Hi y'all,

I have the following on my breadboard. The first couple of stages are more or less based on Baja's UBS (see other forum), adapted for bipolar 15V, and the following two stages are from well known Ampeg and Crate high gain amps.
It sounds great, heaps of harmonics, good dynamics, well voiced for the speakers that i have, overall quite happy with how it turned out, etc.
Except for a tiny glitch:
When doing e.g. a big E power chord, then muting the strings, the sound undergoes a weird subtle "ducking" and it takes around a quarter of a second for it to come back. At first I thought it could be some sort of blocking behaviour, and lowered some coupling caps, etc, but the effect persisted.
If I remove C43 and R15, this behaviour disappears, but the sound instantly becomes more dull and flat, with less gain and harmonics.

So I am left with two question:

- On Ampeg preamps, there is no smoothing cap across inverting input and output on the distortion stages. Is there anything in the Variable Harmonics technology that can cause this glitch if a cap is used there?

- WHY does it sound SUBSTANTIALLY better (and with no audible reduction in top end, quite the opposite, more harmonics) when C43 and R15 are present?

Hope somebody can enlighten me or point at the right direction, thanks in advance

Quote from: Umlaut on December 14, 2023, 12:25:32 PMwhen C43 and R15 are present?

Where is R15?

Or is that R51?

If the wiggle next to C43, then it is oscillating, perhaps supersonic. Not a shock when you cascade THREE brands of "high gain" in one box. (I'm surprised it ever works.) It needs a wider layout or a treble-cut, such as that R-C network.

Yes, I meant R51, sorry.
The circuit is still on the breadboard, so oscillation is likely. But again, the issue is only present when the RC network is on. It disappears when I leave it unconnected. Could oscillation be responsible for a "better" sound? as mentioned, on this type of topology, Ampeg/Crate rely on pre and post treble cut rather than add a cap in the feedback loop. I am guessing that the bias shift network could be responsible for the ducking. WHY is what I am trying to figure out. i'll try to shuffle some components around and see what gives.

>

Quote from: Umlaut on December 14, 2023, 11:37:24 PMthe issue is only present when the RC network is on.

Then I mis-read. Carry on.

Will it be improved by increasing the value of R13 (and R16)?

Increasing those two results in an attenuation of the bias shift effect, at least according to the spice sim. But I could make small increments of, say 2k at a time, and see if it changes the behaviour. Thanks 

Surely the ducking behaviour is from the networks around C11 and C13.

Quote from: Umlaut on December 14, 2023, 12:25:32 PMOn Ampeg preamps, there is no smoothing cap across inverting input and output on the distortion stages. Is there anything in the Variable Harmonics technology that can cause this glitch if a cap is used there?

- WHY does it sound SUBSTANTIALLY better (and with no audible reduction in top end, quite the opposite, more harmonics) when C43 and R15 are present?

C43 and R51 are the same as a cap across the inverting input and output.   Realize R51 is a lot smaller than R12 so it roughly looks like R51 is short and C43 is a filter cap.

The presence of C43 removes some top end before the second clipper stage which will help tweak high frequency intermodulation products.

No filtering on second clipper stage but you have the R19 + C42 post filter.

On the first iteration, I did not have R51. I actually put it there to see if it mitigated that ducking effect. The idea behind C43 was indeed to filter some top end, so as to preserve the frequency response of that gain stage as in the original Uberschall circuit. For some reason, it sounds better than without it, except for that weird effect on sudden string mutings. I would happily get rid of that cap and add some more filtering before the third stage, if it wasn't for the fact that it  makes the overall sound richer and fuller. I'll return to it this afternoon and report

Well, I added a 3.3V zener between R14 and D2 to prevent it from bias shifting before clipping occurs, then changed C11v to 2.2u. Removed R51 and left C43. This seemed to cure the issue. Thanks to all for the insights!