iMX6 board crashing at elevated temperatures

cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

iMX6 board crashing at elevated temperatures

1,308 Views
peter_hardman
Contributor I

We have a bunch of custom boards where the iMX6 is crashing when the board reaches elevated temperatures (SoC temperature 70-80C).  We have swapped the processor between working boards and failing boards, and the problem moves with the processor and does not stay with the board.  This suggests to me that the processor is either faulty or more sensitive to some parameter on the boards that other chips are not affected by.   We are seeing a significant fallout  (~50%) of our latest batch that is built with one specific date code.  

We are supposed to be getting industrial grade parts which should work to at least 100C where Linux will force a shutdown, but many boards are crashing well before that. 

Any suggestions what to look at?

Thanks

Labels (1)
0 Kudos
6 Replies

1,133 Views
igorpadykov
NXP Employee
NXP Employee

Hi Peter

yes reason may be processor as due to process variations some parts may be faster

and heating more, produce more noise, this is described in sect.B.2. Die process variation

AN5215 i.MX 6 Temperature Sensor Module

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

One can try to recheck recommendations given in IMX6DQ6SDLHDG, Hardware Development Guide for i.MX 6Quad, 6Dual, 6DualLite, 6Solo Families of Applic...

and rerun ddr test https://community.nxp.com/docs/DOC-105652 

Regarding "faulty", one can apply to local marketing office for using FA analysis procedure.

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

0 Kudos

1,133 Views
peter_hardman
Contributor I

The app note only points out the variability in the determination of the die temperature, along with the variation in the temperature generated by the processor during operation.  Since my systems are failing well below the expected temperature max, I would be surprised if any error in the monitored temperature was the reason. Chips are crashing when the die temperature is reported as 70C, and I have been able to verify with a thermocouple and reading the die temperauture shortly after boot, that the temperatures reported are relatively accurate.

We will run the DDR test and report back.

0 Kudos

1,133 Views
peter_hardman
Contributor I

We have run the DDR memory calibration and then started the memory tests.  As the board warms up the tests just stop after emitting a few garbled characters to the console.  There is no error message that the test failed the program just stops running.   We see the same behavior if I run the simple mtest from the bootloader.  The test will be running and then just stop in the middle of some iteration with no more output.

Is there a way to take the DDR out of the equation and make sure that the boot loader and any code it is executing are running for the iMX6 cache and not DDR?    If I run a simple loop in the boot loader and we are not accessing the DDR and the problem stops I am willing to believe that the DDR timing is the issue.  If the code still crashes/freezes then we can be more confident that the problem is internal to the iMX.

So how can this be accomplished?

0 Kudos

1,133 Views
igorpadykov
NXP Employee
NXP Employee

>Is there a way to take the DDR out of the equation and make sure that the boot loader and any code it is >executing are running for the iMX6 cache and not DDR? 

one can try to run test code from OCRAM.

~igor

0 Kudos

1,133 Views
igorpadykov
NXP Employee
NXP Employee

If heating up the part causes more failures, then the drive strength is
too low and the signals may not rise/fall fast enough. Drive strength of ddr signals

may be adjusted from i.MX6 side through IOMUXC_SW_PAD_CTL_PAD_x

registers and also from ddr memory side.

Best regards
igor

0 Kudos

1,133 Views
shaneggpierce
Contributor I

Getting same issue here too, in search usps tracking of a solution and already started troubleshooting. Will update if I find anything.

Regards,
Shane

0 Kudos