TMS320F28P650DK: RAM code changed unexpectedly on DK6

Hi Hang,

I'm not sure what could be causing this behavior and will pull in some other expert. ITRAP instruction causes and debugging is documented on the page below just for reference.

https://software-dl.ti.com/C2000/docs/c28x_interrupt_faq/html/index.html 

Best,

Kevin

Thanks for the info Kevin. Customer is stuck at this very early stage of software, please help solving this ASAP.

Hi,

Is there any update on this issue? Is it possible to reproduce this on BU's side?

Regards,

Hang.

Hi Hang,

Sorry for the delay. Still trying to get some help from other experts on the team.

Is this behavior happening on all their devices / boards?

Best,

Kevin

Hi Kevin,

They tested 7 boards and all of them has this issue. Would it be a risk in DK6 devices?

Regards,

Hang

Hang,

This is strange. Are you using same method to load the code on both the boards and also version of SW is same. Also can you disable "Run to main" option and then re-check the content ? If you are loading the code in RAM only and not getting the verification error from CCS loader then content may be getting updated later and not after the load. 

Vivek Singh

Hi Vivek,

Yes, same except choosing DK6 in the project properties:

Also, DK6 is loaded with XDS100 and DK9 is using XDS100V3

Regards,

Hang.

Hi,

Is there any update on this issue?

Did you try disabling run to main option and check it ?

Hi Vivek,

Sorry forget to update here. When halt at code start, the result is the same.

Are you sure it is part of your code ? If you are not getting verification error then it may not be part of your .out. 

Vivek Singh

Do you mean it's not part of the code shown in the memory Brower? When resume the code, we can see it run to the instruction at 0x83d3b.

You have not disable the verification check in flash GUI setting. Right ? 

Fast verification it is.

Can you try with full verification ?

It’s the same. I also tried using hex to load program and dump the memories as hex to compare, the code is changed unexpextedly without verification warnings

Also, manually writing particular values, like 0x180E, either in the memory browser or via array, would not work. Here I use a array to fill 0x180E and some other values to ram, and only 0x180E is changed when written to ram

Look like multiple locations are getting changed. Are these fixed location every time or keep changing ?

Are you loading the code in RAM only or flash as well ? Would it be possible to share the full project or all the linker cmd file ? It'll be also good idea to check the power supply on this board.

Vivek Singh

The location getting changed are not fixed but follow some pattern, for example, the location with value 0x180E would be changed. Loading in RAM and Flash both has this issue, RAM and flash are both getting changed. We have reproduced this issue using the led example in the device_support folder. We are checking power supply now.

Regards,

Hang.

Hang Yang said:

We are checking power supply now.

Thanks. Let me know what you find.

We've checked the 3.3V rail didn't see abnormal during programing. Any idea what might be the cause?

An update here, We tested DK9 on the same board, the issue remains. 

Hang,

I guess, you swapped the working part from TI evaluation board to customer custom board and it's failing now. Correct ? If yes, then it points to some board issue. Please confirm.

Vivek Singh

Vivek, 

No, DK9 is not from EVM but a new sample, we can not use DK9 on EVM since the package is different. The DK9 sample is not tested so it can not confirm it's board issue.

Assuming if there is board issue, is there any suggestion on what's related so that we can check it?