SCELBI 1702 EPROM Contents Troubleshooting

I’ve recently encountered some issues assembling the small program that I use for checking out MEA on my SCELBI 8B. 1702 EPROMs are notorious for “forgetting” data and my first thought was to verify the contents of those EPROMs.

Since the cassette write function still worked, it turns out that the verification process is pretty easy. Here is what I did.

1) I captured the contents of the MEA proms (060-000 to 077-377) to an audio AIFF file on my PC by connecting the audio output of the SCELBI to the mic in port of the PC. This assumes that the MEA cassette driver/utility to write the data out still works.
2) I converted the audio file to an Intel hex file using the utility that I had previously written.
3) I used an UNIX shell program named “diff” to compare this new INTEL hex file with the one that can be found at scelbi.com

What I found, was a couple of errors in the EPROM at page 73. I removed this EPROM from the PROM board, reprogrammed and verified it. After reinserting the EPROM and board, assembly of my little test program, once again, worked as expected.

As far as part 3 goes, I actually did a couple of additional verification steps by loading my OS/X SCELBI emulator with the Intel hex image converted from audio tape format and verifying that assembly of my test program also failed in that environment. I then dumped contents of suspicious memory locations with my emulator loaded with a good version of MEA and compared to dumps executed on the real SCELBI to confirm that those locations really had problems. Those extra steps aren’t strictly necessary, but helped confirm the issue, before I started the process of pulling boards and chips.

Note that I didn’t erase or use a new EPROM in this repair, but simply overwrote and verified contents of the existing EPROM with bad data. It will be interesting to see if these locations act up again in the future.

The whole repair process, including the extra verification steps and updating my ME1702 programmer with new firmware, probably took an hour or so. This was one of the easier vintage repairs that I’ve experienced, at least in recent memory.