Your caps are bad. If you don't replace them, eventually the synth will die. If you don't replace the caps, some CMOS will go out, you'll experience all sorts of erratic behavior, then the Bridge Rectifier or regulators will go. Oberheim's are famous for not tuning if the caps go bad.

Replace the caps. The bridge rectifiers in the original design are simply too small and also need to be replaced. Replacing the regulators won't hurt either; the synth will run cooler.

An extreme mod has been made on my OB-8, probably has close to the same software, in the autotune loop by removing the caps that DC block the VCF from the VCA and I pass autotune every time because the synth has been restored properly.

There isn't a bug in the software.

-Mike

--- In oberheim@yahoogroups.com , "Ricard" <ricard2010@...> wrote:
>
>
>
> Here I go replying to myself, I haven't had any responses here on the forum, but one member responded privately (you know who you are, thanks!, and it's been a real nice chat afterwards too).
>
> What I'd like to ask is to one extent other people also have problems with the autotune.
>
> In searching for the problem with the autotune, I've analyzed not only
> the hardware but also the software program that runs on the Z80 processor.
>
> It turns out to be a fact that the autotune acceptance range is -24 .. 0 cents. First off, the range is rather large, and secondly it is offset in a strange way.The algorithm works like this:
>
> For each oscillator, the synth tries to calibrate the oscillator 50 times. For each time it measures the period of a pulse formed from four cycles of a target tone C4 (1046.502 Hz) tone. The resulting period, measured as 1/4.9152 MHz cycles, should be 4964 (hex). If the period is within +1/-0 of 4964 (hex), then it goes on to the next oscillator. Otherwise, if the last measurement cycle was
> within +256 / -0 from 4964 (hex), the oscillator is considered ok, otherwise it is considered failed, which will cause the corresponding LED to flash after the tune procedure is complete. 256 steps (or FF hex) steps from 4964 works out as 24 cents at the measurement frequency.
>
> I've verified that this is in fact the case by breaking up the autotune circuit and injecting a tone from another (digital) synth where I can set the frequency fairly exactly, and also verify the pitch using a tuner.
>
> So that's the reason for several oscillators failing on my machine; probably the general instability in the synth causes the oscillators not to be spot on, andif they're not, the oscillators are considered untuned if they are even half a hertz over the reference frequency, but not if they are up to 24 cents below.
>
> I believe this is a bug in the software, notwithstanding any instability in the machine due to aging components etc.
>
> All this is rather technical I know, so perhaps not really the right forum. Someone might find it useful I hope. My main reason for this post was to try and poll to what extent others also experience autotune problems, both on unrestored and restored machines.
>
> /Ricard
>