Interesting. I've been studying the schematic, and oddly enough I don't think it would make that much difference, at least if we were discussing something that had a brand-new, decent quality cap in that position, which most x0x'x do. However, if we were taking into account a 30 year-old cap, that may not have been the greatest quality, and considering electrolytic caps back then were not as well made, or efficient as the modern ones, there may be a small difference.
I would suggest that if anyone wants to truly test this on a x0x they should somehow find a 1uf, 25V or higher cap that is 20+ years old and put that in place of C29 and repeat your tests.
Rationale: C29 at first appears to be a smoothing cap (filters out any potential noise and other garbage, as well as preventing voltage drops on quick transients) for the +5.333V line going into the buffer that biases Q20 and Q21. If it was a DC-only circuit, it would be quite a difference because C29 would never charge, and at best it would act like a resistor and leak quite a bit of current to ground. However, in this case you have AC voltage coming into the bases of Q21, and while I haven't tried to calculate the AC voltage at that point (I'd rather just stick the probes of my meter there later and measure reality as opposed to theory), I would imagine there's more than enough AC voltage there to charge the cap. Once the cap charges, it can do it's job, which also is to keep that AC voltage from getting into the +5.333V supply. Do caps behave this way? If they didn't, voltage doublers wouldn't work.
Now, in a modern, brand-new cap, this would happen quickly enough to where it really shouldn't make any difference at all really (in theory
). However, a 30+ year-old cap, that wasn't made as well to start with, and by now probably can't hold a charge well enough anymore to be an effective smoother/filter even if it was biased correctly, I bet it would make a noticeable difference. If it's not filtering correctly, and not shunting all the AC to ground, every +5.333V connection will now have a portion of the signal on it. In some places it may partially cancel out, such as at the collectors of Q26, since the signal is also in coming into both bases. This may explain why the VCO doesn't go crazy under these conditions (or, it may not
). However, unless I haven't counted all the polarity inversions correctly, Q12, Q18, and Q21 don't have that advantage, and well, you can see where they are.
So, I think it would be worth sticking a really old cap at C29, backwards of course, and then see what happens. Think about one thing, and this same issue comes up in high-end audio circles when discussing vintage equipment, none of us are hearing a TB303 the way it sounded when it was new (unless, as the OP seems to be doing, treating one to a full rebuild, and probably with all new electrolytic caps). In fact, this may very well account for almost all the differences in sound that are perceived between a x0x and a TB303. Most of us are using the same active components, and a lot are probably the same age as the originals, but I don't know of anybody that's built a x0x with 30 year-old passive components that have been used for all those years. As an aside, if someone happens to find 30 year-old electrolytic caps that are NOS, don't even bother, as most likely they would not work at all. Like old cars, they have to be driven
I mean, even if I'm totally off base, at least this is food for thought.
Don Taylor, a.k.a. Brassteacher
"Jack of all trades, master of none, proficient at even fewer..."