Simple compressor found on youtube

[11-90-an]

Was seatching for diy octave down sound demos and found this guy who made a really simple compressor... no I/O caps, though, but I guess that was part of the "simplest" concept...

He put some build docs In his description...

I can't wrap my mind around it, though, as most of the circuitry and jacks use ... virtual ground?

It is quite interesting though...

[blackieNYC]

Seems like the attack would be instantaneous. Just needs a little makeup gain stage. Very cool.

[jfrabat]

Interesting...  Needs a couple more components, but it certainly IS simple!  And it works!  (would not trade it for my Engineer's Thumb, but it does work!)

[Rob Strand]

I can't wrap my mind around it, though, as most of the circuitry and jacks use ... virtual ground?

Just think of the ground connection between the input and output being ground.
After that the supply is non-grounded.

To me it's like balancing a broom on the end of the handle.   You need to finely and manually adjust the DC bias on the MOSFET to put it in a workable zone.     The sharp MOSFET characteristic provides the knee.    It's kind of half-wave rectifiying but kind of worse because on one signal polarity it turns off the MOSFET more.

The long time constant of the lamp will slow things down and hides a lot of distortion.

[11-90-an]

I can’t wrap my mind around it, though, as most of the circuitry and jacks use ... virtual ground?

Just think of the ground connection between the input and output being ground.
After that the supply is non-grounded.

So the ground are connected via the 10k resistor?

I redrew the schem a bit to help me understand...



Did I reverse anything accidentally? 

I’m thinking there’s a chance that the MOSFET can be replaced with a BJT, and we could add some I/O caps... maybe a resistor from MOSFET to bulb and a cap from the resistor to ground for a attack/release/decay/whatever-it’s-called control? Probably also a simple LPB-1 boost after?

I think I’m getting overexcited...

[Rob Strand]

So the ground are connected via the 10k resistor?

No, the grounds connect together, then the input and output (the center pins) connect together via the 10k.   The output LDR forms a divider with the 10k and reduces the output level.

[11-90-an]

So the ground are connected via the 10k resistor?

No, the grounds connect together, then the input and output (the center pins) connect together via the 10k.   The output LDR forms a divider with the 10k and reduces the output level

So I've been drawing schems wrong the entire time...

Wait... the  how does the MOSFET get triggered now since it's gate is basically tied to ground?

[Rob Strand]

Wait... the  how does the MOSFET get triggered now since it's gate is basically tied to ground?

The audio input drives the source!!!  and the gate is grounded. 

The current through the lamp passes through the battery and MOSFET but that lamp drive circuit and audio circuit only touch at one point, the source.   It's like two separate circuits.   

The pot biases the MOSFET right at the point where the MOSFET is turning on so a small amount of audio signal pushes past that point and the lamp turns on.   So the audio signal connects to the MOSFET at two points the gate and the source.

Yeah its pretty weird but you can have things like that when you have floating circuits (here the PSU)  and high-impedances (here the gate and source).      If you don't want to go crazy you have to separate the parts of the circuit in your mind.

[11-90-an]

So I'm guessing this pedal wouldn't work with a daisy-chained PSU? Since I/O jack ground will be connected to PSU ground because of possible other pedals?

[Rob Strand]

So I'm guessing this pedal wouldn't work with a daisy-chained PSU?

Not as is.

However, if you add some caps to give DC isolation it's possible to juggle things around so the different parts of the circuit have the same ground.   Roughly like this,

100k pot drawn as 2x50k's:

[11-90-an]

So I'm guessing this pedal wouldn't work with a daisy-chained PSU?

Not as is.

However, if you add some caps to give DC isolation it's possible to juggle things around so the different parts of the circuit have the same ground.   Roughly like this,

100k pot drawn as 2x50k's:

Now that you drew it, it *finally* makes sense... thanks, Rob...

I'll try to find a way to pull this off with a BJT...

[Rob Strand]

Now that you drew it, it *finally* makes sense... thanks, Rob...

I'll try to find a way to pull this off with a BJT...

With the BJT you lose the high impedance into the gate.   Since you already have power you could add a buffer but the on/off threshold probably sharper with a BJT compared to a MOSFET.    See how you go.   The MOSFET version is "rough and ready" anyway.

[11-90-an]

Worked on it the whole day, and here's what I came up with:



It's a nifty little compressor (or should I say, limiter) and squishes nicely...

Transistors can be any generic type, I used BC547s

So I'm just gonna break it down a bit... but I'm afraid I can't explain much at all... 

1. Input buffer - Rob suggested that there should be a buffer up front, so I did that. First I started with a Sziklai pair, but then it used 2 trannies, and it wouldn't be that simple anymore, right? So I chucked out the PNP and it works...

2. Crude Envelope follower (More like signal bleeper, but oh well...) - 100k trimpot biases the transistor that the LED would be slightly on, and with one big strum, should make the LED light up to it's brightest, (limited by the CLR, of course) then go back down to a slightly on state... honestly I can't explain, but the watching the video from the start of the thread should clear some stuff up...

I used a blue LED, but it's only because it was the nearest LED to my hand at the time. Other colors will also have different respones.
I also used 10k as the CLR value, but that was simply because 10k seemed to be the safest value of that LED... of course, different LEDs need different CLRs..

3. Boost - Basically the beavis colorboost with a higher input resistor because of some fizz I noticed...



From the output of that 1u cap, the 10k controls the amount of compression (or depth of compression). I used 2 LDRs because I wanted a quicker and sharper response from it...
And the 100k pot is a volume pot since the boost can give lots of gain. I think around 10 o'clock would be unity signal...

Overall, I'm quite happy with this circuit, simple and works well, of course doesn't compare to a dynacomp or engineer's thumb, I recommend you break out the great and finicky breadboard that you have sitting in the corner and try it out. I wanna see if it works out for others too... also, please, suggestions and corrections are appreciated a lot.. It's still on breadboard, after all...

p.s.1. When sustaining, say, a chord or low note, that was strummed hard, there is a tendency of envelope ripple. Placing the 1u cap from LED cathode to ground helped a bit, but putting a traditional RC network before the transistor doesn't really work, it upsets the balance of the envelope follower I guess...

p.s.2. I don't recommend rolling a vactrol, but placing them right beside each other with the LDR facing the LED it better, without rolling, so you can see the LED blinkity blinking, as looking at the LED makes tweaking the trimpot much easier.

p.s.3 There seems to still be some fizz in the decay of hard-strummed notes... anybody with an idea to remove this?

[Rob Strand]

I haven't gone through all the details of your post yet - have to go do something.

Worked on it the whole day, and here's what I came up with:

Very keen!

There seems to still be some fizz in the decay of hard-strummed notes... anybody with an idea to remove this?

It might be the transistor driving the LEDs.   When the signal is positive a large current flows into the base because the BE junction to act like clipping diode.  You might be able to put a resistor in series with the 1uF cap on the base, try 2k2.    A smaller cap might also help but the resistor is better.

[11-90-an]

Made some revisions... now it's in version 1.1



- Corrected some wrongly written values, 5.1M -> 5.6M
- Component value changes, 470k ->220k, 10k pot -> 5k pot
- Added 2.2k resistor in series with 1u cap to reduce fizz from envfollower (Thanks Rob, fixed the problem entirely... )

I still need a name for it though, any suggestions?
I think I'll just call it Optical Limiter or "OptoLim"...

[11-90-an]

I made a perf layout, could you guys spot any mistakes/errors? Any perfing rules broken? 
I'm basing it off the v1.1 schem ^^

[snk]

Nice
Thank you for the layout, and the schematic!
Any plan to release a veroboard layout as well at some point?

[11-90-an]

Nice
Thank you for the layout, and the schematic!
Any plan to release a veroboard layout as well at some point?

Thanks! The layout isn't verified yet, when I get my trimpots I'll try it. never worked with vero before, but I'll give it  try... 

[deadastronaut]

nifty lil comp/lim.....why 2 ldrs?

[11-90-an]

nifty lil comp/lim.....why 2 ldrs?

I also wanted a steeper and faster volume deduction, since they're wired in parallel, the volume will go down x2 as fast than with only 1 LDR

Also, forgot to add. Don't use ready made vactrols as they're probably too fast...
The best LED's to use are the slow ones. Less ripple.