Compressor design

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

for my forthcoming bass guitar amp I'm planning to provide a compressor.

Let's have a look at the Hammond S-6 chord organ schematics:

An externally hosted image should be here but it was not working when we last tested it.

Next to the upper right corner we find the Percussion section, mainly built around the transformers T1, T2 and the tubes V9, V10, both 6BA6/EF93 remote cutoff pentodes. These act as a VCA to get an output signal with a distinct attack and fast or slow decay.

More than ten years ago I parted out such an amplifier, keeping all valves and all signal transformers. Another partially working S-6 still sits in my flat, waiting to be scrapped.

I think that a VCA built with these trannys and tubes can also be controlled by some DC derived from the guitar signal in order to work as a compressor. Would it be better to get the DC control voltage somewhere from the stages before the compressor (i. e. forward control) or from behind it (backward control)? Compression rate would be a function of AC/DC ratio, i.e. of the AC gain before rectification, wouldn't it? What else do I have to consider?

Best regards!
It's complex stuff. With black magic overtones.

I see 2 options:

1) build it as-is .
We know it works, it's been optimized, etc.
Meant for organs, but you have 2 working attack/release options .

EDIT: forget it.
It's not a compressor at all, has no signal level sensing/rectification, it's a VCA triggered by switch contacts. :(

2) research the great compressors of yesteryear, think Cinema Audio from the 40s and 50s and build one using your tubes and iron.
I see it, but I don't quite believe it. Three octaves of individual oscillators on tubes? Plus all those extra effects, and the "solo"? It must have put out enough heat to cook dinner. I love it.

If you take the control voltage from after the gain cell (like the Fairchild) with a less than infinite attack speed (likely, particularly with vacuum state rectifiers) the signal will go straight through during the attack period, overcharge the time constant capacitor, and then drag the gain down too low, making a sharp transient at the beginning of each new note. Over used it makes a piano sound almost like a harpsichord. On bass, this can sound great (although the gain reduction in the unit I used was optical a balanced vari-µ pentode should be good, nay excellent), but it's spiky; doesn't control the dynamic of the instrument because the sharp attacks are still getting through, before being limited. Nevertheless, with tube stages following, and gentle, controlled clipping…

Whereas, if you generate the control voltage from the incoming signal it'll be a smoother compression, not as aggressive.

Couldn't you do two separate sidechains, switching before or after, as you have all those spare bottles? And put in those lovely timbre controls and switches? It's almost a pity to have it restricted to instrument use; I suspect I'd put a balanced line input and output on it so it could be used as a studio effect, rather than all these digital simulations;).
Pics don't exist on an accessible server. Where were you hosting them?

I did some work with voltage controlled filters with guitar, and individual jacks for synths and mixing boards etc. over the years For auto-wah or trigger-filter etc. for best results you have to sense very early before any source of tube compression, like right after the very first tube stage, and derive the control voltage there from a signal that has maximum dynamics. Then the controlled filter would be very very late in the signal chain, after any distortion made richer harmonics so the filter had maximum effect.

I would imagine that if you wanted a fixed compression ratio or some standard curve or knees in your derived control voltage, you would want to sense early after the very first stage, while there are still good dynamics to sense. Then again, if subsequent stages are altering volume a lot via tube distortion compression when running the tubes in their non-linear region you might have to sense later. More on this later...

For an expander you would ideally want the volume control rather late in the signal path, so that the quiet parts with no program to mask noise would really be quiet. Yet often the output is quiet when you clamp the input much earlier, so you can get away with sensing and VCA all at the same place in the circuit...getting off-track again.

For a compressor, I think it would depend on your intent. Where to sense, and where to apply control, early before any distortion or later applied to the distorted signal.

If you wanted clean compression, like for jazz, you would probably apply the volume control early too, so that you never generate distortion. I call that 'clean compression'...sense early, control early.

But if you wanted more expressive bluesy rock tone, you want the signal to have real powerful dynamics all the way into the stages that generate distorion, so that, like a human voice or a sax, it gets all nasty when played harder; then you want to further emphasize that timbre change rather than just the volume change by minimizing the volume change. I'd call that 'expressive compression'...sense dynamics early, apply volume control (or filter for that matter) late.

But there's another use for compression, when metal and hard rock bands want every hammer-on, pull-off, trill, tap, etc. to have the same envelope. Microsoft Word has 'spell-check' to correct your spelling, and a 5150+ has 'pick correct' to correct your picking and make sloppy technique that no country player could get away with sound really incredible. This is not really the same kid of expressive playing around with corssing the clean/dirty line. This is always pretty dirty, percussive in that the timbre only has much treble for the first milliseconds of attack and you can tell when a power chord is struck, and limits every note or chord to a maximum distorted volume. This kind of compression is what they really want when a guitarist asks for 'infinite sustain', they want a note to come on distorted with a noticeable treble pick attack, then continue to sing on distorted at the same volume until muted. For this, you overdrive like mad and both the sense to drive the control voltage and the placement of the volume control (or filter) are very late in the signal path. Sense late, control late.

Personally I would make the "sense" part of the circuit to derive the control voltage switchable from very early to late, and leave the controlled amp or filter extremely late.
Last edited:
Taking the control voltage for the VCA from a point after the VCA means you have a servo loop. That means that all the usual theory about loop stability comes into play. Ignoring this will create problems, just like it does for global NFB.

Two solutions:
1. don't use a loop - but then put up with the consequences
2. learn about loop stability and design a good loop
I've been wanting real good "sustain" compression in my electric guitar rig. After researching compressors on the web, I bought a Diamond brand compressor pedal, but the attack sounds inconsistent, and the sustain is not much, despite a bunch of really good reviews on the web. So I've researched the cr*p out of compression circuits and find it to be fairly complex.

Apparently the professionals will often cascade 2 compressors, the first having a very fast attack and decay, and the second a medium attack and slow decay. The goal here is to minimize the artifact of the recovery time, when the initial transient causes the typical compressor to temporarily over-react. Another approach is to compress various frequency bands independently. But I want to keep the circuit simple enough to fit it into a small pedal box...

Another approach that is arguably clandestine, is real-time compression, that has no attack or decay time, but increases harmonic distortion, which may or may not be a problem with a guitar rig.

For example, a class AB push-pull tube output stage circuit that has cathode biasing, and no cathode bypass caps... As the grid is driven positive, the more current the tube tries to let loose, the more the bias is increased in real time (since no cap across the cathode bias resistor), so it's half cycle gets real time compression without the attack and decay issues, only potential crossover distortion issues.

I recently built a guitar amp and did this, and it sounds great, but I haven't actually measured the sustain each way, so I don't know how much diff. it made, but the concept seems relatively good.

Another approach to increasing sustain and compression without attack and decay issues is what Mesa and Dumble did, where they cascaded tube stages such that one would just start to soft clip, when the other stage was a little further into soft clipping, thereby spreading out the amplitude range of the soft clipping. That appeared to create quite a bit of sustain in a Dumble I heard recently. Harmonic distortion wasn't obviously audible, but must have been there to some extent.

Any thoughts, or other ideas of how to do realtime compression without the attack and decay issues? That Diamond compressor I bought, it just ain't no good see...
Real time/instantaneous compression will by definition affect the waveshape and thus create distortion; what you seem to be needing is a very fast compressor, close but not the same.

It must not react to the individual wave instant value, but to the signal "envelope" , the line which joins the "mountain peaks".

A pedal is not up to the task, you need a Pro type compressor (probably rack mount) which has adjustable threshold, attack and decay controls for a perfect fit to what you play.

Because of home recording, some are available at quite affordable prices.
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.