjusted to fit each frequency, but the delay routine can be made in a general way.
This software is as follows:
DTMF_ALIAS: ; Call with Code in W
CALL SET_DELAYS ; Set up delay/period & clear
; accum
DTMF_LOOP:
CLRW ; Signal: Add Zero to Accum
CALL GET_SAMPLE1 ; Add ADC to Accum, Start
; ADC
CALL DELAY ; Delay for Period / 4
NOP
NOP
NOP
MOVLW 1 ; Signal: Add ADC to Accum
CALL GET_SAMPLE2 ; Do 16-Bit Add to Accum
CALL DELAY
NOP
NOP
NOP
MOVLW 1 ; Signal: Add ADC to Accum
CALL GET_SAMPLE1 ; ADC is 8-Bit Data, Use Sign
; Ext
CALL DELAY
NOP
NOP
NOP
CLRW ; Signal Add Zero to Accum
CALL GET_SAMPLE2
CALL DELAY
BTFSS FLAGS,DO_RETURN ; This Flag Set in Delay
GOTO DTMF_LOOP
RETURN

158
The “GET_SAMPLE” routines are identical except that they work with different
accumulators. The “DELAY” routine not only delays to produce sampling at one-fourth
of the period in the DTMF_LOOP, but it also counts the number of these delays and sets
a flag to signal the DTMF_LOOP to return to the calling routine.
GET_SAMPLE1:
MOVWF ENABLE ; Store the Code to “Add or Zero”
MOVF ADRESH,W
SUBLW D’127’ ; Convert to Two’s Complement
BTFSS ENABLE,0
CLRW ; Code = 0, “Add Zero to Accum”
MOVWF ADC_VALUE
CLRF EXTEND ; Clear Sign-Extend Byte
BTFSC ADC_VALUE,7 ; Test Sign-Bit
COMF EXTEND ; Sign-Extend as “Negative”
CLRF CARRY
MOVF ADC_VALUE,W
ADDWF ACCUM_LOW1,F ; Add ADC to Accum1 (Low)
BTFSC STATUS,C ; Capture the Carry-State
INCF CARRY,F
MOVF CARRY,W ; Add Carry to Accum1 (High)
ADDWF ACCUM_HIGH1,F
MOVF EXTEND,W ; Add Sign-Extend to Same
ADDWF ACCUM_HIGH1,F
BCF PIR1,ADIF ; Prepare & Start ADC
BSF ADCON0,GO
RETURN
DELAY:
MOVF MASTER_WHOLE,W ; Get “Reference Copy” of
MOVWF WHOLE ; “Whole-Loop” Delay
MOVF FRACTION,W
ADDWF PCL,F
NOP
NOP
NOP
NOP
NOP
NOP
NOP
NOP
DELAY_LOOP:
NOP
NOP
NOP
NOP

159
NOP
DECFSZ WHOLE,F
GOTO DELAY_LOOP
DECFSZ COUNT,F
RETURN
BSF FLAGS,DO_RETURN
RETURN
10.11 DSP Application: Speech Compression
In the early 1970’s a company called Compressed Time, Inc. (CTI) came out with
a novel machine for manipulating the play-back speeds of tape-recorded speech. The
idea was to take the speech of a slow-speaking person on a tape-recording and speed it up
but compensate for the shift in the speech frequencies. That is, the rate of delivery of the
speech would be faster but the person would still appear to speak in a “normal” tone of
voice! The opposite could also be done. That is, slow-down a fast-speaking person and
have him/her appear to speak in a “normal” voice. CTI called this process, “Speech
Compression”.
The equipment that CTI developed was expensive and it contained a tape-recorder
with several play-back heads mounted on a rotating drum. The tape would be fed at the
desired speed while the drum would rotate at a speed that would sample the speech on the
tape and produce a play-back speed difference so that the relative play-back speed was
“normal”.
Doing the same process today is made cheap and easy by using DSP techniques.
Speech compression can be done very easily with the PIC although more analog filtering
would be required.
Let’s consider how we can do speech compression with DSP techniques.
Telephone-quality speech has a maximum frequency of about 4 kHz. To sample
and recover this speech would require a sampling-rate of 8 kHz to avoid aliasing. If the
speech were on a tape-recording and the play-back speed were doubled, the maximum
speech frequency would also double and the new sampling-rate would be doubled to
16 kHz.
There is another kind of sampling that must be considered, however. When we
speak, we hold each consonant and each vowel sound for a minimum of about 20
milliseconds. This figure is only approximate and it varies considerably from speaker to
speaker. If each consonant and vowel sound were considered to be a “bit” of
“information”, the typical “information-rate” is about 50 Hz at most! When the tape-

160
recording play-back speed is doubled this information-rate is also doubled to 100 Hz at
most.
Both of these processes need to be sampled in order to do speech compression.
Sampling each waveform point at 16 kHz is easy but how do we sample the
“information”? The answer is to fill an array of 16 kHz samples over a time-window of
about 10 milliseconds to get a total of 160 sample-elements. Ten milliseconds
corresponds to an “information-sampling-rate” of 100 Hz, but this is not the proper
Nyquist sampling rate for 100 Hz data. In actual practice, a sampling-rate for the time-
window samples may be substantially different from the Nyquist rate. The danger is in
loosing information and the actual number of samples in the array may need to be larger
to faithfully reproduce the speech sounds. This will require some experimentation to get
the speech to sound the right way and with no distortion.
To play this speech back at a “normal” tone of voice, the 160 array points are sent
out at an 8 kHz rate over a 20-millisecond time interval. This process is shown in
Figure 10-2. The speech is recorded at 16 kHz and the 160 samples are spread-out by
sending them out at 8 kHz. In this time interval an equivalent of 160 samples are skipped
in order to send out the first 160 points. The “catch” in doing speech compression in the
“Fast-to-Slow” mode is that some information is lost and cannot be recovered. But if
there is a right mix of sample points and array-sample-windows, the human ear will not
detect the lost information. In addition to a low-pass filter to smooth the output, a high-
pass filter is needed to get rid of the 50 Hz noise that comes in when the array-sample-
windows are “chopped” or “broken” on the play-back process. The human ear will hear
the speaker in a normal tone of voice but at double the speed.
Let’s now consider the “opposite” of the “Fast-to-Slow” mode process. That is,
let’s do the “Slow-to-Fast” mode process.
If the play-back speed were cut in half, the maximum speech frequency would be
2 kHz and we would need a sampling-rate of only 4 kHz. If a 20-millisecond time-
window were used, this would fill an array of 80 sample points.
The “Slow-to-Fast” mode process is shown in Figure 10-3. The trick is to
completely fill the 80-element array and then send it out, twice, each time at a rate of 8
kHz. This time there is no lost information. The human ear will hear the speaker in a
normal tone of voice but at half the speed.
Another application of these techniques is to alter the tone of voice of a live,
normal speaker. This gives some very strange-sounding effects. Several radio and
television science fiction shows have used these techniques to radically change the
speech of their actor’s voices.

161
Input
Sampling
Rate =
16 kHz
Output
Sampling
Rate =
8 kHz
(10 ms) (10 ms) (10 ms) (10 ms)
160 Samples160 Samples160 Samples160 Samples
(Skipped)
(Skipped)
(20 ms)(20 ms)
160 Samples 160 Samples
Figure10-2 "Slow-To-Fast" Mode Speech Compression Process

162
Input
Output
Sampling
Sampling
Rate =
Rate =
4 kHz
8 kHz
(20 ms) (20 ms)
80 Samples80 Samples
80 Samples 80 Samples 80 Samples 80 Samples
(10 ms)(10 ms)(10 ms)(10 ms)
(Repeated) (Repeated)
Figure 10-3 "Fast-To-Slow" Mode Speech Compression Process

163
Appendix A --- The PIC16F877 Instruction Set
The PIC16F877 has a total of 35 instructions, all of which are as a single word.
ADDLW (“Add Literal to W”)
Syntax: ADDLW k
Operand: k, where 0 <= k <= 255
Operation: (W) = (W) + k
Status: C, DC, Z
Cycles: 1 Instruction Cycle
Binary Code: 11 111x kkkk kkkk
Code Example: “ADDLW 7”
Operation Example: If (W) = 3 and “ADDLW 7” is executed, (W) = 10.
Description:
The contents of the W register are added to the constant, k, and the
Results are placed in the W register.
ADDWF (“Add W and file”)
Syntax: ADDWF f,d
Operands: f, where 0 <= f <= 127
And d, where d = {0,1}
Operation: (W) = (W) + (file) if d = 0
(file) = (W) + (file) if d = 1
Status: C, DC, Z
Binary Code: 00 0111 dfff ffff
Code Examples: “ADDWF SUM,W” for d = 0
“ADDWF SUM,F” for d = 1
Operation Examples:
If (W) = 10 and (SUM) = 20,
“ADDWF SUM,W” gives (W) = 30 and (SUM) = 20
“ADDWF SUM,F” gives (W) = 10 and (SUM) = 30
Description:
The contents of the W register are added to the contents of the
Register-file and the results are placed in:
(W) if d = 0
(file) if d = 1

164
ANDLW (“And Literal and W”)
Syntax: ANDLW k
Operation: (W) = (W) & k
Status: Z
Binary Code: 11 1001 kkkk kkkk
Code Example: “ANDLW 0x07”
Operation Example: If (W) = 0xE6 and “ANDLW 0x07” is executed,
(W) = 0x06.
Description:
The contents of the W register are ANDed with the constant, k, and the
ANDWF (“And W and file”)
Syntax: ANDWF f,d
Operation: (W) = (W) & (file) if d = 0
(file) = (W) & (file) if d = 1
Status: Z
Code Examples: “ANDWF SUM,W” for d = 0
“ANDWF SUM,F” for d = 1
Operation Examples:
If (W) = 0xC7 and (SUM) = 0xE5,
“ANDWF SUM,W” gives (W) = 0xC5 and (SUM) = 0xE5
“ANDWF SUM,F” gives (W) = 0xC7 and (SUM) = 0xC5
Description:
The contents of the W register are ANDed to the contents of the
(W) if d = 0
(file) if d = 1

165
BCF (“Bit Clear file”)
Syntax: BCF f,b
And b, where 0 <= b <= 7
Operation: file<b> = 0
Status: None
Binary Code: 01 00bb bfff ffff
Code Example: “BCF FLAGS,3”
Operation Example: If FLAGS<3> = 1, doing “BCF FLAGS,3” will make
FLAGS<3> = 0.
Description:
Bit number, b, is cleared (=0) in the register-file.
BSF (“Bit Set file”)
Syntax: BSF f,b
Operation: file<b> = 1
Status: None
Code Example: “BSF FLAGS,3”
Operation Example: If FLAGS<3> = 0, doing “BSF FLAGS,3” will make
FLAGS<3> = 1.
Description:
Bit number, b, is set (=1) in the register-file.
BTFSC (“Bit Test, Skip If Clear”)
Syntax: BTFSC f,b
Operation: Skip Next Instruction if file<b> = 0.
Status: None
Cycles: 1 Cycle If “NO” Skip
2 Cycles If SKIP
Code Example: “BTFSC FLAGS,3”
Operation Example: If FLAGS<3> = 0, “BTFSC FLAGS,3” will Skip.
If FLAGS<3> = 1, “BTFSC FLAGS,3” will Not Skip.

166
Description:
If the register-file bit, b, is zero, the next instruction will be skipped.
If the register-file bit, b, is one, the next instruction will execute.
BTFSS (“Bit Test, Skip If Set”)
Syntax: BTFSS f,b
Operation: Skip Next Instruction if file<b> = 1.
Status: None
Cycles: 1 Cycle If “NO” Skip
2 Cycles If SKIP
Code Example: “BTFSS FLAGS,3”
Operation Example: If FLAGS<3> = 1, “BTFSS FLAGS,3” will Skip.
If FLAGS<3> = 0, “BTFSS FLAGS,3” will Not Skip.
Description:
If the register-file bit, b, is one, the next instruction will be skipped.
If the register-file bit, b, is zero, the next instruction will execute.
CALL (“Call Subroutine”)
Syntax: CALL k
Operation: (Program Counter) + 1 is moved to the Top-of-Stack
Address = k is transferred to PC<10:0>
(PCLATH<4,3>) is transferred to PC<12,11>
Status: None.
Cycles: 2 Instruction Cycles
Binary Code: 10 0kkk kkkk kkkk
Code Example: “CALL START”
Operation Example: If START = 0x00C7 and “CALL START” is done,
(PC) = 0x00C7.
Description:
Call a subroutine. The incremented program counter (PC) is placed
On the stack, the program counter, bits zero through 10, are loaded
With the target address, k, and the PCLATH bits <4,3> are placed into
The program counter bits <12,11>.
Notes:
Doing a CALL within a 2 K block does not require any manipulation
Of the PCLATH register. If the CALL is beyond the current 2 K block,
The PCLATH register bits <4,3> must be set-up before doing the CALL.

167
CLRF (“Clear file”)
Syntax: CLRF f
Operand: f, where 0 <= f <= 127
Operation: (f) = 0.
Status: Z
Binary Code: 00 0001 1fff ffff
Code Example: “CLRF FLAGS”
Operation Example: If FLAGS = 0x20, doing “CLRF FLAGS” will make
(0x20) = 0x00.
Description:
The register-file is reset. Zeros are written to this RAM byte.
CLRW (“Clear W”)
Syntax: CLRW
Operand: None.
Operation: (W) = 0.
Status: Z
Binary Code: 00 0001 0xxx xxxx
Code Example: “CLRW”
Operation Example: Doing “CLRW” makes (W) = 0x00.
Description:
Zeros are written to the W register.
CLRWDT (“Clear Watch-Dog Timer”)
Syntax: CLRWDT
Operand: None
Operation: (WDT) = 0x00. (WDT Prescaler) = 0x00.
Status: STATUS Flags: /TO = 1 and /PD = 1.
Binary Code: 00 0000 0110 0100
Code Example: “CLRWDT”
Operation Example: Doing “CLRWDT” resets the Watch-Dog Timer.
Description:
CLRWDT resets the Watch-Dog Timer and its prescaler.

168
COMF (“Complement file”)
Syntax: COMF f,d
Operation: (W) = /(f) if d = 0
(f) = /(f) if d = 1
Status: Z
Code Examples: “COMF FLAGS,W” for d = 0.
“COMF FLAGS,F” for d = 1
Operation Examples:
If (W) = 0xE3 and (FLAGS) = 0xF1,
Doing “COMF FLAGS,W” gives (W) = 0x0E and (FLAGS) = 0xF1.
Doing “COMF FLAGS,F” gives (W) = 0xE3 and (FLAGS) = 0x0E.
Description:
Take the one’s complement of the register-file and place the results:
In (W) if d = 0
In (file) if d = 1.
DECF (“Decrement file”)
Syntax: DECF f,d
Operation: (W) = (f) – 1 if d = 0
(f) = (f) – 1 if d = 1
Status: Z
Code Examples: “DECF COUNT,W” for d = 0
“DECF COUNT,F” for d = 1
Operation Examples:
If (W) = 0x3A and (COUNT) = 0x12,
Doing “DECF COUNT,W” gives (W) = 0x11 and (COUNT) = 0x3A.
Doing “DECF COUNT,F” gives (W) = 0x3A and (COUNT) = 0x11.
Description:
The contents of the register-file are decremented by one and the results
Are placed as:
In (W) if d = 0
In (file) if d = 1.

169
DECFSZ (“Decrement file, Skip If Zero”)
Syntax: DECFSZ f,d
Operations: (W) = (f) – 1 if d = 0
(f) = (f) – 1 if d = 1
and Skip the next instruction if the result was zero.
Status: None.
Cycles: 1 Instruction Cycle if non-zero result (No Skip)
2 Instruction Cycles if zero result (Skip)
Code Example: “DECFSZ COUNT,W” if d = 0
“DECFSZ COUNT,F” if d = 1
Operation Example:
If (W) = 0x10 and (COUNT) = 0x01,
“DECFSZ COUNT,W” gives (W) = 0x00 and (COUNT) = 0x01
“DECFSZ COUNT,F” gives (W) = 0x10 and (COUNT) =0x00
In Both Cases: The Next Instruction Will Be Skipped.
If (W) = 0x10 and (COUNT) = 0x23,
“DECFSZ COUNT,W” gives (W) = 0x22 and (COUNT) = 0x23
“DECFSZ COUNT,F” gives (W) = 0x10 and (COUNT) = 0x22
In Both Case: The Next Instruction Will Be Executed.
Description:
The contents of the register-file is decremented and the results are
Placed in:
(W) if d = 0
(file) if d = 1
And the next instruction will be skipped if the result was zero.
Otherwise, the next instruction will be executed.

170
GOTO (“Unconditional Branch”)
Syntax: GOTO k
Operation: Address = k is transferred to PC<10:0>
(PCLATH<4,3>) is transferred to PC<12,11>
Status: None.
Binary Code: 10 1kkk kkkk kkkk
Code Example: “GOTO START”
Operation Example: If START = 0x00C7 and “GOTO START” is done,
(PC) = 0x00C7.
Description:
Jump to a new address.
The program counter, bits zero through 10, are loaded
With the target address, k, and the PCLATH bits <4,3> are placed into
The program counter bits <12,11>.
Notes:
Doing a GOTO within a 2 K block does not require any manipulation
Of the PCLATH register. If the GOTO is beyond the current 2 K block,
The PCLATH register bits <4,3> must be set-up before doing the GOTO.
INCF (“Increment file”)
Syntax: INCF f,d
Operation: (W) = (f) + 1 if d = 0
(f) = (f) + 1 if d = 1
Status: Z
Code Examples: “INCF COUNT,W” for d = 0
“INCF COUNT,F” for d = 1
Operation Examples:
If (W) = 0x3A and (COUNT) = 0x12,
Doing “INCF COUNT,W” gives (W) = 0x13 and (COUNT) = 0x3A.
Doing “INCF COUNT,F” gives (W) = 0x3A and (COUNT) = 0x13.
Description:
The contents of the register-file are incremented by one and the results
Are placed as:
In (W) if d = 0
In (file) if d = 1.

171
INCFSZ (“Increment file, Skip If Zero”)
Syntax: INCFSZ f,d
Operations: (W) = (f) + 1 if d = 0
(f) = (f) + 1 if d = 1
and Skip the next instruction if the result was zero.
Status: None.
Cycles: 1 Instruction Cycle if non-zero result (No Skip)
2 Instruction Cycles if zero result (Skip)
Code Example: “INCFSZ COUNT,W” if d = 0
“INCFSZ COUNT,F” if d = 1
Operation Example:
If (W) = 0x10 and (COUNT) = 0xFF,
“INCFSZ COUNT,W” gives (W) = 0x00 and (COUNT) = 0xFF
“INCFSZ COUNT,F” gives (W) = 0x10 and (COUNT) =0x00
In Both Cases: The Next Instruction Will Be Skipped.
If (W) = 0x10 and (COUNT) = 0x23,
“INCFSZ COUNT,W” gives (W) = 0x24 and (COUNT) = 0x23
“INCFSZ COUNT,F” gives (W) = 0x10 and (COUNT) = 0x24
In Both Case: The Next Instruction Will Be Executed.
Description:
The contents of the register-file is incremented and the results are
Placed in:
(W) if d = 0
(file) if d = 1
And the next instruction will be skipped if the result was zero.
Otherwise, the next instruction will be executed.
IORLW (“Inclusive Or Literal and W”)
Syntax: IORLW k
Operation: (W) = (W) | k
Status: Z
Code Example: “IORLW 0x07”
Operation Example: If (W) = 0xE6 and “IORLW 0x07” is executed,
(W) = 0xE7.

172
Description:
The contents of the W register are ORed with the constant, k, and the
IORWF (“Inclusive Or W and file”)
Syntax: IORWF f,d
Operation: (W) = (W) | (file) if d = 0
(file) = (W) | (file) if d = 1
Status: Z
Code Examples: “IORWF SUM,W” for d = 0
“IORWF SUM,F” for d = 1
Operation Examples:
“IORWF SUM,W” gives (W) = 0xE7 and (SUM) = 0xE5
“IORWF SUM,F” gives (W) = 0xC7 and (SUM) = 0xE7
Description:
The contents of the W register are ORed to the contents of the
(W) if d = 0
(file) if d = 1
MOVF (“Move file”)
Syntax: MOVF f,d
Operations: (W) = (f) if d = 0
(f) = (f) if d = 1
Status: Z
Code Examples: “MOVF TEMP,W” for d = 0
“MOVF TEMP,F” for d = 1
Operation Examples:
If (W) = 0x25 and (TEMP) = 0xC7,
“MOVF TEMP,W” gives (W) = 0xC7 and (TEMP) = 0xC7
“MOVF TEMP,F” gives (W) = 0x25 and (TEMP) = 0xC7

173
Description:
If d = 0, the register-file contents are placed in W, with the register
File unchanged. If d = 1, there is no change in the contents of either
W or the register-file. In both cases the Z flag is set if the data is zero
And it is reset otherwise.
MOVLW (“Move Literal to W”)
Syntax: MOVLW k
Operation: (W) = k
Status: None
Binary Code: 11 00xx kkkk kkkk
Code Example: “MOVLW 0xB5”
Operation Example:
If (W) = 0x03, Doing “MOVLW 0xB5” gives (W) = 0xB5.
Description:
The literal value, k, is moved into the W register.
MOVWF (“Move W to file”)
Syntax: MOVWF f
Operand: f, where 0 <= f <= 127
Operation: (f) = (W)
Status: None
Binary Code: 00 0000 1fff ffff
Code Example: “MOVWF TEMP”
Operation Example:
If (W) = 0xA4 and (TEMP) = 0x3F,
“MOVWF TEMP” gives (W) = 0xA4 and (TEMP) = 0xA4.
Description:
The contents of W are moved to the register-file.
NOP (“No Operation”)
Syntax: NOP
Operand: None
Operation: None
Status: None
Binary Code: 00 0000 0000 0000
Code Example: “NOP”

174
Operation Example: Doing “NOP” does nothing.
Description:
This instruction does nothing but does take one instruction cycle
To execute.
RETFIE (“Return from Interrupt”)
Syntax: RETFIE
Operand: None
Operation: (PC) = Top-of-Stack and INTCON<GIE> = 1
Status: None
Binary Code: 00 0000 0000 1001
Code Example: “RETFIE”
Operation Example:
If (TOS) = 0x0017 and (PC) = 0x108F,
“RETFIE” gives (PC) = 0x0017.
Description:
The program counter is filled with the address at the top of the stack.
Also, the INTCON bit “GIE” is set (=1).
RETLW (“Return from Subroutine with Literal in W”)
Syntax: RETLW k
Operation: (PC) = Top-of-Stack and (W) = k
Status: None
Binary Code: 11 01xx kkkk kkkk
Code Example: “RETLW 0x34”
Operation Example:
If (TOS) = 0x0017 and (PC) = 0x108F and (W) = 0xFC,
“RETLW 0x34” gives (PC) = 0x0017 and (W) = 0x34.
Description:
Also, the W register is filled with the literal value, k.
RETURN (“Return from Subroutine”)
Syntax: RETURN
Operand: None
Operation: (PC) = Top-of-Stack
Status: None

175
Binary Code: 00 0000 0000 1000
Code Example: “RETURN”
Operation Example:
If (TOS) = 0x0017 and (PC) = 0x108F,
“RETURN” gives (PC) = 0x0017.
Description:
RLF (“Rotate Left file Through Carry”)
Syntax: RLF f,d
Operations: See Figure 5-7 in Chapter 5.
(W) = result if d = 0
(f) = result if d = 1
Status: C
Code Example: “RLF POSITION,W” for d = 0
“RLF POSITION,F” for d = 1
Operation Example:
If (POSITION) = 0x8C and (W) = 0x02 and (C-flag) = 0,
“RLF POSITION,W” gives (W) = 0x18 and (C-flag) = 1 and (POSITION) =
0x8C
“RLF POSITION,F” gives (W) = 0x02 and (C-flag) = 1 and (POSITION) =
0x18
Description:
See attached Figure. The results are placed in:
(W) if d = 0
(file) if d = 1
RRF (“Rotate Right file Through Carry”)
Syntax: RRF f,d
Operations: See Figure 5-7 in Chapter 5.
(W) = result if d = 0
(f) = result if d = 1
Status: C
Code Example: “RRF POSITION,W” for d = 0
“RRF POSITION,F” for d = 1
Operation Example:

176
If (POSITION) = 0x8D and (W) = 0x02 and (C-flag) = 0,
“RRF POSITION,W” gives (W) = 0x46 and (C-flag) = 1 and (POSITION) =
0x8D
“RRF POSITION,F” gives (W) = 0x02 and (C-flag) = 1 and (POSITION) =
0x46
Description:
See attached Figure. The results are placed in:
(W) if d = 0
(file) if d = 1
7
7
6
6
5
5
4
4
3
3
2
2
1
1
0
0
C
C
Flag
Flag
Register-File Byte
Register-File Byte
RLF file,d
RRF file,d
Diagram of RLF and RRF Instructions
SLEEP (“Enter Sleep Mode”)
Syntax: SLEEP
Operand: None
Operation: (WDT) = 0x00, (WDT Prescaler) = 0x00,
STATUS flags: /TO = 1 and /PD = 0.
Status: None
Binary Code: 00 0000 0110 0011
Code Example: “SLEEP”
Operation Example:

177
Doing “SLEEP” sends the CPU into sleep mode.
Description:
Doing this instruction sends the CPU into sleep mode.
SUBLW (“Subtract W from Literal”)
Syntax: SUBLW k
Operation: (W) = k – (W)
Status: C, DC, Z
Binary Code: 11 110x kkkk kkkk
Code Example: “SUBLW 7”
Operation Example: If (W) = 3 and “ADDLW 7” is executed, (W) = 4.
Description:
The contents of the W register are subtracted from the constant, k,
And the results are placed in the W register.
SUBWF (“Subtract W from file”)
Syntax: SUBWF f,d
Operation: (W) = (file) – (W) if d = 0
(file) = (file) – (W) if d = 1
Status: C, DC, Z
Code Examples: “SUBWF SUM,W” for d = 0
“SUBWF SUM,F” for d = 1
Operation Examples:
If (W) = 10 and (SUM) = 25,
“SUBWF SUM,W” gives (W) = 15 and (SUM) = 25
“SUBWF SUM,F” gives (W) = 10 and (SUM) = 15
Description:
The contents of the W register are subtracted from the contents of the
(W) if d = 0
(file) if d = 1

178
SWAPF (“Swap Nibbles in file”)
Syntax: SWAPF f,d
Operations: (W) = (f<3:0>,f<7:4>) if d = 0
(f) = (f<3:0>,f<7:4>) if d = 1
Status: None
Code Examples: “SWAPF TEMP,W” for d = 0
“SWAPF TEMP,F” for d = 1
Operation Examples:
If (W) = 0xC7 and (TEMP) = 0xA9,
“SWAPF TEMP,W” gives (W) = 0x9A and (TEMP) = 0xA9
“SWAPF TEMP,F” gives (W) = 0xC7 and (TEMP) = 0x9A
Description:
This instruction swaps the high and low nibbles of the register-file
And places the results in:
(W) if d = 0
(file) if d = 1
XORLW (“Exclusive Or Literal and W”)
Syntax: XORLW k
Operation: (W) = (W) .XOR. k
Status: Z
Code Example: “XORLW 0x07”
Operation Example: If (W) = 0xE6 and “XORLW 0x07” is executed,
(W) = 0xE3.
Description:
The contents of the W register are XORed with the constant, k, and the

179
XORWF (“Exclusive Or W and file”)
Syntax: XORWF f,d
Operation: (W) = (W) .XOR. (file) if d = 0
(file) = (W) .XOR. (file) if d = 1
Status: Z
Code Examples: “XORWF SUM,W” for d = 0
“XORWF SUM,F” for d = 1
Operation Examples:
“XORWF SUM,W” gives (W) = 0x22 and (SUM) = 0xE5
“XORWF SUM,F” gives (W) = 0xC7 and (SUM) = 0x22
Description:
The contents of the W register are XORed to the contents of the
(W) if d = 0
(file) if d = 1

180
Appendix B --- Useful C++ Programs for
Developing PIC ASM Applications
1) PERMS.CPP --- Permutation Hash Tables
2) SINEH.CPP --- Sine / Cosine Tables
3) OPENLOOP.CPP --- “Open-Loop” Time-Measurement

181
//==========================================================
// Program: PERMS.CPP == Generate Permutation Tables
// in PIC16F877 Assembly Language.
//
// Author: Timothy D. Green
// Date: 26 NOV 2005
//
// Compiler: Turbo C++ for DOS, Rev. 3.0
//
// This Program May Be Copied Freely
// It is in the Public Domain.
//
// The User of this Program does so at his or her own risk
// and bears the full responsibility thereof.
// The author, Timothy D. Green, assumes no liability
// for the use of this program.
//
//==========================================================
#include<stdio.h>
int main(void)
{
unsigned int k,Map[8],Power[8],N_Bits,Test,Value,N_Table,j,m;
FILE *out;
printf("Enter Number of Bits = ? ");
scanf("%d",&N_Bits);
for(k=0,m=1;k < 8;k++,m=2*m) Power[k] = m;
N_Table = 1;
k = N_Bits;
while(k > 0){
N_Table = 2 * N_Table;
k--;
}
printf("nnn");
for(k=0;k < N_Bits;k++){
printf("For Input Bit = %d: What is the Output Bit Number = ? ",k);
scanf("%d",&m);
Map[k] = m;
printf("n");
}
out = fopen("PICPERM.ASM","w");
for(k=0;k < N_Table;k++){
Value = 0;
for(m=0;m < N_Bits;m++){
Test = Power[m];
Test = Test & k;
if(Test > 0) Value = Value + Power[Map[m]];
}
Value = Value & 0x00ff;

182
fprintf(out," RETLW %dn",Value);
}
fclose(out);
return 0;
}
//==========================================================
// Program: SINEH.CPP == Generate Sine or Cosine Tables
//
// Date: 26 NOV 2005
//
//
//
//
//==========================================================
#include<stdio.h>
#include<math.h>
int main(void)
{
char H,L,Hex[16];
char S;
int k,N,j;
double Rk,Pi,Peak,RN,Angle,Sine;
FILE *out;
//=================
Peak = 127.5;
N = 128;
//=================
RN = (double) N;
Pi = 3.1415926535;
for(k=0;k < 10;k++) Hex[k] = '0' + k;
for(k=10,j=0;k < 16;k++,j++) Hex[k] = 'A' + j;
//=================
out = fopen("PICSINE.ASM","w");
for(k=0;k < N;k++){

183
Rk = (double) k;
Angle = (2.0 * Pi * Rk) / RN;
Sine = Peak * sin(Angle);
S = (char) Sine;
L = S & 0x0f;
H = ((S & 0xf0) >> 4) & 0x0f;
L = Hex[L];
H = Hex[H];
fprintf(out," RETLW 0x%c%cn",H,L);
}
fclose(out);
return 0;
}
//==========================================================
// Program: OPENLOOP.CPP == Generate "Open-Loop" Code
//
// Date: 26 NOV 2005
//
//
//
//
//==========================================================
#include<stdio.h>
int main(void)
{
int Start,Stop,k;
FILE *out;
//=================
Start = 0;
Stop = 100;
//=================
out = fopen("PICLOOP.ASM","w");

184
for(k=Start;k <= Stop;k++){
fprintf(out," BTFSS PORTC,TEST_BITn");
fprintf(out," RETLW %dn",k);
}
fclose(out);
return 0;
}

185
Appendix C --- Special Function Registers
(RAM Addresses & Bits)
STATUS (Address 0x03, In All Banks)
7 IRP Indirect Addressing Bank Select [0 = Banks(0,1), 1 = Banks(2,3)]
6 RP1 Bank Select: 11=Bank(3), 10=Bank(2), 01=Bank(1), 00=Bank(0)
5 RP0 Bank Select
4 /TO Time-Out Bit 1=(Power-up,CLRWDT,SLEEP) 0=WDT Overflow
3 /PD Power-Down Bit 1=(Power-up,CLRWDT) 0=SLEEP
2 Z Zero Bit 1=Result was Zero 0=Result was NOT Zero
1 DC Digit Carry Bit =Carry-State Out of 4th
Low Bit of Result
0 C Carry Bit =Carry-State Out of Result Most Significant Bit
PCON (Address 0x8E, Bank 2 Only) (Also, Uses only Bits 1 & 0)
1 /POR Power-On Reset Status: 1= No Reset 0=Reset
0 /BOR Brown-Out Reset Status: 1=No Reset 0=Reset
INTCON (Address 0x0B, In All Banks)
7 GIE Global Interrupt Enable Bit
6 PEIE Peripheral Interrupt Enable Bit
5 T0IE Timer0 Interrupt Enable Bit
4 INTE RB0/INT External Interrupt Enable Bit
3 RBIE PortB-Change Interrupt Enable Bit
2 T0IF Timer0 Interrupt Flag
1 INTF RB0/INT External Interrupt Flag
0 RBIF RB0/INT External Interrupt Flag
PIR1 (Address 0x0C, Bank 0 Only)
7 PSPIF Parallel Slave Port Interrupt Flag (Set on Read/Write)
6 ADIF ADC Interrupt Flag (Set when Conversion is Done)
5 RCIF USART Receive Interrupt Flag
4 TXIF USART Transmit Interrupt Flag
3 SSPIF Synchronous Serial Port (SSP) Interrupt Flag
2 CCP1IF CCP1 Interrupt Flag (Capture Mode or Compare Mode)
1 TMR2IF TMR2-to-PR2 Match Interrupt Flag
0 TMR1IF TMR1 Overflow Interrupt Flag

186
PIR2 (Address 0x0D, Bank 0 Only)
6 Reserved --- Never Write to This Bit
4 EEIF EEPROM Write-Operation Interrupt Flag
3 BCLIF Bus Collision Interrupt Flag
0 CCP2IF CCP2 Interrupt Flag (Capture Mode or Compare Mode)
PIE1 (Address 0x8C, Bank 1 Only)
7 PSPIE Parallel Slave Port Interrupt Enable Bit
6 ADIE ADC Interrupt Enable Bit
5 RCIE USART Receive Interrupt Enable Bit
4 TXIE USART Transmit Interrupt Enable Bit
3 SSPIE Synchronous Serial Port (SSP) Interrupt Enable Bit
2 CCP1IE CCP1 Interrupt Enable Bit
1 TMR2IE TMR2 Interrupt Enable Bit
0 TMR1IE TMR1 Interrupt Enable Bit
PIE2 (Address 0x8D, Bank 1 Only)
6 Reserved --- Never Write to This Bit
4 EEIE EEPROM Write-Operation Interrupt Enable Bit
3 BCLIE Bus Collision Interrupt Enable Bit
0 CCP2IE CCP2 Interrupt Enable Bit
OPTION-REG (Address 0x81, Banks 1 and 3)
7 /RBPU PortB Pull-Up Enable Bit 1=Disable 0=Enable
6 INTEDG Interrupt Edge Select Bit 1=Rising 0=Falling
5 T0CS TMR0 Clock Source Select 1=RA4/T0CKI 0=Internal
4 T0SE TMR0 Source Edge Select 1=Falling 0=Rising
3 PSA Prescaler Assignment Bit 1=WDT 0=TMR0
2 PS2 (PS2:PS0)= See Chapter 8, Timer0 or WDT, Prescaler
1 PS1
0 PS0

187
ADCON0 (Address 0x1F, Bank 0 Only)
7-6 ADCS1:ADCS0 ADC Clock Select Bits
00 = Fosc / 2
01 = Fosc / 8
10 = Fosc / 32
11 = Internal RC-OSC for ADC
5-3 CHS2:CHS0 Analog Channel Select Bits
(0,0,0) = Channel 0
(0,0,1) = Channel 1
(0,1,0) = Channel 2
(0,1,1) = Channel 3
(1,0,0) = Channel 4
(1,0,1) = Channel 5
(1,1,0) = Channel 6
(1,1,1) = Channel 7
2 GO ADC Start Conversion Bit (Set this to Start a Conversion)
0 ADON ADC Activation Bit (Set this to turn the ADC unit ON)
ADCON1 (Address 0x9F, Bank 1 Only)
7 ADFM ADC Result Format Bit 1=Right-Justified 0=Left-Justified
3-0 PCFG3:PCFG0 ADC Port Configuration Bits (See Chapter 8)
T1CON (Address 0x10, Bank 0 Only)
5-4 T1CKPS1:T1CKPS0 Timer1 Input Clock Prescale Select Bits
(1,1) = 1:8 Prescale Value
3 T1OSCEN Timer1 Oscillator Enable Control Bit (1=Enable, 0=Disable)
2 /T1SYNC Timer1 External Clock Sync Bit (1=No Sync, 0=Sync)
1 TMR1CS Timer1 Clock Source Select (1=External Clock, 0=Internal)
(External Clock on RC0/T1OSO/T1CKI)
0 TMR1ON Timer1 Activation Bit (1=Turn TMR1 On, 0=Turn Off)

188
T2CON (Address 0x12, Bank 0 Only)
6-3 TOUTPS3:TOUTPS0 Timer2 Output Postscale Select (See Chapter 8)
2 TMR2ON Timer2 Activation Bit (1=Turn TMR2 On, 0=Turn Off)
1-0 T2CKPS1:T2CKPS0 Timer2 Clock Prescale Select (See Chapter 8)
CCP1CON (Address 0x17, Bank 0 Only)
5-4 CCP1X:CCP1Y PWM LSBs (In Order: MSB-LSB)
3-0 CCP1M3:CCP1M0 CCP1 Mode Select Bits (See Chapter 8)
CCP2CON (Address 0x1D, Bank 0 Only)
5-4 CCP2X:CCP2Y PWM LSBs (In Order: MSB-LSB)
3-0 CCP2M3:CCP2M0 CCP2 Mode Select Bits (See Chapter 8)
TRISE (Address 0x89, Bank 1 Only)
7 IBF Parallel Slave Port, Input Buffer Full (See Chapter 8)
6 OBF Parallel Slave Port, Output Buffer Full (See Chapter 8)
5 IBOV PSP, Input Buffer Overflow Detect Bit (See Chapter 8)
4 PSPMODE (1=PSP Mode, 0= General I/O)
2 RE2 Direction Bit (1=Input, 0=Output)
1 RE1 Direction Bit (1=Input, 0= Output)
0 RE0 Direction Bit (1=Input, 0=Output)
EECON1 (Address 0x8C, Bank 3 Only)
7 EEPGD Program/Data Memory Select Bit (1=FLASH, 0=EEPROM)
3 WRERR EEPROM Error Flag Bit (1=Write-Op Prematurely Terminated)
(0=Write-Op OK)
2 WREN EEPROM Write-Enable Bit (1=Enable, 0=Disable)
1 WR Start Write Process (Automatically Cleared by Hardware)
0 RD Start Read Process (Auto Cleared by Hardware)

189
TXSTA (Address 0x98, Bank 1 Only)
7 CSRC Clock Source Select Bit (See Chapter 9)
6 TX9 9-Bit Transmit Enable Bit (See Chapter 9)
5 TXEN Transmit Enable Bit (1=Enabled, 0=Disabled)
4 SYNC USART Mode Select (1=Synchronous, 0=Asynchronous)
2 BRGH High Baud Rate Select Bit (1=High Speed, 0=Low Speed)
1 TRMT Transmit Shift Register Status Bit (1=TSR Empty, 0=TSR Full)
0 TX9D 9th
Bit of Transmitted Data
RCSTA (Address 0x18, Bank 0 Only)
7 SPEN Serial Port Enable Bit (1=Enable USART, 0=Disable USART)
6 RX9 9-Bit Receive Enable Bit (1=Enable, 0=Disable)
5 SREN Single Receive Enable Bit (See Chapter 9)
4 CREN Continuous Receive Enable Bit (See Chapter 9)
3 ADDEN Address Detect Enable Bit (See Chapter 9)
2 FERR Frame Error Status Bit (1=Error, 0= No Error) (See Chapter 9)
1 OERR Overrun Error Bit (1=Error, 0=No Error) (See Chapter 9)
0 RX9D 9th
Received Data Bit
SSPCON (Address 0x14, Bank 0 Only)
7 WCOL Write Collision Detect Bit (See Chapter 9)
6 SSPOV Receive Overflow Indicator Bit (See Chapter 9)
5 SSPEN Synchronous Serial Port Enable Bit (1=Enable, 0=Disable)
4 CKP Clock Polarity Select Bit (See Chapter 9)
3-0 SSPM3:SSPM0 Synchronous Serial Port Mode Select Bits
(See Chapter 9)
SSPCON2 (Address 0x91, Bank 1 Only)
7 GCEN General Call Enable Bit (See Chapter 9)
6 ACKSTAT ACKN Status Bit (See Chapter 9)
5 ACKDT ACKN Data Bit (See Chapter 9)
4 ACKEN ACKN Sequence Enable Bit (See Chapter 9)
3 RCEN Receive Enable Bit (See Chapter 9)
2 PEN “STOP” Condition Enable Bit (See Chapter 9)
1 RSEN Repeated Start Enable Bit (See Chapter 9)
0 SEN “START” Condition Enable Bit (See Chapter 9)

190
SSPSTAT (Address 0x94, Bank 1 Only)
7 SMP Sample Bit (See Chapter 9)
6 CKE SPI Clock Edge Select Bit (See Chapter 9)
5 DA Data/Address Bit (See Chapter 9)
4 P “STOP” Bit (See Chapter 9)
3 S “START” Bit (See Chapter 9)
2 RW Read/Write Bit Information (See Chapter 9)
1 UA Update Address (See Chapter 9)
0 BF Buffer Full (Auto Cleared When SSPBUF is Read) (See Chapter 9)

191
Appendix D --- PIC16F877 Register File Map
ADDR Bank 0 Bank 1 Bank 2 Bank3
0x00 INDF INDF INDF INDF
0x01 TMR0 OPTION_REG TMR0 OPTION_REG
0x02 PCL PCL PCL PCL
0x03 STATUS STATUS STATUS STATUS
0x04 FSR FSR FSR FSR
0x05 PORTA TRISA ==== ====
0x06 PORTB TRISB PORTB TRISB
0x07 PORTC TRISC ==== ====
0x08 PORTD TRISD ==== ====
0x09 PORTE TRISE ==== ====
0x0A PCLATH PCLATH PCLATH PCLATH
0x0B INTCON INTCON INTCON INTCON
0x0C PIR1 PIE1 EEDATA EECON1
0x0D PIR2 PIE2 EEADR EECON2
0x0E TMR1L PCON EEDATH ====
0x0F TMR1H ==== EEADRH ====
0x10 T1CON ====
0x11 TMR2 SSPCON2
0x12 T2CON PR2
0x13 SSPBUF SSPADD
0x14 SSPCON SSPSTAT
0x15 CCPR1L ====
0x16 CCPR1H ====
0x17 CCP1CON ====
0x18 RCSTA TXSTA
0x19 TXREG SPBRG
0x1A RCREG ====
0x1B CCPR2L ====
0x1C CCPR2H ====
0x1D CCP2CON ====
0x1E ADRESH ADRESL
0x1F ADCON0 ADCON1
Notes:
1) Entries marked “====” are reserved – do not use them.
2) Blank Entries are Usable as RAM
3) RAM Addresses from 0x20-Through-0x7F Are Freely Available
4) RAM from 0x70-through-0x7F Do Not Need To Switch Banks
5) Indirect Addresses Must Add 0x80 For Bank 1 & Bank 3

192
Appendix E --- PIC16F877 Pin Function Map
PIN # Functions PIN # Functions
1 /MCLR, Vpp 40 RB7, PGD
2 RA0, AN0 39 RB6, PGC
3 RA1, AN1 38 RB5
4 RA2, AN2, Vref- 37 RB4
5 RA3, AN3, Vref+ 36 RB3, PGM
6 RA4, T0CKI 35 RB2
7 RA5, AN4, /SS 34 RB1
8 RE0, AN5, /RD 33 RB0, INT
9 RE1, AN6, /WR 32 Vdd (+5 Volts)
10 RE2, AN7, /CS 31 Vss (Ground)
11 Vdd (+5 Volts) 30 RD7, PSP7
12 Vss (Ground) 29 RD6, PSP6
13 OSC1, CLKIN 28 RD5, PSP5
14 OSC2, CLKOUT 27 RD4, PSP4
15 RC0, T1OSO, T1CKI 26 RC7, RX, DT
16 RC1, T1OSI, CCP2 25 RC6, TX, CK
17 RC2, CCP1 24 RC5, SDO
18 RC3, SCK, SCL 23 RC4, SDI, SDA
19 RD0, PSP0 22 RD3, PSP3
20 RD1, PSP1 21 RD2, PSP2

193
Appendix F --- Save Register / Restore Registers
on Interrupt
The code to save the W and STATUS registers at the start of an interrupt is as
follows:
W_TEMP: EQU 0x70 ; Storage for W Register
STATUS_T: EQU 0x71 ; Storage for STATUS Register
MOVWF STATUS_T ; Save STATUS
The code to restore the same is as follows:
SWAPF STATUS_T,W ; Get Saved STATUS
MOVWF STATUS ; Restore STATUS
SWAPF W_TEMP,W
Note: The process of saving and restoring the registers can be simplified by using
assembly language MACROs. These are called “PUSH” and “POP” and are defined as
follows:
PUSH: MACRO
MOVWF STATUS_T ; Save STATUS
ENDM
POP: MACRO
SWAPF STATUS_T,W ; Get Saved STATUS
MOVWF STATUS ; Restore STATUS
SWAPF W_TEMP,W
ENDM

194
To use these MACROs, do as follows:
INT_SERVICE:
PUSH ; Inserts the “PUSH” MACRO’s Code Here
---- Do The Service ------
POP ; Inserts the “POP” MACRO’s Code Here
RETFIE
The MACRO-code is translated verbatim into these places where they are called.
They are NOT subroutines! The code is just repeated in each place where the MACRO is
called.

195
References
1) All material covering DSP and the Median filter:
Digital Signal Processing: Theory, Applications, and Hardware
By Richard A. Haddad & Thomas W. Parsons
Copyright 1991 by W. H. Freeman and Company
2) All material & facts about the PIC16F877:
PIC16F87X Data Book
MPLAB Manual
MPASM Manual
Copyright 2001 by Microchip Technology, Inc.
3) Material on FSK, ADPCM, PWM, and Manchester:
Data and Computer Communications
By William Stallings
Copyright 1994 by Macmillian Publishing Company
And
Speech Coding: A Computer Laboratory Textbook
By Barnwell, Nayebi, and Richardson
Copyright 1996 by John Wiley & Sons, Inc.
4) Material on microprocessor & microcontroller interfacing:
Interfacing
By Stephen E. Derenzo
Copyright 1990 by Prentice-Hall, Inc.
5) Material on Dithering (Chapter 7):
FAB
By Neil Gershenfeld
Copyright 2005 by Basic Books, A Member of the Perseus Books Group
6) Material on Direct Digital Synthesis (Chapter 7):
ARRL Handbook (of 1986)
Copyright 1985 by the American Radio Relay League
7) Material on Speech Compression (Chapter 10):
Time-Compressed Speech, Volumes I, II, & III
By Sam Duker
Copyright 1974 by Sam Duker

Esp with thepic16f877

More Related Content

What's hot (20)

Similar to Esp with thepic16f877 (20)

Recently uploaded (20)

Esp with thepic16f877