Here is a followup on this problem.

Over the weekend I built a circuit using an Atmel 89S8252, 32k of external RAM, and associated interface circuitry. I loaded BAS52 (Intel version, since the others are too big to fit in 8k flash), and everything works (including saving program lines).

I did more tests on the Philips board from Futurlec. I hit the external memory pretty hard with a Bascom-8051 compiled program that would generate an array of random address/value pairs and write them, then read them back and compare. I was never able to get any failures writing/reading external RAM. However, I did notice in the spec sheet for the P89C51RD2 that this chip has a bit of internal RAM that is used with MOVX by default. This internal RAM can be disabled by setting bit 1 of "AUXR" (8EH). When I add an assembly language instruction to do this to the first line of BAS52.ASM (Intel version) and assemble this, the Philips board works fine! Thinking the extra instruction had moved the code to another spot in memory away from a bad spot, I tried changing the instruction to move a 0 to that location (for no effect), and the problem comes back. So it is the actual instruction that fixes the problem. I'm not sure why this is the case, since this is not necessary when using Bascom-8051 (and I'm sure Bascom doesn't know about this register, since I'm using a generic chip header file).

Still somewhat of a mystery but at least I have everything working now. I was also able to get PaulMon working with my homebuilt circuit after ANDing /PSEN and /RD together. This should speed up my program testing quite a bit since I will be able to download to RAM instead of waiting for flash to program.

Thanks again to all those who helped.

Frank Kienast