PIC24FJ64GA002 Address Trap Error

[LFern]

Hi All,
I have a target application running on a PIC24FJ64GA002 micro. Used CCS PCD V4.114 as the compiler. It is handling very complex state machine while nested interrupts are enabled.
Also the same target application has been deployed in around 30 boards which I tested long hours.
I saw some boards giving me address trap error but majority is working fine.
I can assure that they all going through the same state machine doing same computation one or the other time but few of the boards keep reporting address trap errors.
I am wondering why only some of the boards reporting address trap error while others working without any problem.
Do you think its could be a silicon errata issue of the micro or component problem?
Number of address errors also different board to board for a period of time like 1 board > 1000 trap errors.
I strongly believe that there could be software bugs that is a resultant to these trap issues. but difficult to understand is why some boards only effected.
Any comments?

[asmallri]

An address trap error can be a word alignment problem when you attempt to access a word variable on an odd address boundary. An intermittent problem is not unrealistic. Try adding an address trap error handler to capture the offending address.
_________________
Regards, Andrew

http://www.brushelectronics.com/software
Home of Ethernet, SD card and Encrypted Serial Bootloaders for PICs!!

[FvM]

In a first place, I would search for user code errors, particularly related to pointer operations, that can easily bring up address errors.

I observed various PCD bugs involving address errors, the latest with V4.112 and V4.121. Besides wrong data alignment, as mentioned by Andrew, also access of non-existing memory addresses, preferably due to non-initialized pointers has been the cause. Most of these have been simple cases, where the erroneous coding can be seen from the disassembly listing, occuring in the instruction before the error address, that can be determined in the address error trap rountine.

More complex cases have been caused e.g. by unpaired push/pop instructions. But these errors aren't accidental.

In your case, I would start to record error addresses with the address error traps to better understand the problem.