SlideShare a Scribd company logo
Testing DDR memory
            with ICT
Problems

Complexity
High data rates
Differences between manufacturers
Lack of access
Complexity

Many instructions needed before the device can be Written to or
Read from.
CK/CK_bar crosspoint needs to be set carefully using a
dedicated logic family for the clock pins.
Burst modes need to be understood, writing 0xA to location 0
and then 0x5 to location 1 is pointless.
Testing ddr memory with ict
High Data Rates

Devices may have a Minimum speed
Can be overridden by disabling DLL
Sometimes get unpredictable behaviour
Differences between manufacturers

Samsung may support CAS Latency (CL) of 2,3,4
Hynix may support CAS Latency (CL) of 3,4,5
Lack of access

Use Bscan but need access to clocks otherwise long test times
and long compilation times.
Long refresh times mean data can be lost.
Problem with noise on clocks.
Test Development Process and
Debug Methodology
DQS testing
DLL disabling
Silicon Nail testing (Boundary Scan).
Dual stage fixturing
Test modes
DQS testing

DQS (if available) always works.
The device does not need any data written to it.
The device should be setup via the mode registers.
A read command can then be issued and just the DQS pins
checked.
This allows the engineer to verify that the Mode register
programming is working, before trying to write and read from the
device.
Testing ddr memory with ict
DLL Disabling

If the DLL is enabled then the device may have a minimum
clock speed which precludes it being tested at ICT.
Some devices allow the disabling of the DLL to allow the device
to run at low speed.
DLL disabling may cause odd behaviour of the device’s CAS
latency which may or may not be documented.
Testing ddr memory with ict
Silicon nail testing

If there is no access then Silicon nails can be used
BUT depending on scan chain length data may be lost.
The memory needs to be refreshed on a minimum time interval,
if this interval is exceeded then the memory will forget what was
programmed.
Try testing the data strobe instead as this a logic function of the
part, not a memory function
Sometimes Data strobes are not used/not available.
Silicon nail testing cont’d

Direct access to clocks is very useful as without it compile times
and run times can be excessive, resulting in a lack of the will to
live on behalf of the test engineer.
A 20 minute compile time is NOT conducive to test debug.
If the clocks are accessible then in a conventional fixture they
can be subject to noise from other parts of the board such as
switched mode power supplies, resulting in an intermittent test.
Dual stage fixturing

Directly clocking the device may result in intermittent tests
caused by the clock probes being affected by noise from other
parts of the board, such as switched mode power supplies.
If clocking the device directly consider using a dual stage fixture
to minimize noise on the clocks.
Test modes

Some GDDR3 devices have a Jedec defined test mode.
This test mode converts the device’s IO pins to a parallel in,
serial out shift register.
This greatly simplifies test development and debug.
It is still important to have direct access to the clock input of the
shift register when the device is in its test mode.
This is NOT the same as the memory clock input.
Dual stage fixtures are beneficial to eliminate noise on this clock
input.
Testing ddr memory with ict
Testing ddr memory with ict
Test modes continued

New IEEE standard for test modes on memories.
P1581
Memory Test Development

If you have multiple device manufacturers and no way of
knowing what’s loaded on the board then don’t even start.
If limited access get access to the clocks, if only for debug.
Disable the DLL so you can run at low speed
Just test the data strobes, don’t worry about writing and reading
yet.
If the device has a test mode, use it.

More Related Content

PPTX
Optimizing I/O’s and ESD protection to reduce power consumption in SOI applic...
PPTX
VARIADORES DE FRECUENCIA ACS880 ABB - EECOL
PPTX
Dynamic Voltage and Frequency Scaling
PDF
MANUAL DE SERVIÇO Alan 8001
PPT
Foundation fieldbus
PDF
BLtouch marlin configuration
PDF
德國量子光動力系統 Supertuning 組合
PDF
OCP Server Memory Channel Testing DRAFT
Optimizing I/O’s and ESD protection to reduce power consumption in SOI applic...
VARIADORES DE FRECUENCIA ACS880 ABB - EECOL
Dynamic Voltage and Frequency Scaling
MANUAL DE SERVIÇO Alan 8001
Foundation fieldbus
BLtouch marlin configuration
德國量子光動力系統 Supertuning 組合
OCP Server Memory Channel Testing DRAFT

Similar to Testing ddr memory with ict (20)

PDF
Reproducible Emulation of Analog Behavioral Models
PPTX
Design for testability and automatic test pattern generation
PDF
Webinar: Practical DDR Testing for Compliance, Validation and Debug
PDF
Get it right the first time lpddr4 validation and compliance test
PPTX
Real Time Debugging - What to do when a breakpoint just won't do
PDF
PowerDRC/LVS 2.2 released by POLYTEDA
PDF
Polyteda: Power DRC/LVS, October 2016
DOCX
RAM TESTBEANCH_VHDL ejemplos ------ .docx
PPTX
Prezentare tcs2011
PPT
Extreme Availability using Oracle 12c Features: Your very last system shutdown?
PPT
Introduction to Blackfin BF532 DSP
PDF
DDR DIMM Design
PDF
Polyteda Power DRC/LVS July 2016
PPT
Overview
PDF
MSc_project_report
PDF
Oracle R12 EBS Performance Tuning
PPT
POLYTEDA: Power DRC/LVS, June 2017
PPTX
Tools using for Repair Electronic devices.pptx.
PPTX
UGM 2015: X1149 workshop
PDF
Win10_Manual_ENG.pdf
Reproducible Emulation of Analog Behavioral Models
Design for testability and automatic test pattern generation
Webinar: Practical DDR Testing for Compliance, Validation and Debug
Get it right the first time lpddr4 validation and compliance test
Real Time Debugging - What to do when a breakpoint just won't do
PowerDRC/LVS 2.2 released by POLYTEDA
Polyteda: Power DRC/LVS, October 2016
RAM TESTBEANCH_VHDL ejemplos ------ .docx
Prezentare tcs2011
Extreme Availability using Oracle 12c Features: Your very last system shutdown?
Introduction to Blackfin BF532 DSP
DDR DIMM Design
Polyteda Power DRC/LVS July 2016
Overview
MSc_project_report
Oracle R12 EBS Performance Tuning
POLYTEDA: Power DRC/LVS, June 2017
Tools using for Repair Electronic devices.pptx.
UGM 2015: X1149 workshop
Win10_Manual_ENG.pdf
Ad

More from Interlatin (20)

PDF
Migration ux to windows - ICT i3070
PDF
In-line fixture design and release to production
PDF
Boundary scan for support engineers and technicians
PDF
Boundary Scan Basics - x1149 de Keysight
PDF
Keysight Mini-ICT - Testing Days México
PDF
Keysight i3070 (antes HP3070) Board Test Interface
PDF
Básicos de Functional Test Systems - Testing Days Tijuana
PDF
Functional Test Systems - Testing Days Bajío
PDF
Análisis de respuesta en frecuencia en osciloscopios
PDF
Presentación de ingeniería digital
PDF
Inter latin ptc i-iot - viaje de transformación de la manufactura
PDF
Elige la fuente de alimentación adecuada para tus aplicaciones
PDF
Sistemas de adquisición de datos en las Fábricas Inteligentes de la Industria...
PDF
Entrenamiento TestExec 8.1
PDF
Introducción a las Redes automotrices - CAN/LIN
PPTX
Pruebas de Campo para la industria de las Telecomunicaciones - Caracterizació...
PPTX
UGM 2015 Automotive Days
PPTX
UGM CAN PXI
PPTX
UGM 2015 Manager's Track
PDF
Interpretación y utilidad de datos estadísticos.
Migration ux to windows - ICT i3070
In-line fixture design and release to production
Boundary scan for support engineers and technicians
Boundary Scan Basics - x1149 de Keysight
Keysight Mini-ICT - Testing Days México
Keysight i3070 (antes HP3070) Board Test Interface
Básicos de Functional Test Systems - Testing Days Tijuana
Functional Test Systems - Testing Days Bajío
Análisis de respuesta en frecuencia en osciloscopios
Presentación de ingeniería digital
Inter latin ptc i-iot - viaje de transformación de la manufactura
Elige la fuente de alimentación adecuada para tus aplicaciones
Sistemas de adquisición de datos en las Fábricas Inteligentes de la Industria...
Entrenamiento TestExec 8.1
Introducción a las Redes automotrices - CAN/LIN
Pruebas de Campo para la industria de las Telecomunicaciones - Caracterizació...
UGM 2015 Automotive Days
UGM CAN PXI
UGM 2015 Manager's Track
Interpretación y utilidad de datos estadísticos.
Ad

Testing ddr memory with ict

  • 2. Problems Complexity High data rates Differences between manufacturers Lack of access
  • 3. Complexity Many instructions needed before the device can be Written to or Read from. CK/CK_bar crosspoint needs to be set carefully using a dedicated logic family for the clock pins. Burst modes need to be understood, writing 0xA to location 0 and then 0x5 to location 1 is pointless.
  • 5. High Data Rates Devices may have a Minimum speed Can be overridden by disabling DLL Sometimes get unpredictable behaviour
  • 6. Differences between manufacturers Samsung may support CAS Latency (CL) of 2,3,4 Hynix may support CAS Latency (CL) of 3,4,5
  • 7. Lack of access Use Bscan but need access to clocks otherwise long test times and long compilation times. Long refresh times mean data can be lost. Problem with noise on clocks.
  • 8. Test Development Process and Debug Methodology DQS testing DLL disabling Silicon Nail testing (Boundary Scan). Dual stage fixturing Test modes
  • 9. DQS testing DQS (if available) always works. The device does not need any data written to it. The device should be setup via the mode registers. A read command can then be issued and just the DQS pins checked. This allows the engineer to verify that the Mode register programming is working, before trying to write and read from the device.
  • 11. DLL Disabling If the DLL is enabled then the device may have a minimum clock speed which precludes it being tested at ICT. Some devices allow the disabling of the DLL to allow the device to run at low speed. DLL disabling may cause odd behaviour of the device’s CAS latency which may or may not be documented.
  • 13. Silicon nail testing If there is no access then Silicon nails can be used BUT depending on scan chain length data may be lost. The memory needs to be refreshed on a minimum time interval, if this interval is exceeded then the memory will forget what was programmed. Try testing the data strobe instead as this a logic function of the part, not a memory function Sometimes Data strobes are not used/not available.
  • 14. Silicon nail testing cont’d Direct access to clocks is very useful as without it compile times and run times can be excessive, resulting in a lack of the will to live on behalf of the test engineer. A 20 minute compile time is NOT conducive to test debug. If the clocks are accessible then in a conventional fixture they can be subject to noise from other parts of the board such as switched mode power supplies, resulting in an intermittent test.
  • 15. Dual stage fixturing Directly clocking the device may result in intermittent tests caused by the clock probes being affected by noise from other parts of the board, such as switched mode power supplies. If clocking the device directly consider using a dual stage fixture to minimize noise on the clocks.
  • 16. Test modes Some GDDR3 devices have a Jedec defined test mode. This test mode converts the device’s IO pins to a parallel in, serial out shift register. This greatly simplifies test development and debug. It is still important to have direct access to the clock input of the shift register when the device is in its test mode. This is NOT the same as the memory clock input. Dual stage fixtures are beneficial to eliminate noise on this clock input.
  • 19. Test modes continued New IEEE standard for test modes on memories. P1581
  • 20. Memory Test Development If you have multiple device manufacturers and no way of knowing what’s loaded on the board then don’t even start. If limited access get access to the clocks, if only for debug. Disable the DLL so you can run at low speed Just test the data strobes, don’t worry about writing and reading yet. If the device has a test mode, use it.