Crossfeeding regeneration to another compander - SHecho project

[Mark Hammer]

I whipped up a pair of Dean Hazelwanter's SHecho boards, and am busy populating them when I have the chance.  My plan is to build a pair of them into a rackmount box with a couple of interesting stereo features.  One of those features involves crossfeeding the regen signal to the other delay channel.  If the two delays are set for different times, and/or different delay tones and/or regen levels, then what comes out of the A and B output jacks will be kind of interesting.

Now, if one were to simply reconnect R4 (see project file at GGG) from channel 1 over to the compander input in channel 2 and do the same thing (i.e., channel swap) with R4 from the second channel, that wouldn't be too much of an issue.  I was thinking, however, of building in a feature where I could send the guitar signal to both delays simultaneously (set for different delay times, and maybe with some hi-cut or lo-cut in the regen path to diversify the regen signal tone) and take the two regens and feed them to the same compander on a single channel.  In other words, R4 from channel A and R4 from channel B would both go to the same point on, say, channel A (pin 12 of compander chip).

Is there something about either the behaviour of companders, or the design of this particular unit (and remember it is similar to the AD-3208) that would suggest this "pooling" of two regen signals is a bad idea, or that some sort of compensation strategy needs to be employed to make it work?

(Just as an aside, I have a bunch of 41256 chips pulled off an old video card.  If anyone needs such a chip and can't find one, drop me a PM.)

[caress]

this sounds like a great idea!  using seperate BPFs on the delays would make an even more distict difference between the two.  your idea is very similar to some synth designs that have delays in them and can use cross-modulation and the like.  maybe you could find some of those resources to help you out...(could be nice to feed both delays into each other, too)  i just started thinking about a delay where the delay time is controlled by sample&hold...might be interesting to make it a dual delay as well  

[Mark Hammer]

Thanks for the encouragement.  I *was* pondering building in some sort of send/receive loop for external processing of the regeneration signal, but I think what I described incorporates enough craziness and options for now.  Independently controllable regen and mix and separate outputs also permits a lot of those send/receive type experiments to be accomplished anyways.

What started me on this path was some stereo experiments with the Line 6 Echo Park.  The EP is not a "true" stereo device, but rather a "mostly stereo" device (I'm reminded of Billy Crystal's speech about someone being "mostly dead" in The Princess Bride).  If you plug your guitar into input A, the repeats will show up in both output A and B, and in some modes will alternate between A and B.  I decided to take the A output, process it, and run it back to input B, sending output B to the amp.  Because the B output includes some of what was already fed into A, as well as what went into B, and because the B output aso gets crossfed to A (where it gets recirculated back into A after some changes), you get these really interesting things where each iteration is entirely different.

As much fun as that is (and it is a LOT of fun), being able to separately adjust the amount and tone of crossfeed from each channel, as well as the delay time on each channel (the EP's controls are all global, even though it is ostensibly a stereo device) starts to open up a world of possibilities.  The stock SHecho shows a 150k min-delay resistor in series with a 500k delay-time pot.  For purposes of being able to more readily dial in differential delay times to achieve specific rhythmic effects, I think I'm going to split the delay time up into multiple ranges, using 1% fixed resistors, and use a smaller value of pot so I can get greater precision.

Low-cut consists simply of using smaller cap values in place of C7, and high-cut consists simply of sticking a 10k+10k resistor in place of R4, and run a choice of caps to ground from their junction.  I've been doing this for decades and it works great.  Though 10k+10k is less than the stock 22k, the risk of irritatingly runaway feedback is reduced by the lowpass and highpass filtering, each of which can reduce overall signal amplitude.  The reduced series resistance is to offset what the filtering takes away.

Without wishing to complicate this posting too much, if each delay channel had a dry-cancel switch (disconnecting R14 from the mixer stage), then I could, for example, dry-cancel B, cross-feed the B regen to A, take the wet-only output of B, process it and return it to B's input.  Each time B repeats, it gets mixed in with A, but it also goes out, gets changed around, and then gets delayed again.