Tool/software:
Hello,
while running my application software with TMS320F28P650DK I have an ECC error.
Unfortunately I'm not able to debug using emulator because I'm using HWBIST at startup and this disconnects emulator.
The problem seems to be present at startup. If I clear the error after it happens it is not present anymore.
Unfortunately the issue can disappears adding portion of code for debug purpose or debug variables
I'm able to add a portion of code and a debug variable in a dedicated flash and ram section.
With this trick I've added the instruction:
"MemCfg_getUncorrErrorAddress"
after the failure.
The returned value is 0x910E
The value is always the same.
The strange thing is that the issue is related to the content of flash memory.
I've a flash area reserved to calibration data. This area is from 0x90000 to 0x91FFF, currently it's used only the area from 0x90000 to 0x90067.
This area is read at startup and loaded into RAM (a different ram section from the one returned by "MemCfg_getUncorrErrorAddress")
But this area is loaded before the error is set (I've added a clear of ECC error after the load)
If the area is blanck, no load is performed and no issue is present.
Loading the flash content from sample to another one gives the same error result (so there's no defect related to a single DSP).
Is the value returned form "MemCfg_getUncorrErrorAddress" reliable? Is there a known issue that generate false ECC?
Can it be related to .cmd filemapping? Can it be related to HWBIST (or other diagnostic check on flash/ram)?
Have you any suggestion to find the issue?
Can a filter on the error at startup be enough?
What I want to avoid is that the ECC error appears again.
Many thanks ,
Barbara
