Discrete compressor. So simple.

[brett]

Hi
I've been working on a simple compressor that uses easy-to-get parts.
The schematic of a working prototype is below.

Q1 and Q2 are any common NPN transistors (I used 2N3904).  The MOSFET is a 2N7000, BS170 or similar.
It seems to work ok.  Two minor issues are:
Don't drive it too hard or it'll distort (in a nice way, though)
I think I'm hearing a little loss of treble. Maybe.

I haven't tested it much. I just plugged it in and it works.  If a coupla people would build it and give me some feedback, I'd really appreciate it.
cheers

[DDD]

Impressive.
I venture to recommend you to reduce the "initial" gain of the input stage to avoid distortion at high input level. The compression rate of course will be less, but they say "the best compressor is a bad compressor" from the point of view of preserving some dynamics.
Also it will make the frequency responce a little bit better.

[cb]

Brett, thanks for sharing your work in this area. I have always been intrigued by compressors with a low parts count. What does Q2 do?

[Peter Snowberg]

VERY COOL!

DDD summed it up perfectly....

Impressive. 

[brett]

Hi
thanks Peter.  That's much appreciated coming from a legend.

What does Q2 do?

Q2 is a funny kind of buffer.  It has high impedance so far as the gain stage is concerned, but very low output impedance.
There's also a "trick" to the use of Q2:  the trimpot controls the amount of both signal and DC output.  As these increase, so does compression because they turn the MOSFET "on".

For the option of "clean" compression, you could have a switch that disconnects the 100hm resistor and 22uF cap.  Then it would be clean except when driven by a huge signal.  Alternatively, the fixed 100 ohms could be made a 1kB pot, wired as a variable resistor.  At 0 ohms it would be high gain and dirty, and at 1k it would be low gain and clean.
cheers

[Peter Snowberg]

Gosh, thanks. Really.... I'm just another bozo on the bus.

We have a number of people here who could easily carry that title, but I don't feel worthy of it.

[Doug_H]

That looks really nice. Good job!

I've got an old hobby music electronics book at home from 1974 that has a diode compressor circuit in it that uses one transistor stage. The Zin is too low to be useful for guitar and there's too much loss in the R/C/diode network. So I subbed the transistor with a MOSFET and added a 2nd MOSFET stage for recovery and it sounded pretty good. Compared to my CS-2 it was "unrefined" but at the same time it did some fun extreme squeezing that the boss circuit won't do. It had the same issue with distortion with a hot input but not much you could do about it as you needed the gain to drive the diode network properly. It's a weird circuit in that it's not the typical sidechain/feedback kind of deal, but functions by slowly turning a diode on and off and shunting signal that way. I never actually built a copy, but it sounded pretty cool on the breadboard!

[Mark Hammer]

One of the tricks for making FET-based phasers less susceptible to distortion from excessive input levels is to run an RC network between drain and gate, as shown here: http://www5b.biglobe.ne.jp/~houshu/synth/PhaseFet0205.GIF

This approach can be seen on a wide variety of commercial phasers, especially Korg, but a number of others as well.  You might note that it is also used in the Orange Squeezer as well (http://www.generalguitargadgets.com/pdf/ggg_osq_sc.pdf). 

Some time earlier this year, Mike Irwin told me that while he used to be a fan of it, he had since switched his opinion.  The reason was that while the RC network raised the threshold where the FET started to distort, when it DID start to distort he found the distortion particularly ugly.  His approach is now that he would rather put up with some minimal distortion that increases linearly, and in a non-ugly way, with input signal level, than have a deceptively clean sound that crossed some threshold to ugliness with increase in signal level.  What was never specified in all of this was the actual signal level at which this occurs.  I draw attention to the fact that Mike works almost exclusively with synths and synth output levels (gotta get that lad a guitar one of these days!), so the "threshold-of-ugliness" might be much higher than most of us here would encounter.  Certainly I've never heard anyone complain about their Orange Squeezer suddenly transitioning from clean to distorted.  And if that wasn't enough, compressors are generally inserted very early in the signal chain, before much gain is added to the guitar signal, whereas phasers tend to be inserted after additional boosting, fuzzing, processing, etc.

So, it may be worth considering use of a similar RC network as found on the OS to reduce the distortion you report....or...it may not be.  At this development stage, it might be worth exploring the audible differences between the one and the other.

Either way, nice piece of work, and a nice example of the K.I.S.S. principle.

[Doug_H]

  The reason was that while the RC network raised the threshold where the FET started to distort, when it DID start to distort he found the distortion particularly ugly.

That's because it's a negative feedback loop and once you break the NFB by overloading the system, things go to hell in a hurry. I had a similar problem tuning the NPN gain stage of the Kay tremolo. I was able to solve it with careful biasing, low hfe, and some degenerative feedback in the emitter circuit (no bypass cap). It sounds really good now. MOSFETs are more difficult at 9v because they have such high gain. Also, as mentioned, you need some drive for the compression to work anyway. It's even more acute with the diode-compressor I mentioned, and is a good lesson on why they use side chains for this kind of stuff. So with the diode comp the little distortion moves from being a "bug" to a "feature"...   FWIW, I was under the impression that the Orange Squeezer distorted a little too, no?

[markm]

I have just completed a build of this comp and the actual Compression control does nothing.
Could be a problem with the layout or perhaps somthing overlooked by me or in the schem.
Here's the layout>>>
Anyone have any ideas?? Or is it possible that the amount of "squish" from the control is so minimal that I just can't hear it.
Also, the trimmer does adjust the compression amount but as I said, the actual B100K pot does zero.

[puzzle87]

Try this modification:

Pot "Compress" really able to work.

[markm]

Okay then.
I'll go back to the drawing board on this one!

[puzzle87]

Okay then.
I'll go back to the drawing board on this one!

...fixed gate and source position of Q4,
sorry

[markm]

Okay then.
I'll go back to the drawing board on this one!

...fixed gate and source position of Q4,
sorry

No big deal, that's an easy fix.
I sent you a PM about this.....let me know.

[markm]

Here's the revised version;

If anyone builds this and can confirm the layout, please post the results 

[brett]

Hi
Markm, did you get any compression? 
Your layout looks good to me.  No obvious "bugs". 
The voltage at the trimpot wiper on mine was about 1V DC (no signal) and about 0.3V DC at the MOSFET gate.  With input of a typical guitar signal, this should rise to almost 2V DC at the MOSFET gate.
cheers

[puzzle87]

This is last (i hope so) mods of NOT so simple compressor

Variant 1

------------------------
Variant 2 (my favourite)

------------------------
Both variants tested on breadboard and really good sounding. Variant 2 have lower noises.

[markm]

Hi
Markm, did you get any compression? 
Your layout looks good to me.  No obvious "bugs". 
The voltage at the trimpot wiper on mine was about 1V DC (no signal) and about 0.3V DC at the MOSFET gate.  With input of a typical guitar signal, this should rise to almost 2V DC at the MOSFET gate.
cheers

I can't help ya on this brett as I have not actually built it yet 

Puzzle87.....
The two variants are just Mods or is this what needs to be done to make the circuit function well?

[puzzle87]

MarkM,
   both variants are complete, full functionally devices. No more changes.

[markm]

MarkM,
   both variants are complete, full functionally devices. No more changes.

Great!
one question;
What would be a substitute for the 2sk170BL?