Idea for better tracking in OC-2 style bass octaver

[Fancy Lime]

Hello Boys and Girls,

I've been pondering the whole octave down topic for a while and was never 100% happy with any of the commercially available or diy-able options. I like the sound of the Boss OC-2 and its many work-alike cousins but I also like me some fuzz and the OC-2 does not like to be fed a fuzzy signal. The main reason for this seems to be the fact that the OC-2 needs to trigger the flip flop at the peaks of the signal and it does so by comparing the signal to its own envelope. While this idea is neat, it works best with sine waves and the like but not at all with square waves, meaning distortion before the octaver will have (potentially undesired) consequences.

It is much easier to get good tracking if we detect the zero transition instead of the peak of a signal. But then we don't get the OC-2 sound. Damn. So that got me thinking: why not get the zero transition and then trigger the flip flop (or a ripple counter) 90° later?

I think this can be done by converting the input to a square wave with a Schmitt trigger, then converting the square wave to a 90° shifted triangle wave with an integrator with a cut-off at ~20Hz, and lastly converting that back to a square wave with a second Schmitt trigger. With the Schmitt triggers properly adjusted, this should track well and the zero transition of the second Schmitt trigger output should be only slightly after the peak of the input waveform. A minimum working example which uses this method to trigger a CD4024 ripple counter is attached.

I was wondering why I have never seen this method used. Seems simple enough, at least way simpler than the OC-2 method, which has been copied a million times. No idea who used it first, does anyone know? Probably wasn't in the OC-2, I think the method has been around since the 70s. Can someone tell me, why the Schmitt-integrator-Schmitt method won't work? Or will it? I have not build it yet. If someone has time to try, let me know how it goes.

Cheers,
Andy

SORRY: I cannot get a connection to the image server, seems to be down or too busy. I'll try to upload the image again asap.
Edit: Now it works again.

[Fancy Lime]

Now it woks again, here's the circuit:



BTW: the bias-network for VR and power supply filtering need to be added, but you'll get the idea, I hope. This is just a minimum example to demonstrate the idea.

Andy

Edit: I found an error on one of the resistor values and replaced the image with the corrected version.
Edit: Added the missing mix resistors. Thanks Paul!

[Fancy Lime]

And here's another idea for the same problem:



This uses the same idea of comparing the envelope of a signal with the signal itself, that the OC-2 uses. But then it resets the envelope follower half way on the down swing. This way we can use an asymmetric Schmitt trigger instead of a comparator, which should help with the stability. I'm not super convinced by this concept, I find the other one more elegant. Anyway, let me walk you through it with some graphs:



Graph a) shows (very approximately) what the envelop follower does without the reset, just as in the OC-2.
Graph b) shows what I think it does with the reset. Can someone tell me if that is really what happens? I neither tried nor simulated it yet, just no time at the moment.
In c) we see the difference between the curves in b), which is what the comparator/Schmitt trigger sees at its inputs and the output this should produce. this should track better than the original OC-2 because once the positive difference on the input of the comparator has set the output low, the input then needs to go very low (up to 0.8V + the diode voltage drop below VR depending on the setting of the trimmer) to swing the output back up. This should kill any and all parasitic oscillations, me thinks.

Anyway, just a concept at this point. Let me know what you think. I'll try to build and refine both of these but that may take a while. Kind of busy these days.

Cheers,
Andy

Edit: Added the missing mix resistors. Thanks Paul!

[Mark Hammer]

It's not the square-wave quality of the signal itself, but the square quality of the overall amplitude envelope.

The triggering of a counter or flip-flop relies on the signal level exceeding some critical value.  Unfortunately, guitar signal amplitude is a highly variable beast.  Sometimes it's hot and sometimes not.  Where it causes problems for octave dividers is when it is at the in-betweenie zone such that maybe it triggers the octave division for a few cycles and then doesn't and then does again.

The ideal circumstance is to feed the counter/flip-flo with a steady level signal from, oh I dunno, a signal generator.  You'll get flawless octave-division.

So how do we mimic that ideal situation?  One way is to boost the dickens out of the input so that it remains above threshold as much as possible.  That could be either via pure boost, or perhaps via compression.  That helps, but is no panacea since it also raises the level of what would normally be low-level signal aspects that the counter/flip-flop would otherwise ignore.

Another strategy, that Craig Anderton used in the PAiA Rocktave was to use a compander chip to maximize signal level prior to octave division, via compression, and gate out the sputter parts via the expander stage of the compander.  That also works; better than most, but not foolproof.

What occurred to me recently was use of series diodes to block any low-level signals, prior to boosting and division.  In a sense, requiring signals to be above some critical "pre-treatment" threshold, so as to be "worth" boosting or compressing, would weed out those parts of the signal that might give you problems further on in the signal path.  It would provide a "harder" threshold for octave division, but wouldn't let any in-betweenies get through where they could cause sputtering.

[Fancy Lime]

Hi Mark,

I don't quite follow. Wouldn't a "square amplitude envelope" mean that the peaks of the signal are all the same hight? And is an audio limiter not essentially an "envelope square wave shaper"?

What triggers the flip flop in an OC-2 type circuit is the output of a comparator, which compares the envelope to the original signal. What comes out of the comparator should be no problem to handle for the 4013 since the signal here should have very steep flanks and no sputter near the zero transition, which is what the 4013 looks for, no? I think the problem lies with the input of the comparator. It the original signal is relatively small compared to the precision of the match between original signal and envelope, then we get sputter from the comparator. A loud signal has less of this problem but not with the wave shape of the input signal.

Square wave signals on the input are a problem because the envelope does not leave the signal at the center of the peak anymore but at the zero transition, which causes the comparator output to be 90° shifted from where it should be, thus messing up thus messing up the delicate stitching of the waveforms after the flip flop.

The series-diodes + boost idea would certainly help with methods that detect zero transitions, because that i where this filters out the dirt. But I don't see what that would help with peak detection other than making weeding out the weak signals that are problematic. Wouldn't it be better to make the weak signal trigger better instead of killing them? Also: Wouldn't a Schmitt trigger instead of series diodes before a comparator achieve the same effect but be adjustable?

Cheers,
Andy

[PRR]

Minor detail: your dry/octave mixer probably wants mix resistors. As-drawn, if either knob is full-down, no signal gets through.

[Fancy Lime]

Hi Paul,

whoops, you're right of course, there are supposed to be two 100k resistors which I forgot to put in. I'll change that.

Thanks,
Andy

[Mark Hammer]

Hi Mark,

I don't quite follow. Wouldn't a "square amplitude envelope" mean that the peaks of the signal are all the same hight? And is an audio limiter not essentially an "envelope square wave shaper"?

What triggers the flip flop in an OC-2 type circuit is the output of a comparator, which compares the envelope to the original signal. What comes out of the comparator should be no problem to handle for the 4013 since the signal here should have very steep flanks and no sputter near the zero transition, which is what the 4013 looks for, no? I think the problem lies with the input of the comparator. It the original signal is relatively small compared to the precision of the match between original signal and envelope, then we get sputter from the comparator. A loud signal has less of this problem but not with the wave shape of the input signal.

Square wave signals on the input are a problem because the envelope does not leave the signal at the center of the peak anymore but at the zero transition, which causes the comparator output to be 90° shifted from where it should be, thus messing up thus messing up the delicate stitching of the waveforms after the flip flop.

The series-diodes + boost idea would certainly help with methods that detect zero transitions, because that i where this filters out the dirt. But I don't see what that would help with peak detection other than making weeding out the weak signals that are problematic. Wouldn't it be better to make the weak signal trigger better instead of killing them? Also: Wouldn't a Schmitt trigger instead of series diodes before a comparator achieve the same effect but be adjustable?

Cheers,
Andy

Those are all fair points.  Ideally one wants the signal to be "binary":  clearly well above or well below threshold...all the time.  Limiting, with heavy compression, would move more of the signal above threshold more of the time.   Most dynamic-control devices will focus their effrts on the attack portion of the plucked note.  Unfortunately, we also have to contend with the decay phase of the signal, and that's where analog thresholds cause us such grief, because unlike synths, decaying strings can't make up their mind about what level they want to be at, moment to moment.  It's the same envelope ripple during the decay phase that drives noise gates, autowahs, and compressors crazy.  My suggestion to use series diodes was to essentially provide primitive gating, and "wipe the slate clean" during the decay portion of the note.