have you ever withnessed calculators not working anymore due to corrupted memory?

[Marc van Lemmen]

Hi, I had a Casio FX-602P where one memory failed after 10 years. When I put 3 in M06 and then recall M06 it returned zero. So one bit failing. Made a control program to check the rest, but only M06 had problems.

[sleibson]

HP made the NMOS ROMs for the HP 9825's HPL language. The 9825A's big HPL language ROM is a slide-out drawer in the side of the machine. Early in the manufacturing cycle, we had a problem with sodium contamination in the ROM processing line. Sufficiently contaminated ROMs would initially work but would die over time as the sodium atoms migrated into the MOS transistor channels and shorted out the transistors. You tested for contamination with a leakage test. Leaky ROMs still worked but you knew they would eventually die. In the earliest days, it was easy to get prototypes of the 9825A hardware but really hard to get tested and certified ROMs, so a few of us got sets of leaky language ROMs and soldered them up so we'd have complete 9825As to work on for developing I/O cards. My "leaky" language ROM died after a year or so, but by then we had the contamination problem under control.

The point of this little history snippet is that many IC manufacturing processes back in the 1970s and 1980s were plagued by some level of contamination. RAMs or ROMs, it didn't matter. A transistor is a transistor. So all of those old ICs are possibly subject to eventual destruction.

[pier4r]

Interesting. Thanks!

[edryer]

Quote:Hi, I had a Casio FX-602P where one memory failed after 10 years. When I put 3 in M06 and then recall M06 it returned zero. So one bit failing. Made a control program to check the rest, but only M06 had problem.

Interesting, although this is an old thread also I experienced this with an FX-602P, storing in M10 (all other of the 89 memories OK) I get corrupt results, that is I store a number it will have the 8th digit corrupt, if it is a shorter integer it will add that "digit" as some (non-random) bit.

Example "123 M IN 10" (store to memory 10)
And then when "MR 10" I get 123.0008, also with other numbers I also get bit corruption (on the the eight place only) these are reflected if I use the number resulting in incorrect results.

Later Casio's (such as the 1986? FX-3900P) I have never seen this issue - in fact the later calculators in the "P" series I think never suffered from memory issues, nor did the contemporary (limited likely to one memory) (1979) FX-80, FX-81 basic scientifics... nor I believe the FX-180P (1981?).

The Memory used in the FX-501/502/601/602P likely changed* in the next generation FX "P" machines emerging around 1984-1985.

Still.. usable...with that quirk, but it is an interesting machine, has a good programming model but the functions are limited, esp the stats stuff as a few years later Casio went way overboard on stats functions shifted on the numeric keys.

Casio's are very interesting... they have kept a uniform naming sequence generally for forty odd years, and even a mid 70's scientific can be seen today in the design of the FX-5800P.

Same with graphing... a 1996 model has mostly the same OS as a 2018 model... only a little has changed (such as RUN and MAT now on the same tab)... can switch between the two without thinking (9750G from 1996 to todays CG series).

*From the NEC uPD444G 1024 x 4-bit S-RAM in an SMD package used (also in the FX-702P)