Unusual behavior of iMAX6Q based custom hardware; when it's going to scale - up !!!

[peteramond]

Dear All,

I have done 50 units of production with my custom hardware and some of the boards are working really well and some are not (10 are working out of 50, others are getting kernel crash). And also previously we did 50 units of production with the same design and out of them 30 boards worked really well. So now we need to isolate the problem for this random issue and need to identify whether it is manufacturing issue or another design issue before we go for mass production. 

Memory AS4C256M16D3A-12BCN

Processor - iMAX6Q

Presently I'm using NXP tools stress test, Stress app test and VTOS (KOZIO) memory verification and stress test tools for memory fault isolation process. But I couldn't identify any issue in memory layout by using those tools. All the tests passed. I have done DDR memory calibration for my imx6q custom hardware design using NXP stress tester version 2.9 and got proper and reliable read,write,dqs gating values for DDR memory. And I updated u-boot configs according to those values.

1) Could you please go through with the attached files of NXP calibration results ? Can you identify any problem with those calibrations ?

2) According to your experience what could be the reason for some boards are working and some are not in same production batch ?

Regards,

Peter.

[igorpadykov]

Hi Peter

>1) Could you please go through with the attached files of NXP calibration results ?
>Can you identify any problem with those calibrations ?

both attached logs are fine showing that tests passed. One can try to rerun ddr test on
problematic boards to try find more accurate settings.

>2) According to your experience what could be the reason for some boards are working
>and some are not in same production batch ?

variations of chip characteristics  (using example of power consumption and associated with it temperature)
are described in AN5215 iMX6 Temperature Sensor Module part B.2. Die process variation

"Variations in die-to-die processing results in devices with a range of power consumptions.
These variations cause two individual parts that are manufactured identically behave
differently. Comparing the upper and lower extremes of the top histogram in Figure 5,
devices can be up to 2 W different. Using a thermal resistance of 12 degrees Celsius
per Watt, there is a difference of up to 24°C, as shown on the bottom histogram in Figure 5"

https://www.nxp.com/docs/en/application-note/AN5215.pdf

So one can pay attention to processor power supplies noise and requirements for them
(like ripples, filtering) provided in i.MX6 System Development User’s Guide
https://www.nxp.com/docs/en/user-guide/IMX6DQ6SDLHDG.pdf

Best regards
igor
-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------