Mosfet Compressor Please RE-respond (previously deleted by server change!)

[comfortably_numb]

So I'm posting this again.  I am building this circuit:



With these modifications per MartyMart:

100k input resistor before first 0.1uf cap
8k2 and 10uf between pins 1 and 8
10k pot and 100k vol pot "as is"
1M log pot replaces the mosfet 1M to ground, wired as variable resistor ( lug 3 -- lugs 1&2 joined  with lug 1 to ground )
4.7uf cap changed to 2.2uf

Here are my questions:

1) Can the diode be substituted for another?  What likely effect will this have?

2) How is this circuit working exactly?  I'm at a bit of a loss on compressors ins and outs.

3) Is MartyMart's 1M pot to ground actually a threshhold control as he is calling it, or is it something else?

4) Marty has said this circuit has "ripple."  Is he refering to the "pumping and breathing" caused by an improperly adjusted attack/release?
He also seems to relate this to the 4.7uF capacitor which he has changed to 2.2uF.  Will the attack/release speed up even more with a 1uF cap.  Lower?

Thanks,
CJD

[Toney]

I replied earlier.
Check the OZ compressor thread... we're working in it...

[Mark Hammer]

So I'm posting this again.  I am building this circuit:

With these modifications per MartyMart:

100k input resistor before first 0.1uf cap
8k2 and 10uf between pins 1 and 8
10k pot and 100k vol pot "as is"
1M log pot replaces the mosfet 1M to ground, wired as variable resistor ( lug 3 -- lugs 1&2 joined  with lug 1 to ground )
4.7uf cap changed to 2.2uf

Here are my questions:

1) Can the diode be substituted for another?  What likely effect will this have?
    The diode allows only one half of the audio waveform to pass.  Any diode will do this..The 1N4148 is a silicon diode, which will only allow signal to pass if it is in excess of around 550mv, so it essentially "deducts" that much from the signal.  Any diode that does the same will leave the circuit performing identically.  The usual replacement is a 1N914 but there are many others.  If you have a germanium diode, that will work too but will deduct less from the incoming signal and alter the performance of the circuit, possibly for the better, possibly not.
2) How is this circuit working exactly?  I'm at a bit of a loss on compressors ins and outs.
    This circuit is a very distant relative of the Orange Squeezer.  The OS uses a kind of electronic volume control made up of an 82k resistor in series, and a FET to ground. When the output signal is fed back through a diode to the gate of that FET, the FETs resistance is lowered and the resistor/FET combination acts like a volume pot that has been quickly turned down, and allowed to be turned up again gradually as the signal dies away.  This operates on exactly the same principle.  One thing I am unsure about is how that action will occur in any calculable way with no resistor beteen the 0.1uf cap and the drain of the BS170.
3) Is MartyMart's 1M pot to ground actually a threshhold control as he is calling it, or is it something else?
4) Marty has said this circuit has "ripple."  Is he refering to the "pumping and breathing" caused by an improperly adjusted attack/release?   He also seems to relate this to the 4.7uF capacitor which he has changed to 2.2uF.  Will the attack/release speed up even more with a 1uF cap.  Lower?
When the envelope-follower portion is crude, the DC signal it spits out on the other side of the diode is not smooth.  It tracks the momentary waveform peaks but does not do a good job of tracking the average over multiple waveform peaks.  Think of it like a mechanical (not LED-based) tachometer that still follows each motor revolution enough that the needle on the dial looks like it is "vibrating" as it moves higher and lower.  In the compressor context, this is referred to as "envelope ripple".  In theory, it is no different than a poorly regulated power supply that has not fully removed the 60hz hum from the AC line and turned it into stable, battery-like DC, except that the "AC line" here is your output signal.
Breathing refers to the manner in which the recovery of volume and the noise of the device mimic human breathing.  Typically, if there is a lot of gain, and a gradual recovery after initial transients (in this case the FET resistance is slowly increased again after momentarily going low), it can sound like someone exhaling.  The way in which this is reduced is by having the capacitor that "averages" all those waveform peaks drain very quickly.  It can do this by being smaller in value (less current/charge stored = faster drain) and/or by having a resistor in parallel to provide a path for it to drain.  In this case, it is the 1M resistor.
It is always a balancing act with regard to how much ripple you want to reduce, compared to how much breathing is a problem.  Going for a big capacitor reduces annoying ripple, but also increases how quickly the unit recovers to be ready for the next picked note or strummed chord.  The ideal can sometimes be located by making that 1M resistance variable.  I would suggest placing a 220k resistor in series with a 1M pot (wired as variable resistor - i.e., wiper and one lug) in place of the 1M fixed resistor.  This will allow you to adjust the discharge time of the cap and get the balance closer to perfect for yourself.

[comfortably_numb]

Why the 220k series resistor?  I had already decided on making the 1M a pot.

Is there another way to eliminate ripple without causing breathing?

[Mark Hammer]

1) You need to have at least *some* smoothing or else it (envelope signal controlling FET) will be all ripple and also severely attenuated.  The 220k sets a minimum decay time.  The 1M pot adds to that.  220k+1M might be more than the original, but that also may be what you want/need sometimes.

2) Ripple is most easily reduced when using a full-wave rectifier.  This unit and many similar ones, use a half-wave rectifier.  Again, to use the power-supply regulation model, if the "hum" were 120hz, and not 60hz, you would not need such a large capacitor value to smooth it.  The full-wave rectifier acts a bit like an octave fuzz, doubling the number of peaks (per unit of time) to be smoothed.  This requires a smaller cap value to do the job than with a half-wave rectified signal.  Smaller caps will introduce less breathing, the same way that faster discharge times will.  This design CAN be modified to use a full-ave rectifier, but at that point it starts to turn into another design.  May as well just make yourself a Ross/Dynacomp, since that already HAS a full-wave rectifier, and is easily modded for faster or slower recovery.  Most importantly, there is a PCB layout already posted and you won't have to "invent" anything yourself.