Hi Glory.

I have developed several sets of code images masked into ROM for a number of processors (some for 8051 derivatives, some for other MCU's) and have a couple of experiences/observations to share.

First of all, I'll bet that we both suffer from a bit of anxiety when having to commit our code to 10,000's, 100,000's or 1,000,000's of copies. An error can cost "a lot of money" (insert your country's monetary denomination here).

Look for things that can vary between what you supplied to the manufacturer and what they returned to you for verification. I am assuming that either your runtime code performs a checksum and reports the result, or that you are check-summing the code via some external means.

If the HEX/BIN file you provided to the manufacturer did not specify the contents of the entire ROM image, the manufacturer may have filled the unspecified contents with a value different that you expect. For example, if you derived your initial checksum from a device programming utility, consider that some utilities initialize unused locations with 0xFF, whereas your manufacturer may have filled unused locations with 0x00 (8051 NOP instruction).

Also, some manufactures might embed test code in unused/reserved ROM locations and that test code can vary between manufacturers. For example, I remember an instance where code I supplied for masking into an MC68HC05 derivative check-summed differently between Motorola-built parts and Harris-built parts (second source). It turns out that each manufacturer embedded some self-test code, but each of their codes was slightly different. If your original checksum checked to the end of ROM without knowledge of the test code's presence, it could be wrong.

I'll try to think of more gotcha's I ran across, but those are a couple I remember right away.

Regards,

--Dan Henry