3 or 4 octave up square wave ??

[markusw]

Besides that my animal is alive ....any comments on how to improve the circuit (MrCoffee, Gez??).

Sorry for not posting the schem but I am too lazy to draw it.

Thanks a lot in advance.

Regards,

Markus

[gez]

Difficult to offer advice without a schematic really.  If it works though...

[markusw]

Difficult to offer advice without a schematic really. If it works though...

OK, I will draw a schematic 

Would a block diagram sufficient ??

Markus

[gez]

Difficult to offer advice without a schematic really. If it works though...

Would a block diagram sufficient ??

Should be enough...depends how simplistic it is though! 

[markusw]

OK, I drew a hopefully not too simplistic diagram. I also included the doubler schem to make it easier to follow.

The cap values for the third doubler stage are different from the ones I posted earlier. Sorry for the mistake. It seems that the values are already pretty close to the lower limit. Reducing e.g. the first stage's values to 470n and 47n already doesn't yield a real triangle wave after the first integrator for low notes (the signal is cut off at the top and bottom). However, 660n and 66n might be possible.








Here's a copy of my previous post (since it is already one page back).


--------
In the meantime I breadboarded the AN-41 doubler. Basically, the schem can be described as follows:

<non-inverting opamp input gain stage> into

either

a: 6 or 12 dB/octave bp filter with Fc of about 30 Hz into LM311 comparator

b: 12 dB/octave bp filter with Fc of about 30 Hz into self adaptive Schmitt trigger (very similar to Boss OC-2) including the two flip flops in series (i.e. output is one octave down) into LM311 comparator

c: 12 or 24 dB/octave Buttherworth lp filter with Fc of 20 Hz  into LM311 comparator

a, b and c go into the first comparator of one of three consecutive AN-41 doublers. The three stages look exactly like in the AN-41 w/o the 100p cap (LM311 for both comparators, MC33171 for integrators). The values for C4 and C2 are

* 1st stage: 1µ and 100n for C2 and C4
* 2nd:        470n and 47n
* 3rd:         220n and 22n

Vref is provided by an opamp buffer and is heavily filtered as V+ (12V) is.

Basically the three octaves up work quite nicely even over the whole 5-string bass freq range (30 to ~450 Hz). However, three consecutive stages seem to be the maximum. After a fourth stage the sounds get too nasty (even for my taste  ). When fed with a sine wave on the oscilloscope the wave shape at the output looks pretty squarish even after two octaves. Symmetry gets worse with each stage which seems to be the reason why 3 stages is the max. This is in accordance with the LTSpice sims although LTSpice exaggerates slightly. When looking at the scope trace of the triangle wave at the integrators' outs the edges have the same slight bump (break, don't really the know the proper expression) as in the sims.

All three variants (a, b, c) display some latency. Variant (c) is clearly the worst, the 24 dB version even worse than the 12 dB. As predicted by Mr Coffee the adaptive Schmitt trigger(b) doesn't improve the situation. The only advantage of (b) is the slightly more symmetric wave shape at the output of the second flip flop. After the first flip flop the wave shape is clearly less squarish than with a "plain vanilla" comparator .

Why variant (c) is that inferior I have no idea for. The signal is clearly heavily lp filtered, but the performance in combination with the three AN-41 doublers is depressing.

Obviously, since the whole mess is on breadboard and since there is no gate either there is a lot of noise when the strings are muted. I increased the offset voltage of the first comparator with a pot. Now the circuit is quiet when I carefully mute the strings.

When used as a carrier for my Ring Stinger clone background noise isn't really an issue since the RS gates anyway. The sound when used together with the RS is already pretty cool but the latency is annoying since immediately after plucking the string the sound is not immediately "present" like with the fixed stock RS oscillator.

The PLL in contrast has definitely higher latency (although I tweaked it quite a lot) but obviously yields a much " purer" square wave.

I also added a orange squeezer to the breadboard but it didn't improve latency of the AN-41 doublers even when heavily compressing.

Some Qs:

* what is that 100p frequency compensation cap for. do you think I need it??

* where does the latency come from? how can I minimise it?

* do you think that adding a gate in front of the first comparator of the first doubler would reduce latency??

* what can I do in general to improve the performance of the whole circuit
--------------------

Again thanks for your help!!

Regards,

Markus

[gez]

At first glance it all seems very complicated...perhaps unnecessarily?  Presumably all these filters are for evaluating what works best, and you'll settle on one then follow it with a sine to square converter?  This is done by the first comparator in the schematic you linked to and once the conversion is done there's no need to include it in the subsequent doubler stages (just the integrator and second comparator in each module are needed). I mention this just in case you've included the input squarer in all subsequent doubler stages - it's superfluous.

Incorporating trim pots into your design would probably help maintain symmetry of the waveform and allow more stages, but would require using a scope to set it up.

I'd guess that the latency you mention is a result of the low cut-off in the filters.  That's going to cause a phase-shift, which will translate as a slight delay.  Presumably you've set the cut off so low to prevent yodelling?  If you set it higher it might improve latency. 

Not exactly sure what you mean by 'adaptive schmidt trigger' so can't comment, but the output of the second flip flop should have perfect 50:50 duty cycle if it's set up to divide down, no matter what.

The compensation cap you mentioned is necessary with some op-amps.  If you're using bog-standard amps, and the data sheet doesn't mention them then just pull them.

Edit:  apologies for the brevity of this reply...I'm knackered! 

[toneman]

Hey Markus,
Thanx 4 the block diagram.... 
Now i see what U have been doing...
Not much time for the QuadDroopler yet....   
I wanted to breadboard the AN41 thingie also.
Just not enough time til the RingOlations are completed.
But, as I suggested 2 U earlier,  the comparator at the front of the AN41
is there to give the sharp SQ to the integrator/triangle generator following it.
My idea was to use the Adaptive Threshold Comparator *in place* of the first comparator.
Escentially, the front end of an OC2, or a DeluxeOctave, or the EH-G_Synth.
From the preamp to the first FF, not the dividor FF..  Output of the first FF to the Opamp integrator.
My thought was that the ATC, after a 12db LPF, would better remove the 2nd harmonic.
Then, the integrator would be able to do it's job "more easily".
A good, simple compressor, at the front, before the LPF, is probably a good idea also.
Since it's warming up here in Sacto, i'm thinking of etching a few DrooplerEXPs pcbs.
Still haven't finished the Rocktive yet, as I wanted to see the tracking characteristics of
a 570based extractor also.  Again, no time....
Oh, how bout 4 pitchshifters??? 
My Ibanez HD1000 will shift up one octave.   Maybe a octave & a half(?)
Feed one into the other into the other etc.   (?)   
Since i have one, i'm gonna try it for the RingOlation carrier.   
Keep it alive!!!!   
afn
T

[markusw]

Hey Gez,

thanks a lot for your answer.

Presumably all these filters are for evaluating what works best, and you'll settle on one then follow it with a sine to square converter?

Exactly.

This is done by the first comparator in the schematic you linked to and once the conversion is done there's no need to include it in the subsequent doubler stages (just the integrator and second comparator in each module are needed). I mention this just in case you've included the input squarer in all subsequent doubler stages - it's superfluous.

I've included the "first" comparator in all stages based on one of MrCoffee's comments:

Or are you running the output of the EX-OR CMOS IC directly into the integrator for the second stage? CMOS output impedance isn't very great or symmetrical.

Incorporating trim pots into your design would probably help maintain symmetry of the waveform and allow more stages, but would require using a scope to set it up.

Trim pots?? For reducing offset voltages??

I'd guess that the latency you mention is a result of the low cut-off in the filters.  That's going to cause a phase-shift, which will translate as a slight delay.  Presumably you've set the cut off so low to prevent yodelling?  If you set it higher it might improve latency. .

I set the cutoff so low to get rid of second order harmonics at low notes. Don't know how to avoid this though.

Not exactly sure what you mean by 'adaptive schmidt trigger' so can't comment, but the output of the second flip flop should have perfect 50:50 duty cycle if it's set up to divide down, no matter what.

the adatptive schmidt trigger bascically looks like in the Boss OC-2. And yes, the output of the second flip flop is really much closer to a 50:50 duty cycle. You loose one octave though. Considering that I would like to go up 3 octaves and 3 octaves up seems to be the max with the AN-41 doublers it's unfortunately not a real option.

The compensation cap you mentioned is necessary with some op-amps.  If you're using bog-standard amps, and the data sheet doesn't mention them then just pull them.

OK, I will check the data sheet for the MC33171!


Hey T,

thanks for your comments too. Will comment on them when I'm back from work.



Regards,

Markus

[gez]

I've included the "first" comparator in all stages based on one of MrCoffee's comments:

Or are you running the output of the EX-OR CMOS IC directly into the integrator for the second stage? CMOS output impedance isn't very great or symmetrical.

Fair point, but have you tried it?  I would.  I always see what I can get away with.  In the integrator, make the cap smaller and the input resistor larger to minimise loading and see if it's ok.  As is, the values don't look too much of a problem.  Like I said, always see what you can get away with, you have nothing to loose if it's on a breadboard.  If it doesn't work, lesson learnt.

Trim pots?? For reducing offset voltages??

Tweak the second comparator's trigger threshold using a scope and you're going to be able to get perfect 50:50 symmetry (well, with a test signal). 

I set the cutoff so low to get rid of second order harmonics at low notes. Don't know how to avoid this though.

Use a more complicated 'fundamental extractor'.

the adatptive schmidt trigger bascically looks like in the Boss OC-2. And yes, the output of the second flip flop is really much closer to a 50:50 duty cycle. You loose one octave though. Considering that I would like to go up 3 octaves and 3 octaves up seems to be the max with the AN-41 doublers it's unfortunately not a real option.

I don't have the schematic for the Boss doo-dah at hand, but if you've squared off the signal already with a Schmidt then you can divide down twice with the flip-flop.  Using a decent square to sine converter will allow you to do this.  If I recall though, the Boss doesn't present a clean enough edge for the clock input of a flip-flop.  Preceding the clock input with a 555 set up as a Schmidt works nicely.  Don't know anything about the XOR chip mentioned, but if there's a spare device in the package you might be able to use it as a buffer to the flip-flop's clock input.  Finally, if I recall correctly, the flip flops in the 74 series have schmidt inputs, though these only opperate from 5V supplies.  Personally, I'd run this whole circuit from a 5V regulator using CMOS op-amps.

[jmusser]

Will this device have individual octaves that you can dial in and out, or will they all be taken to a single output to be blended? This is something I've been interested in for awhile. I believe the E&M Harmonizer was supposed to do something silmilar to this, only it had additional 5th and 3rd tones with it. If I remember right, they were all dialed in individually by a rotary switch. It would be cool to be able to blend the fundamental in with any individual octave up at any level, and any combination. My Arion Octave (down octave pedal) does this wonderfully with fundamental, and one and two octaves down. That's what I've been looking for in an up octave pedal. From what I've read, up octaves are not as easy to differentiate between tones as down octaves. This may also be due to the way the ear perceives the tones. Maybe our ears can't properly separate individual up octaves, but I've also never had them to sample either to varify this. I know R.G. touched on octaves and the ear's perception of sound. That was an eye (ear) opener.

[markusw]

Had a little time to think about all your suggestions. Although some aspects get clearer other questions arise.

My idea was to use the Adaptive Threshold Comparator *in place* of the first comparator.

This is basically variant b1 I suppose. I just have the comparator of the first AN-41 doubler included, but this shouldn't really matter, should it?? The output of the comp should look basically the same as the output of the flip flop (an assumption). So do you think the comparator will hurt ?

From the preamp to the first FF, not the dividor FF..  Output of the first FF to the Opamp integrator.

The problem is, that the signal after the first flip flop is less close to 50:50 duty cycle than the output of a "plain vanilla comparator", even when fed with a sine (or square) wave from a freq generator. You can see it clearly on a spectrum analyser. The out of the dividing flip flop is very nice also on the spectrum analyser. I start to believe, that the secod flip flop is needed to make the whole ATC give a reasonable 50:50 square wave.

A good, simple compressor, at the front, before the LPF, is probably a good idea also.

I put an Orange Squezzer in front of it on breadboard but it did not really improve the situation (neither for the PLL nor the AN-41 doublers). But maybe would be different for guitar. Probably I should try to reduce attack and decay to attenuate the initial after plucking the string even more.

Oh, how bout 4 pitchshifters??? 
My Ibanez HD1000 will shift up one octave.   Maybe a octave & a half(?)
Feed one into the other into the other etc.   (?)

I'd love to use the EH HOG fed into a comparator as a carrier

I'd guess that the latency you mention is a result of the low cut-off in the filters.  That's going to cause a phase-shift, which will translate as a slight delay.

Maybe latency is not the correct description. It's just that the square wave is not present immediately but precceded by some ....hm....junk. It's different from the latency of the PLL. Here there is this short sliding up until it locks in. Interestingly, with the PLL the glitch or whatever is the correct expression also occurs if I hit one string and then while there is a stable square wave hit the string a second time. So I don't think a sample/hold circuit would improve it significantly.

Fair point, but have you tried it?  I would.  I always see what I can get away with.  In the integrator, make the cap smaller and the input resistor larger to minimise loading and see if it's ok.

Does the additional comparator reduce the performance or do you suggest to remove it for economical/space reasons?? Does your suggestion to reduce the integrating cap smaller and the input resistor larger also apply if the additional comparator is present??

Tweak the second comparator's trigger threshold using a scope and you're going to be able to get perfect 50:50 symmetry (well, with a test signal).

Sounds cool! Will this tweak work over a larger frequency range (30-300 Hz)??
Your suggestion and this one by MrCoffee....

Lock-in time is short if you don't overdo the fed back voltage reference filtering. You can also ac couple the triangle wave and use a 1/2 vcc for reference voltage in the second comparator.

....raise some more (dumb) questions for me:

* what is the advantage of having integrator and 2nd comparator AC coupled?
* how is the feed back from the out of the 2nd comparator into the In+ of the integrator supposed to work, and what is the advantage of this feed back thingy versus connecting the non-inv in of the integrator directly to Vref??.

I really would love to understand this

Use a more complicated 'fundamental extractor'.

More complicated than the adaptive schmidt trigger?? 

I don't have the schematic for the Boss doo-dah at hand, but if you've squared off the signal already with a Schmidt then you can divide down twice with the flip-flop.  Using a decent square to sine converter will allow you to do this.  If I recall though, the Boss doesn't present a clean enough edge for the clock input of a flip-flop.  Preceding the clock input with a 555 set up as a Schmidt works nicely.  Don't know anything about the XOR chip mentioned, but if there's a spare device in the package you might be able to use it as a buffer to the flip-flop's clock input.

The inputs to the 4013 are comparators. Shouldn't they have a nice edge?




Here's the schem for the OC-2 btw:

Will this device have individual octaves that you can dial in and out, or will they all be taken to a single output to be blended? This is something I've been interested in for awhile.

Currently I wouldn't call this thing a device Suppose you've seen the mess on the last page 
At the moment I'm struggling as you can see while trying to get the octaves up with a better performance. But basically I planned to include a mixer for all three octaves up (+dry). Also an envelope and some filtering for the square waves would be nice, but my gut feeling is that this might be the easier task (hopefully I'm not completely wrong). Using it as a carrier for the Ring Stinger would probably be fine with just a switch to select the respective octaves directly as a square wave and maybe a mixer for the square waves.


Most likely I forgot some questions but the post is already too long anyway...so I'll keep them for the next one

Regards,

Markus

[jmusser]

Well, at least you're trying. I don't have the technical know how to make it happen, but it was something I threw out as a request once upon a time, and you're the first one to run with it. Someone else was trying this awhile back, but I never heard anymore about it. The thing for most people around here, is the time to actually do what you want to. I have good intentions too, but life in general gets in the way. I'll keep tabs on this thread and see where you get with it. Thanks for taking it on. Gez has been a big help to me in the past, and knows his stuff, so you're in good hands.

[gez]

The output of the comp should look basically the same as the output of the flip flop (an assumption). So do you think the comparator will hurt ?

I'm still not sure what is meant by 'adaptive Schmidt', but if it's all that junk in combination with the first flip flop shown in the Boss schematic then, if I recall, the output symmetry of the first flip flop isn't great...not that it matters in that circuit as the second flip-flop divides down for perfect 50:50 duty cycle.

The outputs of both a comparitor and the flip flop will be square wave.  The comparator won't 'hurt', it's just a question of pruning what isn't needed/concentrating on what works best..

The problem is, that the signal after the first flip flop is less close to 50:50 duty cycle than the output of a "plain vanilla comparator", even when fed with a sine (or square) wave from a freq generator

Have you tried feeding the output of the comparator to the clock input of the flip flop to divide it down?  If it works, you'd be able to divide down twice.  I say if as the chips I use (made by ST) are really, really, picky and I've been unable to get them to work with comparator chips, even though on paper they should.  Only thing that works with them are logic chips with Schmidt inputs.  You might have better luck though...

Interestingly, with the PLL the glitch or whatever is the correct expression also occurs if I hit one string and then while there is a stable square wave hit the string a second time

They're monophonic, you can only play one note at a time.  Not sure what you meant by 'junk' instead of latency, but the fact that the strongest filter produces more of it suggests the problem is due to the low cut-off frequency you've chosen.

Does the additional comparator reduce the performance or do you suggest to remove it for economical/space reasons??

Clutter avoidance strategy! 

Does your suggestion to reduce the integrating cap smaller and the input resistor larger also apply if the additional comparator is present??

No, this measure would be to prevent loading on whatever preceeds the integrator if you decide to do away with the 'buffer' comparator.  I doubt you'd have to change values though: if the buffer comparator isn't loaded by 100k why should the ones at the output be loaded (they're more or less set up the same)?

Like I said, it's all about doing a little pruning and seeing what's really necessary (there's no point creating extra work/expenses for yourself).

Sounds cool! Will this tweak work over a larger frequency range (30-300 Hz)??

Tweaking the threshold of the comparator following the integrator will allow you to adjust each stage for 50:50 duty cycle - it compensates for any DC offsets.

what is the advantage of having integrator and 2nd comparator AC coupled?

It's another way of getting round DC offsets to get better symmetry.

More complicated than the adaptive schmidt trigger??

The person who put the 'fun' in Fundamental extractor was clearly quite mad...

[markusw]

I don't have the technical know how to make it happen,

do you have the impression I do

Someone else was trying this awhile back, but I never heard anymore about it.

hopefully he got distracted by something very special

The thing for most people around here, is the time to actually do what you want to. I have good intentions too, but life in general gets in the way.

Usually this is true for me too but at the moment luckily there is a bit more time.

Gez has been a big help to me in the past, and knows his stuff, so you're in good hands.

Also he seems to be very patient with my questions Thanks again btw for answering!

m still not sure what is meant by ‘adaptive Schmidt’, but if it’s all that junk in combination with the first flip flop shown in the Boss schematic

yes, at least it is called this way in the EH Micro synth schem

if I recall, the output symmetry of the first flip flop isn’t great…not that it matters in that circuit as the second flip-flop divides down for perfect 50:50 duty cycle.

exactly what I observed on scope and on spectrum analyser

The comparator won’t ‘hurt’, it’s just a question of pruning what isn’t needed/concentrating on what works best..

At the moment I try to focus on getting it working Later on I'll start to tweak to reduce the circuit size (if I ever get there).

Have you tried feeding the output of the comparator to the clock input of the flip flop to divide it down?  If it works, you’d be able to divide down twice.

The two ocatves down aiming at a polyphonic synth?? No I haven't tried yet. But IMHO it actually should work. The two comparators in the adaptive schmidt are also able to trigger the 4013...or am I missing something ?  Also the comparators are able to trigger the 4070 XOR gate or am I missing something again ?

Not sure what you meant by ‘junk’ instead of latency,

Think it's time to give you a sample to listen to. Would make it easier

They’re monophonic, you can only play one note at a time.

I know I meant if I pluck one note, wait 1 or two seconds and pluck the same note again I also get that that noise/latency/junk (I definitely will have to record it!)

but the fact that the strongest filter produces more of it suggests the problem is due to the low cut-off frequency you've chosen.

Sounds logic.  IIRC correctly mentioned some posts ago that higher order filters tend to put out their Fc when just fed with a spike. So this would be in accordance with your theory.

Clutter avoidance strategy!

Had to check clutter in the dictionary. Hopefully, I recall the word the next time I need it.

(there’s no point creating extra work/expenses for yourself).

Thats true. But at the moment removing some not-needed comparators would mean extra work

It’s another way of getting round DC offsets to get better symmetry.

So which way would you prefer aiming at optimal peformance?

I know we had this one but may I dare to ask again?

how is the feed back from the out of the 2nd comparator into the In+ of the integrator supposed to work, and what is the advantage of this feed back thingy versus connecting the non-inv in of the integrator directly to Vref??.

The person who put the ‘fun’ in Fundamental extractor was clearly quite mad…

It's just called this way because Pitadamental extractor is harder to pronounce.


OK I will try to post a mp3 ASOP to make everything easier.

Regards,

Markus

[markusw]

Here it is:

http://www.diystompboxes.com/DIYFiles/up/octave_test_11-05-06.mp3

Sorry for loading this garbage to your server aron!

It's split in several parts. In between there are short breaks. In each parts I play always the same notes (on bass): low A, 5th (low E on guitar), octave, 5th above octave. Each tone is plucked once, and then after a while several times in a row. The signal goes either from the 4040 divider in case of the PLL, or the last comparator in case of the doublers (via a cap) into my USB soundcard.

The parts:

* for comparison the first one ist just thorugh the 6db bandpass into the first comparator

* 12 db bandpass, schmidt trigger, both flip flops into PLL : 2 ocatves up
* as last one but 3 octaves up
* as last one but 4 ocatves up

* 12 db bandpass into 3 doublers in series, 1 ocatve up
* as last one but 2 ocatves up
* as last one but 3 ocatves up


* 6 db bandpass into 3 doublers in series, 1 ocatve up
* as last one but 2 ocatves up
* as last one but 3 ocatves up

Interestingly, there's also some kind of glitch in the first part. Didn't check this before.

I know it's hard stuff to listen too especially since it's quite long.

Regards,

Markus

[gez]

The two comparators in the adaptive schmidt are also able to trigger the 4013...or am I missing something ? Also the comparators are able to trigger the 4070 XOR gate or am I missing something again ?

I was wondering if you could get one of those comparator chips (not an op-amp wired as a comparator) to drive the clock input of the first flip-flop wired up to divide down as this would allow you to use the second half of the flip-flop to divide down again.  The comparator chip(s) might be able to trigger the clock input directly and should be able to trigger the XOR no problem, if that's what you meant.

Re symmetry, it all depends how you've done things.  Trimpot would suit a single supply better.

how is the feed back from the out of the 2nd comparator into the In+ of the integrator supposed to work, and what is the advantage of this feed back thingy versus connecting the non-inv in of the integrator directly to Vref??.

I'm guessing here, but I'd say C2 (the .01u cap), in conjunction with the 100k resistor, averages the output of the second comparator.  If duty cycle strays from perfect 50:50 the +ve input is going to be pulled up/down accordingly which in turn provides feedback to the integrator, correcting its output so that the comparator triggers to give a symmetrical output once more.  I should imagine it would take a few cycles to do this though.  The cap value is for a frequency range that starts well above the lowest note of your bass, so you'd need to make the cap bigger for feedback to work properly.  Doing this might make the feedback sluggish though...who knows (all this is guess work on my part)?

The samples sound pretty messy, as though the slightest noise is triggering whatever it is you're using to do the sine-to-square conversion.  Need to have a decent threshold so that string glitch, hum etc doesn't set things off.  Also, it sounds as though the signal might not be strong enough, hence notes being all over the place once plucked.  I'm guessing that a lot of that would be due to the cut-off of the filter being so low.  With octave up circuits, you don't notice 'yodelling' to the extent that you do with octave down circuits.  It might be an idea to bypass the filter just to see if this improves matters.  If so, then you'd need to raise the threshold of whatever filter you end up using.

[toneman]

sounds = Beee  Zarrrrrr !!!   
some of the yodels & computer music is neat!!! 
what's that 60hz at the beginning.... 
Thought U guys used 50Hz??     
But seriously Folks   
Again, get a good fundamental extraction.   
Then multiply UP *past* what U want...  i.e.  x8.
Dosen't matter if it's pulses or what ALA it drives the next stage.   
THEN  divide down....  2  then 2 then 2 = back to fundamental.   
ALL these "divide downs" will  be 50/50 SQ waves.
Then mix individualy.....stir in an enveloped filter.
Not shaken....stirred.

[markusw]

Hey Gez,

thanks once more for your patience.

I was wondering if you could get one of those comparator chips (not an op-amp wired as a comparator) to drive the clock input of the first flip-flop wired up to divide down as this would allow you to use the second half of the flip-flop to divide down again.

Ah, now I understand. Will try this weekend and report.

Re symmetry, it all depends how you’ve done things.  Trimpot would suit a single supply better.

OK, I will try both. Does it make sense to AC couple and additionally use a trim for the 2nd comparator's reference?

Re the feedback thingy: I believe I understand. Your explanation fits very well to the Spice sims I did. I just couldn't interpret them. In my hands decreasing C2 reduced symmetry. I didn't try to increase it though but I fear that it will get too sluggish. Anyway, I'll give it a try.

Need to have a decent threshold so that string glitch, hum etc doesn’t set things off.

Would it work to run the signal through a rectifier, low pass filter it, use the envelope to turn on or off a comparator that subsequently activates or inactivates a 4066 switch (which turns the audio signal going into the first comparator of the first doubler on/off)??

It might be an idea to bypass the filter just to see if this improves matters.  If so, then you’d need to raise the threshold of whatever filter you end up using.

One more option I'll try.

Hey T,

what's that 60hz at the beginning.... 
Thought U guys used 50Hz??

Ahm, we HAVE 50 Hz. This is the great sound of an A on a detuned Jazz bass run through a comparator

THEN  divide down....  2  then 2 then 2 = back to fundamental.   
ALL these "divide downs" will  be 50/50 SQ waves.

My theory: with every doubling stage you get some assymetry. If you now feed the let's say 3 octaves up into a dividing by 2 flip flop you will ideally get the symmetry of the 2 octave signal (but not a perfect 50:50 square wave).

Regards,

Markus

[gez]

Does it make sense to AC couple and additionally use a trim for the 2nd comparator's reference?

I'd just use one or the other.  If you have resistors with a tight tolerance then things should be ok.  It's one of those things you're going to have to try and see what works.

Re the feedback thingy: I believe I understand. Your explanation fits very well to the Spice sims I did. I just couldn't interpret them. In my hands decreasing C2 reduced symmetry. I didn't try to increase it though but I fear that it will get too sluggish

If you're using the cap value shown in the app note it might account for some of your problems.  It would need to be bigger for the frequencies at which the doubler is being used.  It should provide an average DC at the +ve input of the integrator, not an attenuated, distorted version of the output.

Would it work to run the signal through a rectifier, low pass filter it, use the envelope to turn on or off a comparator that subsequently activates or inactivates a 4066 switch (which turns the audio signal going into the first comparator of the first doubler on/off)??

IIR, I mentioned various ways to do this earlier in this thread.  What you've outlined would work very well with an OTA so that you have some dynamics.  It can work with simple on/off switching too, but the comparator used needs some Schmidt action to avoid miss-triggering.

If you now feed the let's say 3 octaves up into a dividing by 2 flip flop you will ideally get the symmetry of the 2 octave signal (but not a perfect 50:50 square wave).

Actually, dividing down with a flip-flop gives you perfect 50:50 symmetry.  Getting a glitch-free signal many octaves up in the first place in order to divide down is the problem though (if it yodels and splutters then so does everything else)...it's all swings and roundabouts. 

[markusw]

Some results:

* bass into 6db bandpass (Fc=30 Hz) into comparator into dividing by 2 flip flop gives an almost perfect 50:50 symmtry

* bass into 6db bandpass (Fc=30 Hz) into 2 doublers in series into dividing by 2 flip flop gives about the same result as just one doubler without the flip flop (but definitely not a 50:50 symmetry)

* bass into 6db bandpass (Fc=30 Hz) into 3 doublers in series into dividing by 2 flip flop is actually worse than just two doublers without the flip flop

If you’re using the cap value shown in the app note it might account for some of your problems.

The cap values are indicated in the block diagram (e.g. for the first doubler:  1µ for C2 and 100n for C4)

It can work with simple on/off switching too, but the comparator used needs some Schmidt action to avoid miss-triggering.

Would it be a good idea to set the comparator's reference slightly below bias (as you suggested for that mono stable approach) to avoid misstriggering?

Also, it sounds as though the signal might not be strong enough,

Acutally, I increase the gain of the unput gain stage until it nearly clips. So I believe the signal should be OK. Maybe all this wires crossing each other could also cause troubles?

Actually, dividing down with a flip-flop gives you perfect 50:50 symmetry.

I suppose I will have to tweak the doublers for more symmetry.

This btw is the scope trace of the signal at the first integrator's out (fed with 50 Hz). Interestingly, these assymetry was also predicted by Spice. I believe, that this is one of the points that causes issues with running several of these doublers in series.



This is after the 2nd integrator:



Any ideas how to get rid of this assymetry??

Regards,

Markus