Guerrilla Compressor..super simple comp.

[dschwartz]

Hi all!!
It has been years since i have posted a circuit here, but i wanted to share this really simple, but effective compressor i have been playing around with.

I assume all of you have seen this simple old passive diode compressor:



it's quite crude , and it works by changing the clipping threshold of the diodes according to the envelope. This is quite dirty and my first attempts produced a very distorted result..it wasn't an ugly clipping, but a relatively nice warm "saturation" effect, but unwanted, anyway.

So i wanted to make an active version and tweak it to get a functional compressor working....

and here is the circuit i have designed so far:



It's so simple, no OTA/Vactrol/Jfets , just simple, ordinary 1n4148's..i tried leds and Schottky's, but they didn't work so well..
I started with a feed-forward configuration, which worked OK, but if you used too much ratio and gain, there was too much pumping (too much compression) and  if i changed from a single coil to a humbucker, the humbucker was too compressed...... so i changed at the last minute to a feedback config, and the pumping was gone, and it stays leveled with different pickup outputs.

Note the Ratio's 680pf bypass cap. This cap made a whole difference, since it lets the higher frequencies less compressed, and the mid and lows are compressed more. Without it, the tone was too dark and distorted. With the cap, the less compressed high end masks the distortion and results in a snappy tone, with more twang and bite, and the distortion actually adds body to the sound and it is not really discernible.

I used LM833 in the proto, since it has a bit more ooomph to drive the envelope, and less noise. there's a slight oscillation tone with the gain at full, i'll have to investigate and come with a fix for that.

well..any questions, improvements, and criticism are welcomed!!

[teemuk]

I think there is a fine line between whether you want this kind of circuit to compress by clipping and with rather ample amount of distortion or to compress by attenuating and with much lesser distortion.

In first case the design is rather straightforward: You just control diode reference voltages with an envelope follower of some sort and tweak control voltage so that signal peaks are clipped.

Second case is much more troublesome: You retain input signal's amplitude in the very narrow region where diode operates as a somewhat "linear" variable resistor. This is the narrow region between beginning of forward conduction and full "saturation". Adjusting diode reference voltage controls diode's resistance, which you then exploit in a resistive voltage divider - type attenuator. The difficult to do this without much distortion AND without excess noise is the reason why you rarely see diodes exploited in such application.

[Vivek]

Thanks for sharing, Daniel

I look forward to refinements on this circuit idea.

[Kipper4]

Thanks Daniel.
Added to breadboard list.

[dschwartz]

I think there is a fine line between whether you want this kind of circuit to compress by clipping and with rather ample amount of distortion or to compress by attenuating and with much lesser distortion.

In first case the design is rather straightforward: You just control diode reference voltages with an envelope follower of some sort and tweak control voltage so that signal peaks are clipped.

Second case is much more troublesome: You retain input signal's amplitude in the very narrow region where diode operates as a somewhat "linear" variable resistor. This is the narrow region between beginning of forward conduction and full "saturation". Adjusting diode reference voltage controls diode's resistance, which you then exploit in a resistive voltage divider - type attenuator. The difficult to do this without much distortion AND without excess noise is the reason why you rarely see diodes exploited in such application.

Thanks for your comment, Teemuk!

Actually, the interesting part of this "compressor", is that it does not work based on the dynamic resistance of the diodes..the dynamic resistance is negligible compared to the "ratio" resistors.
The principle of operation is a "dynamic threshold".. the envelope changes the threshold of when does the diode start to conduct. At high transients, the the threshold is lowered, and reduces gain at a fixed ratio given by the "ratio pot" and the 22k R3&R13, and holds it for the release time..

On the low level signals, the threshold is held high and the signal gets amplified by the output opamp, which is fed back to the envelope, mantaining a relatively constant level, i.e, compressing.

Of course, the compression knee is quite hard and creates distortion, bur the 680pf treble bleed cap "masks" the distortion and it sounds pretty good..

i'm working now on a way to avoid this "treble boost" effect when the ratio is set too high...but at 12'o clock, it sounds just GREAT.

[dschwartz]

OK OK.. i made a few refinements...

i think the Gain/comp may be better as a linear pot...also, adjusted the frequency response, now the response across the different ratios is relatively flat..it is very transparent now..

The attack is fixed at around 10ms, and the release goes from 200ms to about 1 second, and changed the 22uf for 10uf because i got higher envelope voltages with them..and more effective compression.

this is the revised schem:



and some ltspice simulations:

[Vivek]

What's your THD ?

[dschwartz]

What's your THD ?

I don't know..
It's not meant to be a studio comp...just quick and dirty, good for guitars..
Give it a try on the breadboard...

[dschwartz]

i recorded a little test..
i hope you can hear the compression...there's still some distortion going on, it's less noticeable in person, but you can hear it more on the video., specially with the bridge pickup (it's a hot humbucker)

[Nasse]

Someone should try this for solo vocal, something little dirt does not matter

[dschwartz]

Someone should try this for solo vocal, something little dirt does not matter

Yeah i kinda like the dirt, it's like an undertone growl..