THE ENGINEER'S THUMB... At last, a better compressor!

[merlinb]

Last night I had an idea for a different way to add a compression indicator LED. The 4.5V ref only has to supply significant current when the circuit is compressing, so you can just add the LED to the output of the Vref buffer (note that it is inside the feedback loop). I suggest using a super-bright LED since the current variation is between 0 and 2mA at most. What's nice about this approach is that it doesn't use any extra battery power.

EDIT: It worked best for me with a blue or white LED since they were both superbrights. These were so efficient they were dimly lit even with no signal, but if you want to tame this then try adding a resistor in parallel with the LED and adjust to taste (a few k-ohms probably). Most standard LEDs were too dim when I tried it, although one red one was 'OK'.

[deadastronaut]

cheers merlin, i'll give this a go then......report back later. 

[Rob Strand]

Last night I had an idea for a different way to add a compression indicator LED.

I like the idea.  You might get some glitching on the 4.5V rail.  I thought PRR's solution was fine also and perhaps less susceptible to glitching the 4.5V rail.

I think it's going to have the same problems as PRR's solution as far as the LED dimming is concerned; see the comments around the PRR's solution via deadastronaut's link.   

I believe the solution is to put a resistor across the LED, say around 2k2 to 4k7 for yours and maybe lower for PRR's; you will need to experiment.   The idea is when the LED current reaches a minimum threshold it goes out.   Suppose we set the point where it LED extinguished at 0.5mA then the resistor needs to be R = VLED/I_threshold = 1.7V/0.5mA = 3.4k.  When the current drops below the threshold the resistor hogs the current and there's not enough voltage to light the LED.

I've actually used this method for simple battery chargers.

The problem is it's a *compressor* so you have to decide at what point you consider it not-compressing.   If the LED circuit is moved earlier on in the the ckt then you can trigger the LED when the rectifier pumps significant charge into the storage cap.   When the input signal drops there's a natural cut-off, in fact in this case you might need to extend the LED drive time.

[merlinb]

I think it's going to have the same problems as PRR's solution as far as the LED dimming is concerned; see the comments around the PRR's solution via deadastronaut's link. 

There was no problem with dimming. That LED lights according to the gain reduction, and the gain reduction releases slowly (or as fast as you choose to set it to) -that's how compressors work.
If you only want an LED to light when you're actually making a noise then you're talking about a signal presence indicator (e.g. sound2light), not a gain reduction indicator.

he problem is it's a *compressor* so you have to decide at what point you consider it not-compressing.

The comrpessor already does that for you. When the signal is below about 30mVpk it is below threshold, so there is no gain reduction (after the release time) and the LED will go out by itself.
But you can, of course, decide not to let the LED cover the full range of gain reduction, that's up to you. It depends if you just want it to 'look cool' rather than 'show what the compressor is actually doing'.

Now, I suppose I should remind everyone that this is a feedforward compressor, so the side chain is always working at full bore. But the ratio control works like a blend control, mixing compressed with uncompressed. That means even if you have the ratio control set to 'no compression' the side chain is still doing gain reduction as usual, so the LED will still light even though the output signal is purely clean. The LED only tells you what the side-chain / OTA is doing, it can't second-guess what you've set the ratio control to....
...unless you use a dual gang ratio pot and use the other gang to control the LED somehow. Or if you build the ET with the threshold control instead of a ratio control, but that incurs a noise penalty since it now must run at full ratio all the time.

[Rob Strand]

But you can, of course, decide not to let the LED cover the full range of gain reduction, that's up to you. It
depends if you just want it to 'look cool' rather than 'show what the compressor is actually doing'.

For most compressors and limiters it's more conventional to blip the LED when the threshold is exceeded.  If you are in a long section of high input the blips and re-triggers are packed in close enough that the LED is on all the time. 

I don't have a problem with either method.  It's only a problem if you expect/want one thing but it does the other.

[tristanc]

On Merlinb's schematic there are three 1uf electrolytic capacitors, do these have to be electrolytic caps?

Tantalum, plastic/poly, or ceramic caps are preferred. Use electrolytic as a last resort.

What's the impact if I've used electrolytics for the two 1uFs to ground next to the Attack control? Anything major?

[Elijah-Baley]

For a value like 1uF, if you want, try polyester cap. Pay attention to the size.

I just boxed my Enginee's Thumb. But I'm not so happy about it. I built it twice, and both my boards (Sabrotone Layout at 5 knobs, but I included the LEDs) work weird. After the release time the volume rise up louder, and the first attack of the note, indeed, is really loud before the compression. Then, when the compression is high I get more volume from soft touch than the hard picking of the string. Very unnatural playing.
I used a resitor as Release control at 512k. And I could use neither the Threshold pot, but I used two resistors, but not exactly as the pot was in a certain position, because the two resistors are kind asymmetrical, the sum isn't 1M. I dont' know, I tried to work it well I could. (I have the treble boost and the attack pot, even if I prefer it maxed.)
All this volume issue cause a pop when I engage and disengage the pedal. After my mods is a bit softer, but, after the release time, when the compression is not acting I can hear the click of the footswitch anyway.
Further I found the 1M pot for the Ratio excessive, because I get anough comression at half, and at higher setting I still have the volume problem when I play soft and hard.
I tried two different LM13700 bought from two different shop, too.

I'm very surprised that nobody have this same issue!

[jonny.reckless]

What's the impact if I've used electrolytics for the two 1uFs to ground next to the Attack control? Anything major?

Technically, electrolytics have a higher series resistance and a lower self resonant frequency than other classes of capacitor. You have to form a subjective opinion about what that does to the tone and response. I tried electrolytic, multilayer ceramic and tantalum capacitors in this location when I was doing some experiments with my engineer's thumb. You can definitely hear the difference, although it is hard to describe. I quite liked a 1uF 63V electrolytic in parallel with a 100nF 100V box polyester. Paralleling up different types and value of capacitor tends to lower the ESR and the Q, and is a trick we use a lot for high frequency decoupling in digital circuit design, but it doesn't have a lot of effect at audio frequencies.

Tayda sells a 1uF ceramic cap for $0.12 so personally I think using an electrolytic here is a false economy.
https://www.taydaelectronics.com/capacitors/monolithic-ceramic-capacitor/1uf-50v-multilayer-ceramic-capacitor.html

[Ben N]

Wow, a 1uf MLCC! Whodathunkit? I guess that's the solution for all the pcbs that don't leave enough room for the fat 1uf box film caps.

[jonny.reckless]

I regularly use 1uF, 4.7uF and 10uF MLCCs in my designs. They're not expensive and work much better than electrolytics, especially for power supply decoupling and inter-stage coupling. I typically use box polyester for values between 1nF and 220nF, ceramics outside that, and electrolytics for bulk decoupling. I stopped using tantalums for audio a while back.

[rankot]

Wow, a 1uf MLCC! Whodathunkit? I guess that's the solution for all the pcbs that don't leave enough room for the fat 1uf box film caps.

I have recently bought 10uF MLCC from Aliexpress, 100 pcs were around 1USD, and they're just fine. I use them now in all of my builds instead of 10uF electrolytics in non-audio parts of a circuit. Didn't try to use them for audio flow, however.

[tristanc]

I regularly use 1uF, 4.7uF and 10uF MLCCs in my designs. They're not expensive and work much better than electrolytics, especially for power supply decoupling and inter-stage coupling. I typically use box polyester for values between 1nF and 220nF, ceramics outside that, and electrolytics for bulk decoupling. I stopped using tantalums for audio a while back.

Thanks Jonny - this is really useful. It's this type of info I wish I had when making my pedal PCBs, but I guess it depends on the use case, so something like the below is an oversimplification:

1uF upwards - MLCC or electrolytic
1nF - 1uF - polyester
up to 1nF - ceramics

Perhaps even using the same footprints where possible.

I guess the question is if the difference warrants desoldering and swapping over in a tight situation...

[befey]

Does anyone have any thoughts on how to add an external sidechain input to this?

[PRR]

> external sidechain input

[befey]

Thanks so much! I'm still wrapping my head around how the different parts of this come together.

[Ben N]

You'd probably want to use a switching jack on the sidechain input, so that it pulls the sidechain signal from the main input when there is nothing plugged into the external sidechain jack.

[befey]

Based on the above, this is what I've come up with for a sidechain input that switches only when there's something plugged into the sidechain input. I'm unsure how to do the actual switching.

It also seems extra complicated to duplicate so much of the input stuff. Is there some way to eliminate some of that duplication?

<schematic removed>

[merlinb]

Is there some way to eliminate some of that duplication?

Yes, use a switching jack socket:

[befey]

I'm trying to have the threshold pot for both inputs. It looks like this schematic only has the threshold pot on the sidechain input?

Edit:  oh.... I think I can see that the normal input sends up to the sidechain, but gets interrupted when there's something plugged into the sidechain input?

And thanks for your help!