3 or 4 octave up square wave ??

[gez]

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

Sorry, didn't see that.

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?

Only if you use a monostable as outlined, otherwise you'll still get miss-triggering.  This comparator would only be used for switching, not to feed your doubler circuit.

Re asymmetry of the triangle from the integrators, there shouldn't be any.  You should have nice straight sides.  Distortion can be caused when AC coupling stages with caps that are too small, using amps with poor slew rate (though I doubt that would be a problem here), not having a nice clean square wave to present to the integrator.  It might even be a result of the feedback setup (if the resistor and cap used aren't decoupling the outputs of the comparators properly).  In your simulator, try referencing the comparators to half supply instead of setting them up to feedback to the integrator.  Does this help?

[markusw]

Only if you use a monostable as outlined, otherwise you'll still get miss-triggering.  This comparator would only be used for switching, not to feed your doubler circuit.

Hm, could I somehow use the monostable to on/off switch the signal to the first comparator of the doubler??

It might even be a result of the feedback setup (if the resistor and cap used aren't decoupling the outputs of the comparators properly).  In your simulator, try referencing the comparators to half supply instead of setting them up to feedback to the integrator.  Does this help?

Seems you wre right once more Tried it both in LTSpice  to give V+/2 as reference for the integrator and it  clearly  gives a nicer triangle wave. On breadboard the triangle wave is now cut off at the top??? Does it mean that I have to set the reference for the intergrator with a trim??

Markus

[gez]

Hm, could I somehow use the monostable to on/off switch the signal to the first comparator of the doubler??

With a CMOS switch, sure.  If you mean disable the comparator...possibly.  A hack/poor way of doing this would be to set up the output of a retriggerable monostable so that it forms one of the 'anchor points' for the divider that biases the comparator feeding the doubler, ie it acts as either ground or V+.  When the monostable finally flips, its output goes hi/low (depends on how you've got it set up) and pulls the -ve input of the comparator away from half supply towards one of the rails (again, which one it gets pulled to depends on how you set up the monostable).  This will cause the output of the comparator to stay in one state.  The question is, do the comparator/op-amp (if you're using them instead) chips allow the inputs to be pulled to at least one of the rails?  You'd need to check data sheets.

Tried it both in LTSpice to give V+/2 as reference for the integrator and it clearly gives a nicer triangle wave. On breadboard the triangle wave is now cut off at the top??? Does it mean that I have to set the reference for the intergrator with a trim??

Glad you understood me (accidentally typed comparator instead of integrator...sorry).  Back to the question.  It depends.  The clipping might be due to asymmetrical output swing of your op-amp.  Again, check the data sheet.  If you don't have much upward swing then comparators and integrators all need to be referenced to a voltage slightly below half supply.  Either that, or use CMOS output op-amps. 

[gez]

PS  Increasing either the input resistor or the cap in the feedback loop of the integrator will reduce the amplitude of the triangle.  If you can get away with it, this might be a better option than to change the bias.

Something just occurred to me.  Even when you've got the trigger threshold sorted out so that the thing doesn't miss-trigger at the slightest breeze on your strings, you still might have a problem.  When the circuit is lying dormant, the first comparator's output will either be hi or low (depending on how things are set up) so eventually the output of the integrator which follows will ramp up/down until it hits the rails.  When this happens, feedback to the -ve input stops so the cap in the feedback loop continues to ramp up/down (though exponentially from now on) pulling the input with it to one of the rails.  When operation starts again (input signal above the threshold set) then it will take a few cycles until the -ve input is back at half supply.  This might account for the 'latency' you described:  The octave effect won't kick in immediately.

If this is the case, then a couple of back to back Ge diodes between the -ve and +ve inputs of the integrator should sort things out...or at least help.

[markusw]

Thanks a lot for your invaluable input!

With a CMOS switch, sure.

So a 4066 for switch would be fine?

The data sheet for the LM311 I'm using for the comparators says +/- 15 V input at +/-15V supply. But it also says common mode voltage range is typically 13.8V down to -14.7V. So pulling the -ve input with the retriggerable monostable down should be OK? But does it mean that the output swing of the comparator isn't symmetric ??

The clipping might be due to asymmetrical output swing of your op-amp.

I checked it again more carefully having the first integrator's +ve inpiut tied to V/2 instead to the 100k feedback and it seems even more wierd now:  sometimes the triangle is cut off at the top, sometimes at the bottom. Even when playing the same note. Vref is rock stable at +ve input. ?? The data sheet for the MC3317 opamp says: Large Output Voltage Swing: –14.2 V to +14.2 V (with ±15 V Supplies). Looking more carefully now, it also says 4.3V to 0.05V at +5V/0 supply. So I assume that at +12V/0 supply it will be assymetric too.

If you don’t have much upward swing then comparators and integrators all need to be referenced to a voltage slightly below half supply.

At the moment I have Vref provided by two 100k as a voltage divider buffered by an opamp. This Vref is used as a reference for all comparators.  Is this OK?

If yes, I theoretically could replace the two 100k by a trimpot to adjust Vref?  Or would it be better to do it with individual trim pots for each comparator?
The funny thing is, that if the +ve input is fed by the 100k feedback resistor no triangle clipping occurs  ?? This I definitely do not get.

From the AN-41:
A comparator is used to shape the input
signal and feed it to an integrator. The shaping is required
because the input to the integrator must swing between the
supply voltage and ground to preserve symmetry in the
output waveform.

But at 5V supply like in the AN-41 the output of the LM111 for sure won't swing rail to rail??


Does this all mean I have to use a rai-to-rail opamp?? And will the assymetrical output swing of the LM311 cause problems?? If, yes I actually could use rail to rail opamps for comparators and integartors.

Increasing either the input resistor or the cap in the feedback loop of the integrator will reduce the amplitude of the triangle.

Will give it a try. Interestingly, the triangle is clipped also higher notes with much lower amplitude.

Something just occurred to me.  Even when you’ve got the trigger threshold sorted out so that the thing doesn’t miss-trigger at the slightest breeze on your strings, you still might have a problem.  When the circuit is lying dormant, the first comparator's output will either be hi or low (depending on how things are set up) so eventually the output of the integrator following it will ramp up/down until it hits the rails.  When this happens, feedback to the -ve input stops so the cap in the feedback loop continues to ramp up/down (though exponentially from now on) pulling the input with it to one of the rails.  When operation starts again (input signal above the threshold set) then it will take a few cycles until the -ve input is back at half supply.  This might account for the ‘latency’ you described:  The octave effect won’t kick in immediately.

I believe you hit the bull's eye again

This is exactly what I observed in the sims. I just though that it is an artifact.

If this is the case, then a couple of back to back Ge diodes between the -ve and +ve inputs of the integrator should sort things out…or at least help.

You mean like you'd protect an opamp from high voltage spikes (in parallel but different orientation)? Will give it a try. How exactly are the didodes supposed to help??


Regards,

Markus

[gez]

So a 4066 for switch would be fine?

Yes

The data sheet for the LM311 I'm using for the comparators says +/- 15 V input at +/-15V supply. But it also says common mode voltage range is typically 13.8V down to -14.7V. So pulling the -ve input with the retriggerable monostable down should be OK? But does it mean that the output swing of the comparator isn't symmetric ??

The data sheet should have figures for output swing, check it again.

Will the comparator chip work properly or be damaged with an input directly connected to a monostable's output which switches rail-to-rail?  No idea, I don't use comparator chips any more.  However, as it does switch switch rail-to-rail and common mode voltage range is within this limit, connection would best be done via a resistor (EDIT: If you're using the divider at the output of the monostable idea that I outlined, then that resistor wouldn't be needed...sorry all this is getting a bit confusing!    All this is accademic if you use a CMOS switch.

I checked it again more carefully having the first integrator's +ve inpiut tied to V/2 instead to the 100k feedback and it seems even more wierd now: sometimes the triangle is cut off at the top, sometimes at the bottom. Even when playing the same note. Vref is rock stable at +ve input. ?? The data sheet for the MC3317 opamp says: Large Output Voltage Swing: –14.2 V to +14.2 V (with ±15 V Supplies). Looking more carefully now, it also says 4.3V to 0.05V at +5V/0 supply. So I assume that at +12V/0 supply it will be assymetric too.

It sounds as though the output is more or less symmetrical and should be fine.  V ref should be rock solid though, but your schematic is too sketchy to see what is going wrong.  The + input of the integrator should be referenced to half supply - just to get things clear, you're using a 12V single supply...yes? - and C2 & R5 would be dispensed with.  Yes, an op-amp used for a Vref is ideal.

The funny thing is, that if the +ve input is fed by the 100k feedback resistor no triangle clipping occurs ?? This I definitely do not get.

The amplitude of the triangle is going to vary with frequency (reducing as frequency increases).  It sounds as though you need to reduce it quite a bit so that you avoid clipping.  The feedback that you were using originally might have been reducing amplitude, who knows, but you clearly have a problem now so need to compensate.  Just follow the advice I gave in one of my previous posts about reducing it (scope it and check for clipping with the lowest frequency you intend to run through the doubler stages).

You mean like you'd protect an opamp from high voltage spikes (in parallel but different orientation)? Will give it a try. How exactly are the didodes supposed to help??

A couple of diodes in parallel but different orientation -'back-to-back' - between the + and - input of the integrator (pins 2&3 in the app notes).  The - input can only be pulled up/down by the cap (as it ramps up/down towards one of the rails when idle) within a  few hundred millivolts of half supply: the diodes clamp the input which means when the integrator is brought into use with an incoming signal, the cap doesn't take so long to return to half supply for the integrator to work.  Ge have lowest threshold, therefore it will take less time for the integrator to 'recover'.

[markusw]

Hi Gez,

thanks for your patience!

The data sheet should have figures for output swing, check it again.

I checked the data sheets for the LM311 form TI and NS but this is the only figure I found that makes any sense to me 



Unfortunately, the data sheets don't say anything about output voltage swing.

All this is accademic if you use a CMOS switch.

I think I'll go with a CMOS switch

It sounds as though the output is more or less symmetrical and should be fine.

Do you think an output swing from 0.05V to 4.3V at 5V supply for the MC33171 opamp  is symmetric enough?

you’re using a 12V single supply…yes? - and C2 & R5 would be dispensed with.  Yes, an op-amp used for a Vref is ideal.

Yup, 12V supply and Vref from an opamp

The amplitude of the triangle is going to vary with frequency (reducing as frequency increases).  It sounds as though you need to reduce it quite a bit so that you avoid clipping.  The feedback that you were using originally might have been reducing amplitude, who knows, but you clearly have a problem now so need to compensate.  Just follow the advice I gave in one of my previous posts about reducing it (scope it and check for clipping with the lowest frequency you intend to run through the doubler stages).

With the integrator's + input tied to Vref I tried everything you suggested: increasing the input resistor to the integrator, increasing the integrating cap.....nothing really helped. The only way to somehow improve the situation was to tweak the offset trim of the preceeding comparator. However, the wierd thing is: I turn on the signal generator, the triangle is clipped, I tweak the offset trim, the triangle is fine, I turn the signal generator off and on, bingo...the triangle is clipped again. In contrast, with the integrator's + input commected to the 100k feedback there is absolutely no clipping of the triangle.

So at the moment I believe that the issues with the clipped triangle are somehow connected to either voltage offset and/or assymetric output swing from the comparator. So maybe it would  be the best to replace all comparators and integrators with the same type of  rail to rail opamp? Do you have any recommendation for a cheap one?

A couple of diodes in parallel but different orientation -‘back-to-back’ - between the + and - input of the integrator (pins 2&3 in the app notes).  The - input can only be pulled up/down by the cap (as it ramps up/down towards one of the rails when idle) within a  few hundred millivolts of half supply: the diodes clamp the input which means when the integrator is brought into use with an incoming signal, the cap doesn’t take so long to return to half supply for the integrator to work.  Ge have lowest threshold, therefore it will take less time for the integrator to 'recover'.

Thanks for the explanation!
I  tried to add the two Ge diodes. With the integrators + in connected to Vref the diodes seem to mess up everything. With the feedback version there may be a slight improvement but I can't say for sure. Maybe it will be easier to hear the difference once I've added a gate.

EDIT: I know tried to run a software synth (set up as a square wave oscillator) into the whole mess and now the improvement with the Ge diodes is clear. Also it is clear that this latency until the ocatve cuts in is not present if I play a note and change to another one (e.g. a 3rd or 5th above). Latency is much lower than with bass anyway. Therefore, I believe it might help to compress the bass signal in order to reduce the strong initial signal immidiately after plucking a string.

Regards,

Markus

[gez]

Sorry Mark, things are getting a little muddled in my mind.  I was forgetting the LM311 was a comparator chip.  Presumably 'saturation voltage' in the data sheet is indicative of how low the output can go?  If so, then the figures given look reasonable enough. 

Do you think an output swing from 0.05V to 4.3V at 5V supply for the MC33171 opamp is symmetric enough?

Symmetry isn't really the issue (it's good enough), amplitude is the problem.  You mentioned that when you tried increasing the cap/input resistor of the integrator, nothing happened.  Make either one (or both) much bigger.  You need to get amplitude down to get the circuit working properly.  It will - or rather should (I'm assuming everything is connected up correctly)- happen. 

So at the moment I believe that the issues with the clipped triangle are somehow connected to either voltage offset and/or assymetric output swing from the comparator.

Perhaps, but if you make the amplitude of the triangle smaller, none of this will be an issue and there'll be no need to replace anything: the output will swing within its limits and clipping won't occur.

So maybe it would be the best to replace all comparators and integrators with the same type of rail to rail opamp? Do you have any recommendation for a cheap one?

ICL7621, but try the above first.  The circuit should work with what you're using.  Focus on the integrator for the moment, get this sorted out.

I tried to add the two Ge diodes. With the integrators + in connected to Vref the diodes seem to mess up everything

That shouldn't be the case.  It might be that it isn't working properly because the output of the integrator is clipping; reign it in and it should work fine.  As I keep saying, get the amplitude of the triangle down so that there's no chance of clipping.

Therefore, I believe it might help to compress the bass signal in order to reduce the strong initial signal immidiately after plucking a string.

Don't really understand you.  I'm presuming the signal generator is of constant amplitude and fires up immediately, so why would you want to reduce the attack of your bass notes?  If you meant drag everything up that follows the initial attack so that you get as even an amplitude as possible across the decay of a note, then yes that will help.  But then we get back to the low cut off of your filters again, I'm sure this isn't helping.

If you're still getting problems after trying all the above, it might be time to start thinking of drawing up a more detailed schematic. 

[TELEFUNKON]

sorry for the disturbance:

anybody ever tried doubler`s feedback - eedback - dback - ack - k?

[gez]

sorry for the disturbance:

anybody ever tried doubler`s feedback - eedback - dback - ack - k?

A pleasure to see you Mr Funk-On!  Not sure what you mean.  Care to elaborate (general will do)? 

[TELEFUNKON]

feeding the (frequency-doubled) output of (any) frequency-doubler
back to the input of said frequency doubler,
to have the (once) frequency-doubled output
frequency-doubled again (and again - and again?)?
(of course after having thought about Mr. Barkhausen, beforehand!).

[markusw]

Hi Gez,

ymmetry isn’t really the issue (it's good enough), amplitude is the problem.  You mentioned that when you tried increasing the cap/input resistor of the integrator, nothing happened.  Make either one (or both) much bigger.  You need to get amplitude down to get the circuit working properly.  It will - or rather should (I’m assuming everything is connected up correctly)- happen.

I will try to make them much bigger.

Focus on the integrator for the moment, get this sorted out.

I will do! Thanks for guiding me through this project!!

Don't really understand you.  I'm presuming the signal generator is of constant amplitude and fires up immediately, so why would you want to reduce the attack of your bass notes?  If you meant drag everything up that follows the initial attack so that you get as even an amplitude as possible across the decay of a note, then yes that will help.  But then we get back to the low cut off of your filters again, I’m sure this isn’t helping.

Sorry for the explanation. I meant to "drag everything up that follows the initial attack".

If you're still getting problems after trying all the above, it might be time to start thinking of drawing up a more detailed schematic.

Would probably be a good idea anyway

Hi Funk-on,

feeding the (frequency-doubled) output of (any) frequency-doubler
back to the input of said frequency doubler,
to have the (once) frequency-doubled output
frequency-doubled again (and again - and again?)?
(of course after having thought about Mr. Barkhausen, beforehand!).

Wouldn't this circuit oscillate like hell?? Will have to google for Mr. Barkhausens ideas.

Regards,

Markus

[markusw]

Some more results

At the moment I'm just playing with the first doubler (preceeded by by the 12dB bp filter). The circuit is fed by a 100 Hz sine wave. Supply is 11.6 V.

If the +ve input of the first integrator is connected to the 100k feedback I get a triangle wave as supposed. If I measure DC with a true RMS DMM at the integrator's out it gives me 5.73 V while the circuit is fed with the sine. When the input to the whole circuit is grounded (i.e.the comparators are not triggered and the output of the circuit is quiet) I get about the same DC V at the output of the integrator. The ouput of the preceeding comparator is high (11.49V) as is the -ve in of the integrator.

When the +ve input of the first integrator is connected to Vref, it's entirely different. With the 100 Hz sine the output is at about 0.68V DC, with the input grounded I get 85 mV DC.
Increasing the integrating cap to 570n gave about 180 mV DC at the integrator's out when fed with the sine and the triangle's ampliude clearly goes down. The subsequent comparator never triggers if  the +ve input of the first integrator is connected to Vref regardless if the integrating cap is 100n or 570n, thus the XOR gate just gives the fundamental.

With 100n as integrating cap the tringle's amplitude btw is about 1/10th of the comparator's square wave regardless whether the integrator's  +ve in is connected to the 100k or Vref.

(I’m assuming everything is connected up correctly)-

Actually, it was my first asumption   that there is NOT every thing connected properly but in the meantime I checked everything several times and it definitely works with the integrator wired like in the AN-41. Changing this one connection is sufficient to make the DC voltage at the integrator's out drop  from Vref to 0.68 V. This is about the same result I get in my Spice sims btw.

Hopefully, these measurements are somwhow useful and are not just confusing.


Regards,

Markus

[gez]

DMM measurements are pretty meaningless really as you can't tell (well, I can't) if there's any distortion of the waveform.  Do you have a scope?

You don't want to reduce the amplitude of the triangle till it's so small it no longer triggers the following comparator.  All you needed to do was reduce it slightly so that you don't get clipping when a test signal set at the lowest frequency you intend to run thru the circuit.

[markusw]

Hey Gez,

thanks for your reply!

DMM measurements are pretty meaningless really as you can't tell (well, I can't) if there's any distortion of the waveform.  Do you have a scope?

I checked the tiangle's amplitude with my software scope. It's clearly about 1/10th of the comparato's square wave amplitude with 100n integrating cap regardless which way the integrator's +ve in is wired. So I'm pretty sure that it's amplitude is not the reason for clipping. With the DMM I just measured DC voltages which I suppose should give me the DC voltage the various waves are centered on. At least it clearly gave Vref when I measured DC at the comparator's out fed with a sine wave. Please correct me if I'm wrong!

All you needed to do was reduce it slightly so that you don't get clipping when a test signal set at the lowest frequency you intend to run thru the circuit.

The triangle's amplitude is the same with the +ve in of the integrator connected to Vref but it seems to be centered at about 0.68 V instead at Vref, which IMHO is an easy explanation for the clipping observed. Therefore, I' now quite sure that it's some kind of biasing issue. This morning while browsing through "The Art of Electronics" I found that obviously integrators like in the AN-41 when connected with the +ve in to Vref tend to drift towards on of the rails. There were some solutions given for this issue (e.g. adjusting offset, putting a large resistor in parallel to the integrating cap, some switching thingies I don't really remember..)  Will check more carefully this evening and post the details.

In conclusion I'm pretty sure that connecting the +ve input to the 100k feedback also somehow avoids this issue.

Regards,

Markus

[gez]

I checked the tiangle's amplitude with my software scope. It's clearly about 1/10th of the comparato's square wave amplitude

This amplitude is too small IMO.  Negative feedback ensures the - input of the integrator is held at the same voltage as its + input.  With Vref at half supply, whenever the comparator goes high a voltage equivalent to half the supply will develop across the integrator's input resistor (assuming rail-to-rail output from the comparator's output).  Because of the virtual earth effect at the - input of the integrator this voltage will be constant, so current is constant too and the cap will ramp down (inverting amp) in a linear fashion for as long as the comparator is high.  When the comparator goes low, the same principals apply, only the cap ramps up. 

The value of the input resistor and cap in the feedback loop determine charge time for the cap, which also determines how far the cap ramps up/down, so it's easy to see that frequency will determine the amplitude of the resultant waveform: the period of time that the comparator is high/low determines how far the cap ramps down/up so higher frequencies (less time) = smaller amplitude.

Ideally, you want the triangle to have the largest amplitude possible in order to get a wide range of frequencies: there comes a point as frequency increases where the amplitude is so small that it no longer triggers the following comparator if it has a DC offset (wave is always above/below the trigger threshold).  So, a 100 Hz test signal isn't appropriate.  What you need is a test signal equal in frequency to that of the lowest note of your instrument.  For a bass, that's going to be roughly 40 Hz for a low E.  With this frequency as you test signal, adjust the integrator (trim pot for input resistor would be ideal) until the output is a large as you can possibly make it before clipping sets in. 

The triangle's amplitude is the same with the +ve in of the integrator connected to Vref but it seems to be centered at about 0.68 V instead at Vref, which IMHO is an easy explanation for the clipping observed

Measure the output of Vref (I think you said you were using an op-amp).  It should be half supply.  If it is, then the integrator should be centred round half supply too (with a test signal).  If it isn't, then you need to double check everything is wired up correctly.

This morning while browsing through "The Art of Electronics" I found that obviously integrators like in the AN-41 when connected with the +ve in to Vref tend to drift towards on of the rails

That's why I suggested using the diodes.  Read thru what I wrote again and it should make more sense now.

[markusw]

Hey Gez,

thanks a lot for your explanations and once more for your patience!!

Read thru what I wrote again and it should make more sense now.

I definitely will!

Regards,

Markus

[markusw]

Hey Gez,

here's a pic of the integrator's out compared to the out of the doubler (fed with 31 Hz which is the lowest note on my 5-string). I believe the amplitude should be fine. Maybe I can reduce the 100n a bit but I believe for now it should be OK (47n already is too small for 31 Hz).



Sorry for the quality, photobucket decided to downsize the 250kb original.

However, I found another issue that might have to do with the already quite large circuit conglomerate on my breadboard
With the +ve input of the integrator connected to Vref I observed the following:

* the DC voltage at which the triangle is centered is dependent on the sine input signal strenght (which actually shouldn't be the case as long as the comparator is triggered  I suppose)
* also Vref measured at the +ve input of the integrator slightly varies (a few mV) with the input signal strenght
* for a given input signal strenght I can adjust the DC voltage at which the triangle is centered by tweaking the offset pots I added to the 1st comparator and the integrator thus making the doubler work.

There are quite a lot of ICs connected to one Vref provided by one opamp: several opamps, and about 10 comaprators. Do you believe that this number is already too much for one Vref opamp? Also all these Vref connections are daisy chained across the breadboard although I'm sure that all these connections are fine. Could this be a reason for the slight instability of Vref??

Regards,

Markus

[gez]

In the image you've posted there's a bit of distortion of both waveforms.  It looks as though bass response is somewhat lacking.  Probably nothing to worry about, I should think this is an input cap on the computer's soundcard causing the problem.   If you have one of those plug in scopes, does it have a DC option?  Anyway, should be ok, what you've shown is often what you see when a low frequency signal is AC coupled.

If I recall, you said you were using an op-amp as Vref.  Did you decouple the divider for this amp with a cap?

[markusw]

If you have one of those plug in scopes, does it have a DC option?

I'm using my Tascam US-122 USB interface. So unfortunately there is no DC option.

If I recall, you said you were using an op-amp as Vref.  Did you decouple the divider for this amp with a cap?

There is a massive cap after the Vref opamp (1000µ because I had it home). Will try to add another one directly at the divider.

Regards,

Markus