From Michael E Caloroso Sent Sat, Mar 10th 2018, 02:49
I studied microprocessors during college and I know how they operate inside. CPUs can steer data into multiple registers for processing. Besides data, there are registers for program counters, interrupt vectors, relative addressing, etc. Two things clued me in: 1) it would crash on occasion and wipe out the patches, 2) when I ran the C7 routine to monitor autotune it would report rational numbers then as it warmed up they would be all zeros. So I had an intermittent temperature related malfunction - the worst to troubleshoot. After being lead down a few wrong paths, it dawned on me that there may be a register on the Z-80 that was going bad. I happened to have a surplus MOSTEK Z-80 in my spare stash - swapping it out solved the problem. There was indeed a register on the CPU substrate that was going bad, apparently one that didn't affect the operation of the rest of the synth. A very misleading symptom and the one IC I least expected to be bad. I've done this long enough to NEVER rule anything out, no matter how remote... MOSTEK made Z-80s under license for Zilog before they owned their own fab labs, so for all practical purposes the MOSTEK Z-80 is an authentic Zilog part. I've pulled completely dead CPUs before, but that's the first one I've pulled that was fractionally dead. Shortly after that a friend asked me to look into a malfunctioning guitar preamp and the reason the front panel buttons didn't work was a missing clock signal which was generated directly from the 8031 port pin. Yet the rest of the preamp worked. Sadly, that 40-pin 8031 was not socketed. Sent from my iCPUWrench, MC On 3/9/18, Brian Willoughby <xxxxxx@xxxxxxxxxxxx.xxx> wrote: > Thanks for the report, Michael. Should be great information for the > archives. > > Meanwhile, I had intended to respond to your earlier message about the = =E2=80=9Cbad=E2=80=9D > Z-80 with a hypothesis. I recently repaired a Roland CR-78 where one bit = of > a latch had partially burned out. The whole thing was working except for = the > Bass drum, and I tracked it down to that one bit. My thinking on the Z-80= is > that it must have been mostly functional if it was able to run the synth > apart from tuning, but perhaps there is a Port IO pin that is exclusive t= o > tuning and which had gone bad. Not that you can repair an individual pin > without replacing the whole chip, but I was curious. Typically, a CPU is > either completely dead or totally functional. It=E2=80=99s actually quite= rare to > have some features fail while most are still working (at least not unless > there is a bug in the firmware - in which case the same problem would app= ear > on all units). > > Again, thanks for detailing your work and discoveries. > > Brian Willoughby > > > On Mar 9, 2018, at 5:32 PM, Michael E Caloroso <xxx.xxxxxxxxxxx@xxxxx.xxx= > > wrote: >> The last mile to the finish line was the toughest... autotune >> wouldn't. Even though the display reads "6 TUNED" does NOT mean that >> it tuned the VCOs! After isolating many of the support ICs and >> confirming the audio path between the VCOs and autotune circuit the >> root problem turned out to be.... the Z80. EVEN THOUGH THE REST OF >> THE SYNTH FUNCTIONED. > >