Tell me what You think of my compressor design.

[Brisance]

Hello.

First of all an introduction. I am an Estonian beginner electronics hobbyist and have been trying(and loving) this hobby for the better part of last year. Mainly interested in DIY audio gear and among others, stompboxes! I have been facing a bad truth - While I have friends, I have no one to discuss electronics with, so long story short I made an account here and hope, you will welcome me. I have been trying to not do any clones and if, then heavily modified, I mainly desire to design stuff myself. And thus:

I have lately been thinking a lot about compressors, optical in particular. So today I designed the circuits bare bones, I only have simulated parts of this circuit briefly, not the whole thing yet. So I am looking for experienced peoples feedback on this, whether I should try to go on with it.

Note that the circuit is only a simplified signal path, no values calculated yet.
The circuit:


(click for full size)

I was thinking of powering it from +/- 12V, enough headroom and common, when the circuit works, I will build a dedicated PSU in the case.

Since the design uses 4 op amps, all dual supplied, I think I could use a nice quad one to save space.

My theory on this:

• The signal enters the first opamp, with negative feedback set by an LDR(which will be in a dark enclosure along with light sources), with also a manual gain pot, probably the same value as the LDR in complete darkness. there is also a bypass switch, which could be engaged, thus setting the circuit on unity gain. Now to think of it, I should bypass the output volume pot with it as well, so one can A/B nicely.
• The signal then splits off to an output buffer and to the gain reduction stage. One could add an extra sidechain input, switch the control switch and control the dynamicks with an external source.
• Then comes the threshold voltage divider, which will only attenuate the signal gain, if the divider's output is above diode forward voltage.
• Then another opamp comes, this time with variable gain, so one could set the compression amount.
• After that the signal gets rectified, full wave by a pair of shottky diodes.
• Next comes a parallel capacitor used with charge and discharge time pots, to control attack and decay(it worked in a simulation at least and that made me pretty happy since I came up with such a solution) The cap and pot values to be calculated and twaked yet.
• There will be a switch to switch between 2 mosfets driving either a LED for harsher characteristics(instanteneous) or a smoother filament bulb
• Since it is a boosting circuit, output levels are regulated down instead of using makeup gain.

I would be very thankful for constructive feedback! Keep in mind, i'm a beginner.

[PRR]

Welcome!

There is SO much missing (bias and blocking connections, values) that it is hard to comment.

Why are you re-inventing this wheel? There are many-many compressor plans published. And I think you could find some good points by studying them.

Your connection of LDR and "input gain" probably does not do what you think. When dark (silent) the gain goes toward infinity, which is more than you want (brings up the last bit of background noise). OTOH at maximum gain-reduction the gain becomes Unity, and it is sometimes necessary to reduce gain more than that.

You connection of Attack Release throws the transients into the lamp driver-- "attack time" is zero.

VR3 (you have *two* of these) and VR4 are connected so gain goes to infinity (but not below unity). This is more gain than you need.

A gain-pot between last amp and jack begs for that last amp to overload, no way to turn-down.

"Output to panel" probably needs to be buffered. This suggests another opamp.

The default Attack/Release bias is zero Volts, and the FET is biased to zero. No-signal, the FET and LED are full-ON, the LDR is low-gain. Probably not what you want. Then when signal pulls the A/R filter positive, FET Gate forward-conducts. LED does not get any brighter, but FET Gate is strained.

Not sure why you need two FETs. Perhaps the one could be switched LED/Lamp? But I think you have other things to deal with first.

Quad opamps can be awkward. Too many parts around too few pins. Different signals (audio and control) laying right next to each other. Duals are usually better layouts.

[Brisance]

Thanks for commenting!

Why are you re-inventing this wheel? There are many-many compressor plans published. And I think you could find some good points by studying them.

While that is true, I think I will learn most when I am doing exactly that, forces me to think about every issue and not just blindly follow a schematic. I could take a look though

Your connection of LDR and "input gain" probably does not do what you think. When dark (silent) the gain goes toward infinity, which is more than you want (brings up the last bit of background noise). OTOH at maximum gain-reduction the gain becomes Unity, and it is sometimes necessary to reduce gain more than that.

Now come to think of it, yes, you are right. Probably will have to use the LDR to short to ground(resistor of course added) and still use makeup gain on the last opamp.

You connection of Attack Release throws the transients into the lamp driver-- "attack time" is zero.

Hmm, that may be, although I simulated that part of the circuit on the online java simulator applet(falstead?) and there the lamps current worked as intended. Will have to put some thought into this.

"Output to panel" probably needs to be buffered. This suggests another opamp.

That's the one point I HAD thunk of "panel meter was supposed to be a quad comparator with leds indicating the level of compression

The default Attack/Release bias is zero Volts, and the FET is biased to zero. No-signal, the FET and LED are full-ON, the LDR is low-gain. Probably not what you want. Then when signal pulls the A/R filter positive, FET Gate forward-conducts. LED does not get any brighter, but FET Gate is strained.

Thank's for pointing that out, will give some thought.

Not sure why you need two FETs. Perhaps the one could be switched LED/Lamp? But I think you have other things to deal with first.

Well to be honest I initially thought I would have to bias each FET or NPN as in the first thought, the original pencil and paper schematic had a circle around them with "bias these!" next to it.

Quad opamps can be awkward. Too many parts around too few pins. Different signals (audio and control) laying right next to each other. Duals are usually better layouts.

Thanks, I always thought they would be more convenient since there are less rails to consider for a not-so-perfectly etched one sided PCB(altho I have only used 741s before since I have them in abundance thanks to an accidental misunderstanding at ordering parts, I ended up with a 100) .

I am really grateful for your input and I want to improve this circuit even more now, I will take it as a challenge. I deliberately did not add the values etc, because I wanted to know if my approach works at all(I am used to just drawing the circuit like that and then calculating, adding and breadboard tinkering stuff until right)

[midwayfair]

Agreed that it's hard to comment without values etc.

I have a couple other comments:

1. Your threshold and compression pots are doing almost identical tasks. The compression pot is better IMO because it is buffered from the audio path.

2. I think D1 and D2 are superfluous: Your envelope already will not trigger unless the signal amplitude exceeds the Fv of D3. I'm not even totally sure D1 and D2 will do anything of value in actual practice anyway, just based on past experimentation. Chances are you wouldn't notice a difference if they were removed.

3. I know lots of people find optical compressors nice, but really all they do is set a limit on how fast your attack can get. The LDR has a slow turn on time. The sluggishness acts as a longer attack. A properly implemented attack control would do everything the LDR does except a FET costs less.

4. Speaking of which -- your rectifier produces a negative voltage. Applied to the gate of the FETs, the resistance between the FETs and ground goes UP. That's useful on its own. Just connect the FET's gate to VR1 and it's your variable gain. You might want a capacitor in series with both (i.e., capacitor > VR1 > FET). I think if you swap the FET for an NPN transistor, you'll end up with something that drops the resistance to ground and makes the compressor work as desired.

5. Move your attack control so it's in series with D3. The attack calculation is Attack resistance || decay resistance * C1. Make it 20mS-70mS and you'll have the same attack of your typical photocell.

6. You will want a resistor in series with the release control, soemthing to keep it from going below 50mS (you could try to go lower, but you'll probably get noticeable ripple even at 50mS).

7. Unlike PRR, I don't think that dropping the signal below unity is strictly necessary, and there are plenty of guitar and bass compressors that never reduce the signal even that far. A guitar isn't going to overdrive an op amp at unity gain running on 24V.

8. You can avoid an extra device for the panel indicator by just putting the LED for the panel in series with the one for the optical. Assuming, of course, that you leave it as optical.

9. You will want a resistor in parallel with the LDR if you leave this optical to manage the maximum gain.

10. Check out the Flatline compressor for an optical design with the LDR in the effect loop and the ratio control as the gain of the first op amp -- this is similar to your first stage and will give you some idea of what works well.

11. Check out the Rothwell Lovesqueeze for something similar to the FET compression method I described in paragraph 4; and I did something similar with transistors for my "Bearhug" compressor.

Overall there are some interesting ideas here. I like the sidechain switch. I like that you left room for a make-up gain stage. I like that you included enough controls to get the most range out of the compressor. But obviously it's in a very nascent stage at the moment.

[Thecomedian]

The best reason I could think of to reinvent the wheel is a clear conscience when selling them as your own design 

[Brisance]

Thanks for all the feedback, here's an updated version, still a work in progress.