Univox compressor growl issue

[snk]

- I don't have any BAT46 : could I try BAT48 instead ? Or should i use another Germanium diode (like 1n34a or 1n60) ? Is it advised to have the same diodes for all the 3 diodes in the circuit ?
- For the 330nF cap, I don't have any polarized or Tantalum cap : would a regular green Mylar cap do the trick fine ?

[Mark Hammer]

Near as I can tell, any Schottky will do.  It's the lower forward voltage that's important here.

[duck_arse]

- For the 330nF cap, I don't have any polarized or Tantalum cap : would a regular green Mylar cap do the trick fine ?

always. better than fine - can't be backwards. yes.

[edit :] under sage advisement.

[Mark Hammer]

Lest there be any misunderstanding, I think what ducky means to say is that a non-polarized mylar cap has the virtue of being immune to accidentally being installed "backwards", because it has no particular direction.  One might accidentally read his post as "DON'T install it backwards!", and then wonder how to identify the "right direction".  At least I did, initially.

[snk]

No, I get the polarity thing
My question was mainly about the sound (sometimes, a ceramic is to avoid, sometimes it is a must-have...).

[rankot]

I had some spare time last night, so if anyone's willing to build this one, there's a PDF with PCB design to fit 1590A. 22p capacitor is there in case you want to use NE5534 instead of 741.
https://docdro.id/qJA93Wh



Untested, because the guy who produces PCBs for me is currently refurbishing his workshop.

[snk]

there's a PDF with PCB design to fit 1590A.

Thank you, Rankot !

22p capacitor is there in case you want to use NE5534 instead of 741

What do you mean ? the 5534 is not a dropin replacement to a 741 ? Where would you put this cap (and is it mandatory) ?

[rankot]

In this case, you can use both. Put a socket and try. NE5534 shall have less noise, but it needs that cap installed. You can also try any other single opamp, but check the datasheets for compensation caps - some of them need different pins connected to the cap.

For NE5534, there's a 22p cap at the right of IC, connected to pins 5 and 8. Simply omit it if using LM741.

[snk]

So, i tried to build on veroboard the layout i draw and posted above, and (without much surprise) it didn't work... It's not a big surprise, as it is an unverified layout made from another unverified layout, so chances to fail are higher than usual 
I used the layout with 2 x 10µF caps (not the one with a 33µF cap), with a NE5534 (without any 22p cap) and a VTL53C, and didn't have time to investigate & double-check everything yet. I willl take some time to troubleshoot my build, and if it fails, I might try to build the other layout (with a 33µF cap).

[tubegeek]

The "hand drawn schematic shown below" has become a victim of Internet Aging. Do you still have it, Mark? Or anybody? Thanks!

[tubegeek]

Here's the schematic, I think?

[Mark Hammer]

Yes, as tended to be the case with many Univox pedals, they came out under a variety of brand names, Melos among them.

While we have the schematic here, the question will come up, "Why the recommendation to use a Schottky diode for D1?".  I'll explain.

There are many ways to transform an AC audio signal into a DC-like control signal.  It needs to be DC to provide "control" because AC/audio will get larger on both sides when there is "more" of it.  So, to be a control signal, there needs to be a "minimum", against which increases can be meaningful.

D1 "dumps" the unwanted half-cycle to ground, such that only the other half-cycle is turning Q1 on and conducting through the LED.  Is ALL of the unwanted half-cycle dumped to ground?  Depends on the diode type used.  If it's a silicon diode, then as much as 650mv of the unwanted half-cycle is still feeding Q1's base, because D1 won't dump that content until it reaches the 650mv forward voltage.  I opted for using a Schottky diode, because the forward voltage can be as low as 180mv or so, removing more of the unwanted half-cycle.  If one wishes to assure zero unwanted half-cycle, you'd stick another Schottky in series with the 5k6 fixed resistor (arrow pointing towards the Q1 base).

Look at a range of autowahs that also use a simple half-wave envelope extractor, and you'll frequently see that sort of two-diode arrangement: one in series and one to ground.

When I used a silicon device, I found there was too much unwanted half-cycle retained, producing an annoying ripple that was almost a hum.  Changing to Schottky, cured that.

Kudos to either Paul/PRR or Rob Strand, I forget which of the two, who noted a while back that the back-to-back pair of diodes at the output has been a common strategy for older broadcast-studio limiter/compressors.  The LDR takes a few milliseconds to spring into action.  The diode pair take care of initial peaks before the LDR starts to do its thing.  Because that diode-to-LDR changeover period is so short, you don't hear it as clipping.  In a sense, it takes advantage of the fact that the most harmonic content in a plucked string is in those first few milliseconds.  So 2msec more of what's probably not a significant increase in harmonics (produced by the diodes) is unnoticeable.  Clever.

Incidentally, if you look at the Dynacomp and Ross designs, you'll see a similar use of a single diode to "dump" unwanted half-signal.  In those instances, however, they opted for combining two complementary "chopped" half-cycles for a kind of full-wave rectifier.  I've ever tried it, but suspect that if one swapped the stock silicon diodes for Schottky, the result would be cleaner-sounding, because less of the unwanted AC would find its way into the control signal feeding the OTA.  OTOH, a diode swap may not matter, given the 10uf averaging/storage cap that I'm sure does some smoothing.

[rankot]

Here's the schematic, I think?

Yes, I used that one for my PCB design.

[tubegeek]

For completeness sake, here's David Morrin's hand-drawn schematic:


It's a really small world: David Morrin works at the recording studios of the Clive Davis School (media arts, part of NYU.) My nephew Clarke is currently supervising the renovation of a big office building about a mile from my house, in downtown Brooklyn. This building will be the Clive Davis School's new home when it's completed, and Clarke's been telling me about the bitchin' studio spaces that are built out in there. I have a feeling I may get to meet Mr. Morrin before too long....

Meanwhile:

I just love simplistic circuits - I now have this almost built already, with a metal-can 741 and a TO-18 transistor, 2N4386, hFe 430, "low noise." It looks cool next to the 741. I'm using one of those small Radio Shack experimenter perfboards with the little 3-pad busses everywhere. The board is half-empty, this thing has such a low parts-count.

This kind:



Should be able to get a test run going in the next couple of days!

I just need to place a couple of jumpers at this point and then it's wired up.

[tubegeek]

OK, it works, but with a few issues yet. I have a slow (subsonic) oscillation wiggling the sidechain, so it clamps down on signal nicely but it also clamps down periodically all by itself. Nothing's in a box yet, maybe that might get better when I put it into something.
 
It DEFINITELY compresses. It does this thing at max compression setting where it ends each note with almost like a volume swell, as it makes a ramped return to full gain.

I used a 5mm red LED (ordinary, not superbright) and a photocell that goes above my meter's 2M range/dark and down into the 1k-ish range/light. I think I got those cells from Tayda and they all matched each other pretty well, which was interesting. A metal can 741C and the 2N4386 in its little metal can I mentioned before are the actives.

The oscillation was different but still present when I first got it to work. Since then I added a chunk of wire to beef up the ground bus and I raised the main power supply cap from 100uF to 1000uf. I think these changes made the oscillation slower and deeper, and introduced a little POP on some attacks, seemingly ones that are catching the wave on the right part of it maybe?

I'll try and drill some holes in a project box later if I get time. If I am to get more serious about tweaking this thing all the way, I would work on the 1M compression pot and its 6K8 series resistor to get a different behavior - it goes from not doing much to cartoonishly crazy in about the last 1/8th of its rotation. I think I have some reverse log 1M pots, maybe that'd be the ticket.

Amazing what such a small handful of parts can do, though! It's definitely a "character" compressor for sure.

[Mark Hammer]

Like I said, it's an "effect".  Clean, as most optical compressors are, but NOT transparent.

[tubegeek]

Like I said, it's an "effect".  Clean, as most optical compressors are, but NOT transparent.

Yeah, your description was on the money, which is good, because it's why I built it!

Any guesses about the POPs on the attacks?

Also: you kinda sucked me in with the "magic telecaster" stuff. Guess what cheap Chinese guitar kit I've just finished building and am in the process of tweaking the setup?

[Mark Hammer]

The diode pair on the output is intended to "take care of things" until the LDR kicks in.  What sort of diodes are you using for them?

(I'm sure that audible pops can be produced in some fashion without there being uncontrolled peaks at the output.  I trust your troubleshooting instincts.  I'm just trying to cross off no-brainer things from the list.)

As with any optical compressor, the brightness of the LED, and the responsiveness of the LDR to sudden intense flashes, can yield some deviations from what is expected.  We may be building it in 2020, but the drawing stems from the '70s, when garden-variety LEDs kind of topped out at 600mcd.  Cheap LEDs these days can be rated at 10,000-20,000mcd.  So maybe the "cure" lies in using a 10k-25k trimmer in place of the 2k2 resistor, to dial back that first bolt of LED lightning a bit.  Or use 2k2 and go for an old school low luminance LED.

[tubegeek]

Diodes: a pair of 1N4148 that have about 700 mV forward drop (I measured.) The rectifier diode is a Schottky with about 230 mV forward drop. Seemed to conform to what you described up above.

LED: I suppose I could dig through enough stuff to find an ancient one. Is there a way to measure and select a dimmer one that would be more systematic than just comparing by eye?

I suppose a slower photocell could be another idea....

I'm wondering if it's a layout issue: that if I give the LED a path to the power filter cap that doesn't go past the op amp maybe it won't kick the op amp's power rail when it turns on?

Or maybe if I put another filter cap right by the 2K2/10u connection to 9V, and from there to ground? But the LED seems pretty well isolated already by the two caps above & below it.

I do have - before anybody asks - a .1 ceramic within a mm of the op amp's + and - rail connections.

My strongest hunch is that the slow, slow "motorboat" (like 1 hZ) is just too disruptive and needs to be addressed first. But I'm not sure how.

[PRR]

A new/hot LED will get dimmer if you use a larger series resistor. Even 10X larger.