Enveloping Timbre: envelope distortion

Keep the clarity of your attack before letting the distortion embrace you: the best of both worlds!

At least that was the original idea behind this effect, to keep the gain low for the short-lived, high-amplitude note attack of a plucked instrument, then ramp up the gain and the clipping for the decay when the level becomes steadier. This is quite the opposite of what happens with conventional fixed-gain distortion circuits, where the attack of the note is the most distorted part, having higher amplitude. Of course there’s nothing wrong and much to like about that, but going the opposite way would be cool and unusual.

This circuit falls in line with my recent interest in unusual VCRs and level triggered envelopes, elements present in the VCA-1 and the PUP. This time, I thought it would be interesting to try to use plain old BJTs as voltage-controlled element. Nothing new so far, BJTs, and diodes, have been used often in the past this way, with the caveats of keeping the levels low and accepting a bit of asymmetrical distortion, in part caused by a single transistor having to rely on reverse beta when the collector-emitter polarity is flipped by the AC signal.

I thought they would be perfect in this case, since how I was planning to use them forced the voltage across them to be equal, at worst, to the guitar input, and a tiny bit of distortion, while unacceptable for hi-fi, would go unnoticed in a circuit that has distortion as its primary goal, and in the world of electric stringed instruments in general.

Without getting needlessly accurate about it, I tested this was the case by ear and didn’t notice any effect on the sound with a configuration equivalent to the one in the circuit coming below.

The signal path

The basic circuit is my own take on the usual diode clipping distortion recipe (you can call it overdrive if you want): high input impedance, going straight into a high gain amplifier for best noise performance; clipping diodes; a shelving treble control; a recovery amplifier and out. Unimpressive, although probably good sounding, but this circuit has other tricks up its sleeve to be interesting.

The side-chain

I’ll start with the side-chain before getting to what’s actually being controlled. Originally I had tried a high gain amplifier or comparator going into a rectifier and peak detector, not unlike in the VCA-1, in parallel with the clipping stage. While this works in principle, using the other half of the TL072 this way led to a quite unpleasant signal bleed from the saturating side-chain amplifier to the signal path, most noticeable when gain was low: this repeated with different op-amps and I’m confident to exclude capacitive coupling since the two sections were as separated as possible, unless I reduced the gain of the side-chain to barely anything; LEDs to prevent saturation, capacitors to limit the slew rate didn’t do much to help, so I think the cause is to be found inside the chip and the only solution to use a separate one. It wouldn’t be the first time I encounter this problem: rejection between the two op amp sections is good but not perfect.

This was unfortunate, because I didn’t want to use a whole 8-pin package for the side-chain and I didn’t have much use for a second device there, which I couldn’t put in the signal path either. So I started working on an alternative side-chain exploiting BJTs.

First, why does the side-chain input amplifier need to saturate? Sure, it would need some gain, but how much? Many envelope-controlled effects such as compressors and filters react differently to different signal levels after all. My goal with this circuit was to get a timed delay on the gain though, not just because I wanted this delay to be consistent, but most importantly because this circuit becomes worryingly similar to an expander if the sidechain doesn’t respond equally to all signal levels: after all, the gain is normally low, and only goes starts increasing once the envelope is turned on, so effectively signals below this threshold would receive lower gain than the ones above.

This leads to a couple of important characteristics of the circuit. First of all, being an envelope controlled effect means that to be triggered again, a short recovery time between signals is necessary (this is made as short as possible); this means that there must be a level that’s considered “off” for the envelope to reset, so that noise and signal both below that will be ignored. This makes this circuit what you can call “gated” in a way, which is great in that the noise and pickup interference it outputs when not playing (which is when it’s really audible) is greatly reduced, since the gain will be low (as low as 4 already before the volume control, so usually unity or less). It also means that the gain will cut out if you let the notes ringing for too long though and that very quiet notes might be stuck at low gain: while there will be ways to alleviate that with the right settings, this also requires some cooperation from the player.

The side-chain is fed from the output of the circuit, and there are a few reasons for this: to drive the low-impedance grounded-emitter transistor and to drive it with an amplified signal already. The recovery amplifier at the output already usefully divides that threshold by 4, referred to the input, and any additional baseline gain set in the input amplifier will cut that down further. Besides this, I think taking the output signal gives some useful hysteresis: on the note start the gain is low, so it takes a higher input signal to trigger the envelope, giving noise rejection; once the signal is present and the envelope is high, the gain is higher so the note can ring a bit further before the envelope turns off.

It sure would be nice to bias this transistor at a predictable voltage above the threshold of the next stage, so that any positive peak can readily send it below that, while negative peaks have little or no effect: the shown arrangement for Q2 achieves just that, in a way reminiscent of a famous distortion (fuss) circuit at that. This stage makes up the level detector: as soon as signal is present, the transistor quickly discharges C6, which stays that way with any frequency of interest because the 1M pull-up is much slower and negative peaks can’t let the collector go that high anyway because of the diode; actually very large negative peaks have a chance to go straight through the diode, rectified, and further contribute to pulling C6 down (this effect is intentionally made negligible in this case).

Q4 makes up the actual “envelope generator”: triggered by Q2 pulling the base down, C9 slowly starts charging through the two resistors and the "Attack" pot, and this is the attack time. Very soon after the signal ends, C6 gets charged enough that the base can be pulled up again by R9, and very quickly C9 is discharged, turning the envelope off. R13 isn’t really necessary, since the discharge time is already limited by how much current R9 and the transistor’s beta can sink, but it’s there just so the transition isn’t so abrupt. The envelope is then fed to the two VCR transistors with one resistor each.

The side-chain makes a clever use of which polarities have the fast transition of the transistor pulling down vs. the slow pull-up from the resistor. To choose the values, I started from the end and worked backwards. I’d have liked to keep at least some of the caps at most 100n for size reasons and to reduce electrolytics so that’s one starting constraint. I’ve recorded some notes and analyzed them to estimate the “attack”, that is the time before the amplitude suddenly drops, which seemed to be about 100-150 ms. This is reflected in the C9 charge time, which turns out to be, considering the drop across the base resistors plus Vbe, about 50-60ms with attack at minimum, which is subtle and only really catches the attack itself, about 600 at maximum which is audibly slow and will give a swelling effect, cool in its own regard.

The resistances and capacitances needed here were doubly constrained: if the resistors were larger than this I couldn’t have used a standard value pot for the attack control (but could have used a switch) and they risked being too large to give enough current to turn on the VCR transistors even at those small signal levels, even if it possibly meant a smaller capacitor. Vice versa, smaller resistor values would have resulted in a larger capacitor and again a similar issue with R9 not being able to turn on Q4 at the currents involved: the latter effect was somewhat fun to observe in practice, with the envelope output slowly creeping back up and turning on by itself after some seconds if left to itself. To get around that, I’d have had to make R9 smaller and C6 larger, giving up one small capacitor and possibly having to rework more values upstream, so I settled on this compromise. C6 was chosen so that the recovery on note off was quick enough to not affect playing. but not so quick to cause ripple with lower frequencies. In this case it’s about 7 ms: in case of ripple the capacitor can be made larger but I found this to work well for guitar.

C8 and R7 finally might seem a strange combination, but with low frequency input and a smaller capacitor, the output of Q2 crept back up even while there still was signal. A 10μ capacitor eliminates the effect. R7 doesn't have a big effect unless it's too large, but this value avoids negative peaks making C6 negative right through the diode and making the recovery slower than intended.

The VCRs

Finally we get to what the envelope is controlling: remember that ENV is normally low with no signal, and somewhat slowly goes up with a signal before going back down when the signal ends. The positions of the two VCR BJTs are already nicely AC coupled and going to ground, which makes things easier and requires less capacitors.

Q1 controls the gain of the first stage so that it’s 1+[Gain/(R1+Ratio)] at the beginning and 1+(Gain/R1) then. “Gain” sets the overall gain, while “Ratio” roughly corresponds to how much lower it is when the envelope is off. Unlike most swell and other volume effects, this controls a non-inverting amplifier, so the gain will always be at least 1 and the attack at least a bit audible. I like to set “Ratio” together with “Gain” so that the attack has just enough gain to be clearly audible while still being clean, but you can set it how you want: higher ratio for a muffled attack and a swelling effect or lower ratio for more “bark”, up to shorting out the VCR and turning it into a more conventional distortion. You want to set “Gain” first and adjust “Ratio” accordingly.

The second VCR, Q3, lifts the treble control for the attack, connecting it back to normal for the rest of the note. The attenuated treble is most useful together with distortion, and taking it out for the attack really makes it pop, both in volume and definition. Additionally, even with “Ratio” taking Q1 out, you still get a brighter note attack by this, which is really interesting. I didn’t add an additional control on this transistor, to keep the controls low in number, but you can if you want, like a switch shorting it out or a second 1M ratio control.

I’ve spent some time trying different diode forward voltages, both hard and feedback clipping, trying to get the most even volume between the attack and the tail of the note: “surely a schottky, with the low forward voltage, is the best option so that even when clipping the level can’t rise much”, I thought. I even went from a BAT41 to the even lower forward voltage of the 5819 (while changing to hard clipping so that capacitance wasn’t an issue), then realized that by adjusting the “Ratio” control, I could get it so that the attack is the same loudness without it clipping: do I have a Vf three times higher? I just need three times more gain for the same effect! Schottky diodes resulted in a little bit of clipping even with gain all the way down, and meant at higher gain I couldn’t get a clean attack unless I used a larger value for “Ratio”, which made most of its range useless at low gain since it can’t drop below unity. Also, with a larger amplitude of distorted signal, you keep the option of keeping the attack quiet with low “Ratio”. U1B has just enough gain to give plenty to output and side-chain without clipping itself, which is easy to do with the well-defined hard clipping amplitude. Since noise isn’t an issue at this point of the circuit, I used values that allowed me to go for a 100n capacitor.

Demo

This is another of those effects that is best experienced directly, but I've tried to give an idea of the possible options and control ranges:

Conclusions

While I didn't quite end up with a realization of my original idea, I instead made a very interesting, dynamic and unique distortion if I may say so, and a pretty simple one at that, probably more than what I had planned. Even if it's not your thing, this shows interesting possibilities in envelope controlled pedals, in combining envelope and distortion, and as proof that cheap and reliable BJT VCRs can still be more than decent for effects pedals.

Thanks as usual to aotmr, for the helpful suggestions and even more for the feedback and discussion through the design.