Identifying and Overcoming Noise in Data Acquisition

Identifying and Overcoming
Noise in Data Acquisition
William Chen
Product Manager

2
Copyright © Yokogawa Meters & Instruments Corporation
Kristina Neahr
Marketing Specialist
Yokogawa Corporation of America
Newnan, GA
kristina.neahr@us.yokogawa.com
1-800-888-6400 ext. 2611
tmi.yokogawa.com
Host & Panelist

3
PC’s Speakers - Audio Broadcast
To hear the audio through your PC, select the Communicate Tab
and Join the Audio Broadcast.
Recorded Presentation
A recording of this presentation will be posted under our
technical library of our web page.
Chat: Select and send all questions to “Panelist” during the
Webinar presentation.
Yokogawa Webinar – Housekeeping Issues

4
William Chen
Product Manager
Yokogawa Corporation of America
Newnan, GA
William.Chen@us.yokogawa.com
1-800-888-6400 Ext 2537
tmi.yokogawa.com
Presenter

5
If you have any questions for one of these
Webinar Topics, please send them to the
below e-mail. I will try to answer them
during the Webinar or directly afterwards.
Ø webinarwednesdays@us.yokogawa.com
Questions

6
■ Data Acquisition Overview
• Applications By Speed and Signal Types
-  Low speed monitoring and recording, High speed single shot, Repetitive
waveform monitoring, Memory blocks (Sequential store), High Speed
continuous monitoring.
■ Quantization noise
• Vertical resolution, LSB, Gain
■ Internal A/D noise
• What do accuracy specifications mean and how do they reflect the noise
characteristics of the DAQ hardware?
■ Power line noise
• Filtering, Integrating A/D
Overview

7
Overview
■ Time skew
• Inter channel skew, and simultaneous sampling
■ Aliasing noise
• Nyquist theory, Sampling rate/interval and frequency spectrum, AAF
■ Common mode noise
• Ground loops, common mode, isolation
■ Radiated noise (EMI)
• Crosstalk, DAQ product shielding, cable shielding
■ Application Example
• Fuel Cell Impedance measurements

8
DAQ Applications
DC-1kHz
Temperature, pressure,
Static load, displacement
Production monitoring
Power line monitoring
1kHz-100kHz
Mechanical Electronics
Sound and Vibration
Automotive, Aerospace
100kHz-20MHz
Electrical performance
Digital, timing, pattern I/O
Consumer electronics
20MHz-1GHz +
Component design
RF, Microwave
PC-based Internal (most PC-centric)
Pro: low cost, multi-function, bus speed
Con: little/no signal conditioning,
almost always multiplexed, poor noise immunity Benchtop (least PC-centric, DL850E)
Pro: separate/isolated power, portable,
Better quality measuring hardware
Con: less channel density, higher cost,
large footprint, slower for automation
PC-based External (SL1000, MX100, GM10)
Pro: most scalable, better quality measurement HW,
fast performance for PC automation,
lower cost, high bus speed, high channel density

9
DAQ Signal Types
■ Analog Input
• DC/AC Voltages
•  “Special” sensors
-  Accelerometer, ICP Microphone
-  Strain Gage, Load Cell
-  RTD, Thermistor, Resistance
■ Analog Output
• DC, Function Generation, Arbitrary,
Sweeping
■ Digital Input / Output
• TTL/CMOS, Static & Buffered
■ Timing Measurements
• Event Counting, Delay, Period,
Frequency, Tachometer, Encoder, Time
Stamps

10
Analog Input Applications
■ Low Speed Monitoring and Recording
•  Machine monitoring
•  Process monitoring
•  Certification testing
•  Reliability testing
■ High Speed Single Shot
•  Startup/shutdown monitoring
•  Electrical response
•  Device characterization
•  Sweep testing
•  Destructive/explosion testing

11
■ Repetitive Waveform Monitoring
•  Vibration
•  Test stands/cells
•  Engine or combustion monitoring
•  Glitch measurements
■ Memory Blocks (Sequential store)
•  Low re-arm time
•  Continuity/glitch testing
•  Engine R&D

12
■ High Speed Continuous Monitoring
•  Also called: Free-Run, Streaming to PC, Circular buffered acquisition, FIFO buffer
•  In-vehicle/flight DAQ, high energy physics, real-time monitoring of multi-hour tests

13
Source 1: Quantization Noise
■ Most commonly affects:
•  Thermocouple measurements
•  Low voltage measurements
•  Ripple measurements
•  High speed measurements on
any voltage

14
Principle of Quantization Noise
■ Decimal System (Base 10)
•  Used by humans
•  Number of Digits
•  1 digit = 0-9 [10 counts]
•  2 digits = 00-99 [100 counts]
•  3 digits = 000-999 [1000 counts]
•  counts = 10(Number of Digits)
■ Binary System (Base 2)
•  Used by Analog-Digital converters
•  Number of Bits
•  1 bit = 02-12 [2 counts]
•  2 bits = 002-112 [4 counts]
•  3 bits = 0002-1112 [8 counts]
•  counts = 2(Number of Bits)
8 bit A/D = 28 counts = 256
10 bit A/D = 210 counts = 1024
12 bit A/D = 212 counts = 4096
14 bit A/D = 214 counts = 16384
16 bit A/D = 216 counts = 65536
Binary (8
bits)
Decimal
(raw)
Voltage
(scaled)
00000000 0 -10V
00000001 1 -9.921875
00000010 2 -9.84375
00000011 3 -9.765625
00000100 4 -9.6875
00000101 5 -9.609375
… … …
11111111 255 +10V

15
Principle of Quantization Noise
■ Bit Resolution determines how many counts exist across the full scale
■ A Least Significant Bit (LSB) is the smallest voltage change that can be measured
•  1 LSB = 𝑅 𝑎𝑛𝑔𝑒/𝐶𝑜𝑢𝑛𝑡𝑠  𝑜𝑟 𝑅 𝑎𝑛𝑔𝑒/2↑𝑏𝑖𝑡𝑠 𝑜𝑓 𝑟𝑒𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛  
•  8 bit: 10 𝑉 −(−10 𝑉)/2↑8   = 20 𝑉/256 = 78.125 mV
•  12 bit: 10 𝑉 −(−10 𝑉)/2↑12   = 20 𝑉/4096  = 4.883 mV
+10V
-10V
b0
b1
b2
b3
b4
b5
b6
b7
A/DPGIA
Full scale
1 LSB

16
Quantization Noise Solution
■ Typical result achieved with digital software filter (poor)
■ SW Filter used: Sharp, Lowpass, fcutoff=2%*fsample, 88th order

17
Quantization Noise Solution
■ Hardware solution: Decrease (match) the range, or increase the bit
resolution (1 LSB = [range / counts])
■ Quantization noise exists even if it is not visually present
■ Improving this will improve analysis accuracy

18
AC Coupling
■ AC coupling will dramatically reduce quantization noise when:
•  You are interested in analyzing the AC content of a waveform
•  A DC offset is present
Must Use
±2V Range
Due to DC
Offset
Enable AC/DC
Coupling
Circuit,
allowing use
of a lower
Range
(±50mV)

19
Source 2: A/D Internal Noise
■ Noise described by printed specifications
(accuracy)
■ Main factors contributing to internal noise
•  Nonlinearity of the A/D converter itself
(differential & integral nonlinearity)
•  PGIA & A/D block – gain and offset
•  Thermal effects and thermal stability of the
entire digitizer
■ Accuracy specifications are highly
inconsistent in literature across vendors

20
Source 2: A/D Internal Noise
■ Various methods of reporting A/D internal noise
•  Accuracy (gain and offset)
•  Absolute Accuracy
•  Signal to Noise Ratio (SNR)
•  Effective Number of Bits (ENOB)
•  Noise Floor (dB)
•  Spurious Free Dynamic Range (SFDR)
■ Improving A/D internal noise:
•  Choose a higher precision digitizer
•  Maintain a stable environmental temperature
•  Reduce the Bandwidth used

21
Source 3: Power Line Noise
■ Power Line Noise also called “Pick Up” or “Hum”
■ Can be conducted (through DAQ device or power supply) or radiated
(through EMI), internal or external source
■ Noise frequency always occurs at the power line frequency (50/60/400Hz)
Clean Signal
Signal with power
line noise

22
Integrating A/D Converter
■ Integrating A/D converters quantize by time rather than by voltage
■ Better linearity and accuracy, no missing codes
■ Eliminate noise occuring at the integration frequencies (and multiples)
■ i.e. 16.67msec integration period will filter out (60Hz, 120Hz, 180Hz, …)
noise
■ Best low speed solution
Normal Sampling
Time
Voltage
Signal with power
supply noise
Voltage corresponding
to the value after the
conversion
For Integration-type A/D
Time
Voltage
Signal with power
supply noise
Voltage corresponding
to the value after the
conversion

23
Other Power Line Noise
■ For Slow speed signals measurements: A “Moving Average” software filter
is nearly as effective as an integrating A/D converter
■ Power Line Filters to remove high frequency RF noise from power
■ Isolated Input to Data Acquisition hardware

24
Source 4: Time Skew
■ Many Data Acquisition systems do not measure time aligned data
■ Differential measurements can have time skew between + and – terminals
(pseudo-differential)
Actual Signal Measured Signal

25
Principle of Time Skew
■ Multiplexing Sampling
channel clock [inter-channel skew]
scan clock
Ch1
Ch2
Ch3
Ch4
…
■ Simultaneous Sampling
Ch1
Ch2
Ch3
Ch4
…

26
Source 5: Aliasing Noise
■ Aliasing or Fold-Over Distortion
■ Occurs when higher frequency content exists beyond the sampling
bandwidth

27
Principle of Aliasing
■ Any Analog Signal or Non-sinusoidal waveform, actually consists of Sine
Waves of various:
■  Amplitudes
■  Frequencies
■  Phase
fmax

28
■ Red trace is actual signal, Green dots are measured values
■ Original waveform with 20
samples per period (fs = 20 f0)

29
■ Simple Example
•  fs = 16 samples/sec
■ Adequately Sampled
•  f0 = 1 Hz
■ Near “Nyquist frequency” or
sampling bandwidth
•  fN = ½ fs = 8 Hz
•  f0 = 7 Hz
■ Near “Nyquist frequency” or sampling
bandwidth
•  fN = ½ fs = 8 Hz
•  f0 = 7 HzUnder sampled
•  f0 = 11Hz
•  In this case we should see a flat line (since
the signal is above our measurement
bandwidth)
•  The incorrect signal introduces a type of
distortion into the measurement.

30
■ More realistic example
■ Original Signal
•  3 Hz content, 1Vpp
•  50 Hz content, 0.2Vpp
■ Aliased noise at 5 Hz
■ Sampled at 55 Hz
■ (fN = ½ fs = 27.5 Hz)
■ 27.5 – [50 - 27.5] = 5

31
Effect of Aliasing
■ Some Signals (i.e. a sawtooth
wave) have infinite harmonics
■ If the fundamental frequency of
the sawtooth wave is f0, and we
sample at 20*f0, what happens?
Result = increased noise floor

32
Aliasing Solution
■ A Hardware filter (low pass) eliminates fold-over distortion
■ Set the filter frequency close to Nyquist frequency (½ fs)
■ The filter must reside in Hardware – Software filters are ineffective
■ Especially important for sound and vibration testing
■ External HW filter is an option, newest technology uses SW-selectable, built into
instrument

33
Reconstructing the Signal
■ In theory, the reconstructed waveform must not possess any frequency
content >= sampling frequency
■ To perfectly reconstruct, bandwidth limit the re-creation/output
•  (apply an ideal low pass filter with (cutoff frequency = Nyquist frequency))
■ You can also use curve fitting to approximately reconstruct the signal
•  Linear interpolation, Sinusoidal interpolation, Spline
■  In practice, most engineers do not reconstruct waveforms prior to
analysis!
■  Therefore, apply a Practical over-sampling criteria fsample > 4x fmax

34
Source 6: Common Mode Noise
■ Original signal with no common mode
■ Signal with Common mode input to a
pseudo-differential device
■ Signal with Common mode measured
by isolated digitizer
* Noise depends on CMRR
■ Signal with Common mode noise
greater than digitizer specification

35
Principle of Common Mode
■ Voltage difference between sensor “ground” and instrument ground
■ Can cause permanent damage to measurement hardware
5V
0+Vcm
Vcm+5

36
Common Mode Noise
■ Typical causes of common mode voltages include:
•  Thermocouple measurements of powered devices
•  Battery or fuel cell testing
•  External sensor power supply
•  “Floating” sensors in noisy EMI environment

37
Common Mode and Normal Mode
Common Mode Noise
  Noise
Normal/Differential Mode Noise
  Actual signal and noise

38
Ground Loops
■ Return paths for current referred to as “ground”
■ Occurs when more than one ground connection path exists between
devices
DAQ DUT or sensor
Ground path through shield
or negative terminal of SE
measurement
Ground path through building
or earth ground

39
Ground Loops
■ Three ways ground loops cause equipment problems
•  Low currents circulating in the grounds generate voltages that can cause data
errors such as 60 Hz humming or high-frequency oscillations
•  High-energy transients will clear through circuit ground instead of earth ground
causing inrush or switching currents to damage equipment
•  Ground loops can cause common-mode noise between phase, neutral and
ground in a power distribution system. Noise injected into the power supplies will
pass on to the electronic components.

40
Common Mode Solution
■ Isolation Barriers
•  Prevents ground loops and negates common mode voltage
■ Safety: keeps high voltage/current away from people and equipment
■ Integrity: rejects unwanted voltages from affecting measurement accuracy

41
Common mode solution
■ Built-in Isolation Systems
■ Four isolation specifications to consider
•  [V1] Channel-to-ground isolation
•  [V2] Module-to-module isolation
(in a modular system)
•  [V3] Channel-to-channel isolation
•  [V4] Transient overvoltage protection
(or maximum withstand voltage)

42
Channel to Earth Isolation
  [V1] Channel-to-ground isolation
■ Accessible parts of instrument is safe
■ Prevents ground loops and common mode noise
Isolation Barrier
Analog
Module
A
Analog
Module
B
Analog
Module
C
Backplane

43
Module to Module Isolation
■ [V2] Channel-to-Channel isolation
■ Each module is isolated from each other
■ Provides noise immunity and overvoltage
protection

44
Channel to Channel Isolation
  [V3] Channel-to-Channel isolation
  Provides channel to channel protection from:
  Crosstalk
  High energy transients

45
Source 7: Radiated noise (EMI)
■ Certain environments are especially prone to noise:
•  Industrial or manufacturing facilities
•  Power engineering labs (supplies, UPS, etc.)
•  Motor or drive companies
•  High Energy Physics Laboratories
■ Emission sources
•  Fluorescent lighting (120 Hz sinusoidal)
•  Bipolar Power supplies
•  Internal components of desktop PC
•  Crosstalk (other sensors, particularly
active sensors)
■ Manifest as Common mode and
Normal mode Voltage

46
Reducing EMI noise effects
■ Faraday Cage principle – Electric field within a closed surface
is zero
■ Shield Cabling
•  Use standard shielded cable types (coaxial/BNC, twisted pair)
•  Use an external cable shield around each sensor-to-digitizer cable
•  Tie the cable shield to ground on only one side
•  Consider optical connections when feasible
■ Shield the Digitizer Module (vendor)
■ Shield the Station or Chassis (vendor)
■ Shield the Rack

47
Reducing EMI noise effects
■ More about Cabling
■ Avoid ribbon cables and unshielded terminal blocks at all
costs
■ Plug-in boards use high density connector, require terminal
block or external screw terminal
■ Many digitizer instruments have direct connections (clamp
terminals, NDIS, BNC)
■ Use true differential hardware with isolation to reject EMI
radiated as common mode voltage
■ For low speed/industrial applications, use current to transmit
signals (4-20mA)

48
Elements of a Noise Minimizing Digitizer
■ High Resolution A/D Converter (quantization noise)
■ Isolation barrier (common mode noise)
■ Low pass or AA filter (power line, aliasing noise)
■ Programmable gain (quantization noise)
■ BNC input (radiated noise)
■ Simultaneous sampling, independent channel hardware (time skew)
■ Attenuation
■ Mechanically shielded and enclosed hardware (EMI)
■ Acquisition Memory

49
Application Example: Fuel Cell Impedance Testing
When a cell is used for a long time, its
impedance will increase.
This causes a degradation or
inefficiency.
By measuring the fuel cell impedance,
we can verify the electrical nature of
the internal configuration of the fuel
cell.
Current Density (Output current from Cell)
Conductor Resistance of electrode
Reaction Resistance at Anode side
Reaction Resistance at Cathode side
Electrolyte Resistance
OutputVoltagefromCell
1.03V
(Theoretical)
Anode side Cathode side
Reaction Resistance
Electric Bilayer
Capacitance
Solution
Resistance
Fuel Cell
Equivalent Circuit

50
Application Example: Fuel Cell Impedance Testing
■ Key requirements:
•  Isolated output
•  Isolated input
•  High resolution, 16 bit
•  AC/DC coupling
•  Programmable gain
•  Hardware Filtering
•  Simultaneous Sampling
Hydrogen flow
Hydrogen outlet
Air(Oxygen) flow
Water and
air outlet
Electric Load DC
component
Electric Load AC
component
DC Voltage(cell voltage)
AC Voltage(ripple)
(optional)
switch box
Measured load
current

51
Isolated DAQ Instruments
Copyright © Yokogawa Electric Corporation
ScopeCorder DL850E Series
  1000 Vrms isolation
  100 MS/s high-speed sampling
  12-bit A/D resolution
  Complete built-in isolation system

52

53
High-Speed PC Based SL1000 Series
  12 types of input modules for
measuring:
  Voltage
  Strain
  Temperature
  Acceleration
  Frequency
  1000 Vrms isolation
  100 MS/s high-speed sampling

54
Summary – What We Hope We Did
■ Data Acquisition Overview
•  Applications By Speed and Signal Types
-  Low speed monitoring and recording, High speed single shot, Repetitive waveform monitoring,
Memory blocks (Sequential store), High Speed continuous monitoring.
■ Quantization noise
•  Vertical resolution, LSB, Gain
■ Internal A/D noise
•  What do accuracy specifications mean and how do they reflect the noise
characteristics of the DAQ hardware?
■ Power line noise
•  Filtering, Integrating A/D
■ Time skew
•  Inter channel skew, and simultaneous sampling
■ Aliasing noise
•  Nyquist theory, Sampling rate/interval and frequency spectrum, AAF
■ Common mode noise
•  Ground loops, common mode, isolation
■ Radiated noise (EMI)
•  Crosstalk, DAQ product shielding, cable shielding
■ Application Example
•  Fuel Cell Impedance measurements

55
Questions?

56
Thank You for Attending

Identifying and Overcoming Noise in Data Acquisition

More Related Content

What's hot (20)

Viewers also liked (7)

Similar to Identifying and Overcoming Noise in Data Acquisition (20)

More from Yokogawa1 (20)

Recently uploaded (20)

Identifying and Overcoming Noise in Data Acquisition