[sdiy] Bad offset on VCA output

Wed Sep 3 12:39:21 CEST 2025

Hi Naoki and all,

You also need the exact same resistor values  also at the bases not only at the collectors.
There is a base bias current, and even with a 1% imbalance this can equate into some uV difference at the base, which will result in mV offsets at the output. But this doesn't explain the large offset.

I concur with the others, that it is quite difficult to do this VCA in a trimmless design. At least with your current part choices. This calls for pre-selected components (at least 0.1% resistors) and transistors that are matched to a uV level. If I introduce the 2mV worst case mismatch of the BCM847 in a simulation (of the circuit on my website, similar to yours, and no oscillations). Then I obtain offsets in the same order of magnitude to what you are seeing.

Best,
 René

--
synth at schmitzbits.de
 http://schmitzbits.de

________________________________
Von: Synth-diy <synth-diy-bounces at synth-diy.org> im Auftrag von Naoki Iwakami via Synth-diy <synth-diy at synth-diy.org>
Gesendet: Mittwoch, 3. September 2025 03:51
An: Magnus Danielson <magnus at rubidium.se>
Cc: synth-diy at synth-diy.org <synth-diy at synth-diy.org>
Betreff: Re: [sdiy] Bad offset on VCA output

Hi Magnus, thanks for the reply

In my experience, any diff-pair OTA like this needs triming, because reality does not give us very well-matches components and also one needs to be quite protective about diff-pair traces and keep them away from other sources, potentially use guard-rings.

This might be the issue. I removed a part of the plated ground from the PCB. The offset changed then. Strictly symmetric routing may be necessary.

Temperature will be a factor, so just heating/cooling becomes an issue. The BCM847 is a pair of transistors in one package, this helps to thermally connect them, but for precision work like the MAT series of transistors, you actually have 4 transistors in parallel for each, and they interconnect such that temperature gradients cancel first degree. This is why such transistors is prefered in expo-converters, but it is really just he same diff-pair or "long tailed pair" as you do an VCA.

BCM847 are matched pair transistors.  I first tried BC847DW1 that are not-particularly-matched pair transistors and picked up mached ones.  Only 1 out of 10 was usable.  BCM847 are reasonably matched and I can pick up better-matched ones if necessary.  It's a kind of trimming though :-)

I don't think you can avoid trimmers, not with these components at least. You really have to choose to either have trimmers or DC-blocking caps in the audio path, you really can't have both unless you spend more on the components, and even then you would improve with trimmers.
Actually, I'm planning to use this for CVs, too.  So DC capability is a must.  This VCA is meant to be used in a larger circuit. When a coupling capacitor is needed, I'll put it externally. I can put a trimmer terminal as a last resort, but will try trimless a little more.  I'm pretty sure it's doable if the PCB is larger.

-- Naoki

On Tue, Sep 2, 2025 at 3:13 PM Magnus Danielson <magnus at rubidium.se<mailto:magnus at rubidium.se>> wrote:

Hi Naoki,

Den 2025-08-27 kl. 09:28, skrev Naoki Iwakami via Synth-diy:
I'm developing a small trimless VCA of size 7/8" x 3/8", but suffering from bad offset on output such as 0.4V for zero input with 5V CV.  I tried the same circuit on a breadboard and 3" x 2" universal PCB.  Both worked fine (no bad offsets).  I suspected crowded PCB layout did some harm so let the gain unit transistor pair hang in the air to take some distance from the PCB — the problem disappeared then.  I will redesign the PCB to eliminate this issue but cannot figure out what to move to take distance from the transistor pair.

Does anyone have similar experience, such as cupper beneath a semiconductor changes the behavior, transistors and opamps work correctly only when they keep certain distances, and so on?  I'm using relatively small SMD components for this project (SOT-363, TSOT23-8, 0603, etc.).

I posted a blog article about this issue https://gaje.jp/2025/08/26/7810/

In my experience, any diff-pair OTA like this needs triming, because reality does not give us very well-matches components and also one needs to be quite protective about diff-pair traces and keep them away from other sources, potentially use guard-rings.

Temperature will be a factor, so just heating/cooling becomes an issue. The BCM847 is a pair of transistors in one package, this helps to thermally connect them, but for precision work like the MAT series of transistors, you actually have 4 transistors in parallel for each, and they interconnect such that temperature gradients cancel first degree. This is why such transistors is prefered in expo-converters, but it is really just he same diff-pair or "long tailed pair" as you do an VCA.

A challenge you also get is a hidden one, the input audio and output audio both convey DC and thus DC offsets is not blocked. This put requirements on the design, and the electrolytic caps is the easy way to cheap around from trimming DC, but here you try to avoid both and that is a higher challenge.

I don't think you can avoid trimmers, not with these components at least. You really have to choose to either have trimmers or DC-blocking caps in the audio path, you really can't have both unless you spend more on the components, and even then you would improve with trimmers.

Cheers,
Magnus

--
岩上直樹
Naoki Iwakami
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/20250903/d06df3a3/attachment.htm>