Univox compressor growl issue

[Mark Hammer]

I used to love my little grey Univox compressor before I had to sell it in 1980.  Not a particularly sophisticated or subtle effect, it could give seriously audible squish, and to my ears brought a single coil bridge pickup to life.  Back in the day, I also used it as my clean booster, and connected an EHX Hotfoot cable to the output volume pot on it, so it could also double as a volume pedal.

I ran across a vero layout on tagboardeffects ( http://tagboardeffects.blogspot.ca/2017/03/univox-uni-comp.html ) and a handdrawn schematic, shown below, which seems to align with the vero layout.  I find vero layouts annoying and confusing, so I used the schematic to perf it, using a nice bright red LED and a high dark-value LDR from Tayda (generally something over 4meg), and a BC184 for the transistor.  All diodes were silicon type, and I had to sub a 220n cap for the 330n shown, shaving off a tiny bit of bass.  Initially I tried out a TL071 as the op-amp, but when I heard the problem I'm about to describe, I switched to a nice old metal-can 741 I had in the parts drawer.  It seemed a little brighter, not discernibly noisier, but the problem was still there.

I fired it up last night and it sounds every bit as good as I remembered it.  So what is the problem?  Well, there is a low growl/modulation, whose frequency seems to track the note being played.  The fact that the modulation-rate changes with note pitch suggests it is not simply hum (I recently switched to all LED lighting to reduce such issues where I build and test)  And the slow recovery time of the LDR, aided and abetted by the 33uf between the emitter and collector of Q1, would suggest it is not necessarily envelope ripple.

So I'm kinda stumped.  Ideas, ladies and gents?

[digi2t]

The vero states a 33uF cap, but the drawing doesn't seem to state any value. Maybe 33uF is too big?

[Mark Hammer]

Thanks for the suggestion, Dino.  Personally, I think the 33uf value is part of what I like about it - long gain recovery time.  Looking at it with a magnifying glass, I think I might have installed the rectifier diode backwards.  Hard to tell, because things are a little cramped, but I'll see if flipping it around is a) possible, and b) solves the problem.

[R.G.]

You're pretty well versed in the usual suspects, Mark. It still might be one of those.  My only other guesses from looking at the schematic is that it might be to do with the rectification's loading on the output stage of the 741, or a power supply interaction.

The folks at Univox were clearly trying to keep costs down by not buffering the signal into the rectifier side chain. The opamp is loaded by two resistor->diode pair networks, one being the sidechain rectifier and the other a clipping network on the output. 741s are not great with heavy loading, and the 1.8K output clippers plus the sidechain rectifiers qualify. Maybe one half of the output stage is getting funny and causing the growl, or tipping over the current limit.

I would scope the power supply for signal related wiggles, and try a temporary buffer to drive the sidechain.

It's not much, but it is a theory. 

Saw your post: If the diode at the base of the rectifier sidechain transistor was backwards, that could do it.

[Mark Hammer]

Can now confirm the diode is installed correctly. No idea if it matters or not, but the BC184 I used has an hfe of around 450.  Of course the schematic only specifies minimum hfe. 

I forgot to mention that I replaced the 2k output pot with a 5k unit, although I honestly can't see that having any relevant impact on what can often sound like a ring modulator in the background.

Bear in mind that all of this rests on the assumption that: a) the schematic is traced and drawn accurately, and b) my recollection of how it sounded is accurate.

The scope is currently buried under a pile of stuff.  I'll clean up a bit, so I can get at it, and report back later today.

[Mark Hammer]

Incidentally, and nothing at all to do with the roblem, I filed down the "dome" of the LED, and buffed the now-flat surface smooth, so that the LDR could rest face to face with the LED, and put a piece of black shrink-tubing around it.

[Mark Hammer]

Here's a sample to illustrate what I'm talking about.  You can hear the undertone essentially following the pitch of the note being played.  I added a bit of reverb since that seems to highlight it a bit better.

https://soundcloud.com/mhammer1/compressor

[Johan]

It's probably the ldr attack time.
You probably have a half wave rectified guitar signal riding on a slower rising/falling dc, controlling the led..see if you can implement at tiny bit of attack time into the circuit. Perhaps a small cap at the transistor Base to ground?

Or try a vtl5c2, they work well for half wave rectified opto compressors
j

[Mark Hammer]

Thanks.  I was thinking that maybe some sort of modifcation of the envelope might be appropriate.  I'm also wondering if the use of a silicon diode in that position might leave too much AC in the envelope.

[Tony Forestiere]

I thought the effect was rather cool. It is not a defect, but an "Added Feature". 

[Mark Hammer]

Maybe, but it's a feature I would like to subtract.

I'm going to try a few diode manipulations later today and see if that improves matters.

[reddesert]

The tagboardeffects post and thread says that although the part number is for a silicon diode, people tested the actual parts and the voltage drop suggested germanium diodes.

[Mark Hammer]

Replaced the silicon diode with a Schottky (BAT46) and that did the trick. Clean, audibly ripple-free compression.  Im in love with this little baby all over again.  Thanks to all for resurrecting one of the first pedals I ever bought, 40 years ago.  Double and triple stops on the bridge pickup sound magical, like a pedal-steel.  Mark is one happy boy. 

[reddesert]

There are three schematics for the Uni-Comp on the web and they're all a little different.  There's the hand drawn one that Mark posted from davidmorrin.com, a traced one on FSB, and one at effectsdatabase.com, http://files.effectsdatabase.com/docs/files/melos_minicomp_schematics.jpg.

The hand-drawn schematic above has a 3.3K resistor and a 1K resistor in parallel from +9V to the anode of the vactrol/LED.  That doesn't make much sense.  The effectsdatabase schematic has a 2.2K resistor and a 10uf cap in parallel there. So I think David Morrin probably misdrew the 10uf cap as a 1K resistor. (The FSB schematic is also missing a 33uf cap, but it's in the photos).

I'll probably try to build it with the 10uf cap and see what happens. I'm not expert in the workings of compressor sidechains, but it seems like that cap would increase recovery time and decrease power supply interaction.

[rankot]

Interesting, I'll try to breadboard it with GL5539 photoresistor.

[Mark Hammer]

There are three schematics for the Uni-Comp on the web and they're all a little different.  There's the hand drawn one that Mark posted from davidmorrin.com, a traced one on FSB, and one at effectsdatabase.com, http://files.effectsdatabase.com/docs/files/melos_minicomp_schematics.jpg.

The hand-drawn schematic above has a 3.3K resistor and a 1K resistor in parallel from +9V to the anode of the vactrol/LED.  That doesn't make much sense.  The effectsdatabase schematic has a 2.2K resistor and a 10uf cap in parallel there. So I think David Morrin probably misdrew the 10uf cap as a 1K resistor. (The FSB schematic is also missing a 33uf cap, but it's in the photos).

I'll probably try to build it with the 10uf cap and see what happens. I'm not expert in the workings of compressor sidechains, but it seems like that cap would increase recovery time and decrease power supply interaction.

Whn I calculated what the 3k3/1k parallel combination would work out to, I realized it worked out to a rather uncommon resistance value   In my mind, that justified this unusual pair.  There are no Youtube videos to compare to, but fortunately I used to own one, and my build from the posted Davidmorrin schematic nails it.  So I'm not questioning its accuracy, simply based on results.  That doesn't make what you have inferred from theory "wrong".  I'm just saying that what I made works.

The thing to understand about the Uni-Comp is that it is an effect.  It is not transparent in the least.  You can certainly make it more transparent by turning the compression way down.  And certainly, the use of photocell-based gain-reduction helps in that regard.  But there is a pleasing "pump" to it, when the compression is cranked, that allows a Tele bridge pickup to sound like a pedal steel.

As for LDR properties, it is certainly more convenient to use an LDR whose dark resistance is 4-5meg or more, and leave all other audio-path component values as shown.  But if the LDR you have has a lower dark resistance value, simply recalculate the other bandwidth-determining values so that you can have relatively flat bandwidth for guitar, and a gain range of around 2x-150x.

I'm glad I stumbled onto this circuit. I missed it.

[PRR]

> calculated what the 3k3/1k parallel combination would work out to, I realized it worked out to a rather uncommon resistance value

It is unlikely to be a Magic Value. It sets the Max LED current. Most of the time the LED current is lower, governed by signal level. Also if we did want to limit LED _light_, or LDR resistance drop, we'd want to know the LED+LDR response. I think it just keeps the LED from burning-up, or throwing huge thumps. I think 500r or 1K would be fine. With newer LEDs, a higher value would do about what the low value did in 1979.

[Mark Hammer]

I'm inclined to agree, particularly given that LEDs available now are far more efficient than those available in 1977, when I bought mine.  Photocells may be smaller, now,  but unlike LEDs have not really changed.  So yeah, 1k is probably quite sufficient to limit the current needed to light the LED up.  Indeed, 3k3 might even be enough if you're using a superbright.

[snk]

Hello
I recently got advised to give the Univox Mini-Comp a try, and attempted to draw a layout (taking this one from Gfxlayout, and making it slightly  more compact).
I cannot guarantee that it is working (because i haven't built it yet, and the original is still unverified).

From what i have read, i can replace the opamp with a TL071 or a 5534, the transistor by a 2N5088, and use germanium diodes (Schotty Bat46), right ?
Has the layout above been verified (I mean : it is the layout with two 10µF caps, not the one with a 33µF cap) ?

If I wanted to play with release time, which value should I try to adjust ?

[Mark Hammer]

I built mine from that layout.  I had to adapt a few things because of the size of some caps, and cutting the board too small.  But apart from that, it worked.