A simpler octave down concept

[Fancy Lime]

Hi folks,

so a while back I analyzed the classic Boss OC-2 and came to the conclusion that this is quite possibly the ideal way of making an analog octave down but Ileft off with a feeling that there must be some way to make this a bit more DIY friendly, aka simpler. Quick recap of the important parts of the OC-2:

There are two peak detectors, one for the positive, one for the negative peak. They work by comparing the positive or negative envelope of the signal to the signal itself. These two detectors feed the set and reset inputs of a CD4013 flip flop, which then switches the output of the flip flop high and low consecutively at each "set" but only if there has been a "reset" in between. This means that the output of the flip flop changes at (or actually shortly after) each wave form peak at the input of the peak detectors. The flip flops output is then used to essentially shifting the "DC" reference (not really DC anymore but it would normally be where a DC reference goes; I'm sure this has a proper name, I just don't know it) of a summing amplifier with exactly half the frequency of the original tone, making the output of said amplifier a kind-of-sort-of sine triangle wave (not really but not to far off), assuming that the input was roughly sine shaped. And it turns out that that sounds pretty good and tracks fairly decently, considering it's an analog device that tries to transpose a frequency, which is pretty darn impressive, if you ask me. Caveat: The approach works poorly with square wave inputs but that is dealt with by preceding low pass filtering to an extent that seems sufficient for guitar or bass use.

So the idea I got stuck in my head is basically as follows: Couldn't we make a comparator that compares the input signal with a phase delayed copy of itself and use that to feed a ripple counter, to get a very similar effect (counting only up to the output of the CD4013 in the OC-2)? The point being, that the signal and its copy cross over just after the high and low peaks of the input sine wave, just like in the OC-2. If we make the comparator hysteretic (so, a Schmitt trigger) and choose the delay of the copy right, that should, hopefully, track OK (given similar low pass filtering preceding the whole shebang as is done in the OC-2). For a ripple counter one could of course use a CD4024 or something alike, which would also allow a second octave, third octave... But I thought: why not make a ripple counter from an inverter based flip flop by triggering the flipping via a JFET, which is fed the high-pass filtered output from the comparator, thus flipping only on the uptick, hence producing on it's output an octave down square wave with the zero transitions just after the highest point of the original input sine wave, same as the OC-2?

Here's a haphazardly cobbled together schematic:



Yes, the JFET symbol is wrong, I didn't have the right one and did not bother modifying the existing symbol. Also, since this is in the brein-fart stage of development, most R and C values need working out and of course none of this is tested. The buffer up front just provides a DC reference, in the real device there would probably be a 3rd order Sallen Key low pass biased to V1/2 in this position.

All that being said, I cannot shake off the feeling that this is one of those "too good to be working in real life" ideas. Normally I would just throw it on the breadboard, try getting it to work, give up after a few fruitless days or weeks and bury the idea in quiet shame. Unfortunately, I am really strapped for spare time at the moment and it would probably take months before I would get around to even try this in its most basic form. So my question is essentially: please tell me why this won't work anyway, so I can get on with more promising projects.

Thanks a bunch,
Andy

[anotherjim]

I'll have to digest that Andy...
...are you sure it won't still mistake the extra zero-crossing caused by a strong harmonic as marking the fundamental frequency period? That's the usual problem facing these things.

[merlinb]

are you sure it won't still mistake the extra zero-crossing caused by a strong harmonic as marking the fundamental frequency period? That's the usual problem facing these things.

Exactly. If a single comparator could do the job there would be no need for the phase delay in the first place. The problem is how to make the comparator not produce the red trace when there are wiggles on the waveform. The wiggle is on the peak in this example because I was talking about the OC-2 method, but if you look at a guitar note you'll see the harmonic wiggle 'ride' the fundamental and slide down into the zero-crossing region as the string vibrates.

https://www.diystompboxes.com/smfforum/index.php?topic=76997.20

[Fancy Lime]

Hm, well. The variant I am proposing is certainly not as robust in terms of tracking because of the single comparator. In my experience that ca still work fairly well (tested with signal zero crossing detectors, though, but same principle) if the hysteresis is chosen well and the signal is filtered properly. But yes, the "riding harmonics" will be more of of a problem here than in the OC-2. It's a tradeoff between stability of tracking and complexity. My question was more: Would the thing I'm proposing work at all. In other words: would it be worth trying to see if I can get it to track decently enough or are the fundamental problems with the idea, which I am overlooking that mean it'll never work or is not likely to work anywhere near well enough to be useful enough to justify the reduced complexity?

Thanks,
Andy

[merlinb]

It'll 'work', it'll be glitchy that's all. The phase shift in your design is going to be very slim and variable with frequency -I think you'll get very poor tracking on real audio as opposed to sine waves. But if it's very glitchy then you can advertise it as an octave scrambler! 

[Fancy Lime]

Hm, you're probably right. The frequency dependence of the tracking got me worried too. I have never managed to reign that in enough to make a reasonably useful device in the past with similar, although not identical, concepts. But the idea of making an intentionally glitchy, badly tracking "octave scrambler" is actually not bad. Not that there are not plenty of those around already, but so are fuzzes, so why not. I have an old Rocktron Purple Haze Octavider sitting around that does just that: Awful tracking and lots of chaotic glitching. Not the most useful thing in the world but an interesting "effect effect" IYKWIM. I've never seen a schematic for that one, maybe I'll get around to tracing it one day. Trashy sounds are fun sometimes. But for a well tracking octave I might indeed be better off with the original OC-2 type design. Oh well...

Thanks,
Andy

[merlinb]

I think a more effective 'simple' approach would be first to filter and amplify the signal by, like x50. This might clip the peaks but we don't care about those- we just need the signal to be big enough to overcome diode drop [see next step].
Now pass the signal through a couple of series diodes to 'core out' the zero crossing, you know, like crossover distortion. Lightly filter it again too, to remove any discontinuities. This should erase most of the wiggles and garbage around the zero crossing.
Now pass that signal to a zero-crossing comparator (hysterisis may further help?), and then on to a frequency divider in the usual way (e.g. U-boat).
The amount of 'pregain' we apply would then become the tracking trimmer, or maybe put a variable resistance in parallel with the 'coring' diodes.
I think fair results might be possible this way?

Or maybe, just maybe, amplify the signal, zero cross detect, then filter the s*** out of the output, and then into another zero-cross detector...

[Mark Hammer]

That's the stick in the spokes of octave dividers.  Flip-flops will have a threshold for being triggered.  If the input signal were steady-state, life, and your circuit idea, would be peachy-keen.  The trouble is that the input signal is variable not only in amplitude, but also in pitch, as well as harmonic content.

As Merlin wisely suggests, improvement in the behaviour of any triggering sub-circuit can be fostered by making the input signal more like a steady-state oscillator.  In that spirit, I'd suggest a gain stage with fairly strong clipping via feedback diodes, lowpass filtering via a feedback cap in parallel with the diodes, and maybe even a Q&D gate via a back-to-back pair of diodes in series with the output of that op-amp stage.

I suspect  one of the reasons why the MXR Bluebox is a fuzz+suboctave is so that the signal triggering the flip-flop is boosted enough to be almost always squared.  Once you have that squared-up signal, may as well make productive use of it, eh?  In their case, they gate the overall output to reduce sputtering, as opposed to gating what feeds the flip-flop.

[EBK]

For me, the most annoying part of analog octave circuit design is that, when I am thinking my way through a complicated but elegant design, a little angel appears on my shoulder and whispers into my ear, "You can spend more time with your family, save money, and have much better results if you'd just buy a POG."  Analog octave circuits are very fun to think about though.

[Fancy Lime]

What you describe is almost exactly what I did when I first started messing with octave down effects. Filter, amplify excessively, filter again, Schmitt trigger (which could also be implemented with crossover distortion diodes although I did it differently). It tracks pretty well. However, I used the output(s) of a CD4024 directly as audio and just ran them trough another fundamental extractor. Not bad for it's simplicity but of course very different from the OC-2 in that the octave has no dynamic whatsoever and sounds very synthy. And that just wasn't what I was looking for.

What I am looking for is what you (Merlin) call the "boss method" (if I remember correctly) in the U-Boat article, so flipflopping on the peaks rather than the zero transitions of the input signal. The advantage, to my ears or at least for what I am looking for with this particular thing here, is that the output retains most of the input dynamics and does not need much filtering to sound more or less sine-like, which I like for a sub-organ kind of sound. So maybe the OC-2 method remains the way to go for now. One can always additionally take the output of the flip-flop directly for a synth sound.

Andy