AnsweredAssumed Answered

Inconsistent results on DDR calibration for custom i.MX6ULL board using DDR stress tester v2.70

Question asked by Chad Wolter on Feb 20, 2018
Latest reply on Feb 22, 2018 by Yuri Muhin


We are currently doing evaluation testing on a custom board utilizing an i.MX6ULL with 512MB of DDR3.  Using the "ddr_stress_tester_v2.70" tool from NXP and putting the board in Serial Downloader mode, and with the attached custom script, the Calibration test is run with the settings indicated by the attached screenshot.  The issue I am having is that when I run the calibration, and make the recommended settings to the *.inc file, and run the calibration again, it suggests different settings than it did during the previous run, which causes me to not trust what the tool is telling me.  Steps used are:


1.  Run the tester with the default *.inc file included with the tool (

2.  Using the settings from the attached screenshot, run the calibration test

3.  Put the calibrations into the custom *.inc file (attached)

4.  Run the calibration again, using settings from the attached screenshot

5.  Suggested calibrations are different than those recommended in step 2


As an illustration, I ran several iterations and recorded one particular register:  MPWLDECTRL1.  Initially, the tool suggested a value of 0x001F001F for this register.  I then modified the setmem /32 0x021B0810 line in the *.inc file to reflect this.  I ran the calibration again, and the tool now suggested a 0x001B001B setting for the MPWLDECTRL1 register.  I then ran the tool again, and a value of 0x001C001C was suggested.  I then ran the tool again, and a calibration of 0x001F001F was suggested (the original value!).  I attempted one final run using the same iterative process and a value of 0x001D001D was suggested.


Please give me an indication of what I might be doing wrong or if I am using the tool incorrectly.


Chad Wolter