Why using Chopper circuit in H&MM Harmony Generator?

[Mr. Lime]

Hi guys!

I ordered a Harmony Generator clone PCB from circuitbenders uk and aksed myself why there's a chopper at the output from the octave divider/multiplier?
There are other circuits like Schumann PLL or some CD4046 octaver that use a direct output of the divider and sum it with other outputs.
Couldn't someone take the signal directly from the CD4040 or at least the seperate octavces and add interstage dividers and summing amp to have a better control like the Barton Waveshaper below does? What's the point of the chopper (IC9B) here?

For my understanding the chopper is a very high frequently switching device but how does it differ from a tremolo beside the rate and the given squarewave?

Here are the schematics and articles:


http://hammer.ampage.org/files/EMMHarmoGen.PDF

Thanks a lot!

[anotherjim]

I'm not sure it's a chopper. I think it's a sample & hold. When it "thinks" it should have locked on the right pitch, it switches the VCO to run only according to the control voltage left on the loop filter capacitor C11. Primarily IC9a switches off preventing the PC2 output changing the frequency. There a few more bits and pieces I think are there to smooth transitions when it decides to move to a new pitch.

[Rob Strand]

I think the idea is C5 tracks the input envelope, ie. the instantaneous input level.  The digital circuits lose the signal envelope and only generate signals with the intended frequencies.  The chopper chops the envelope so the heard frequencies are in faster digital signals and the instantaneous level is the level on C5.  I'm surprised they don't smooth the output of the chopper with a filter at least a small amount.

[anotherjim]

I still don't see any choppers
It's just an envelope (rectifier to C7) derived note-on/note-off gate switching. Thresholds set by the IC4 Schmitt trigger gates. Note the common C control pins to the other switches from the envelope gate.
The Harmonic output is muted by IC9b at note-off.

[Mr. Lime]

Thanks for the fast response!

Alright I just read about the chopper in the kit describtion but I'm looking forward for other theories!
The essential question for me is if I should keep IC9B or ditch it and use IC3 as summing amp for the octave outputs like in the second schematic.
Does a guitar signal profit from IC9B?
Otherwise how about spst switches for each octave output before going into IC9B or can there only be one octave at a time to trigger the ,,chopper"?

[anotherjim]

I see the chopper! Well, kind of. It's actually a "swing VCA", since the control is simply setting the maximum peak swing of the output signal.
Below is the more usual form, but you can see the BJT in this isn't expected to be linear but just a switch. The input signal is "chopping" the envelope - or the envelope is defining the swing of the output.


Yes, you could say the harmony signal is "chopping" the envelope.
Clever use of a spare CMOS switch, but its control input is limited to logic level input, unlike the BJT version.
The envelope on C5 gives the harmony something like the input signal envelope.
The voltage on C5 provides the peak positive level of the harmony signal. The Peak negative level is effectively fixed at 0v.
As the harmony square wave switches logic high, the harmony signal leaving the switch peaks at whatever C5 voltage is. When harmony switches low, the signal "drifts" (since switch IC9d is off) to 0v. Then the harmony output level is attenuated 1/10th by R6/R7 ratio.

...can there only be one octave at a time to trigger the ,,chopper"?

The problem if you mix divider outputs is that some switching of the logic levels will coincide perfectly and not quite give the expected audible result of the different frequency cycles in the total output. However, if you mix octave related outputs you will hear something of each, but bear in mind the "chopper" only responds to logic levels.
For example, if logic high is 9v and low is 0v, then 2 counter outputs simply mixed together will at times produce 4.5v which the CMOS switch just won't know what to do with.
You could add pulse generators to turn the square waves into shortened pulses which don't overlap, but it can be tricky to tune that since the pulse widths will be fixed whatever the period of the counter outputs are and only sound right over a limited musical range.

[Rob Strand]

Here's a slightly idealize simulation of the chopper and envelope circuit.   
Note that the chopper is unidirectional and relies on the DC offset removal by the output cap.   
A cleaner output would result from a +/- envelope and alternate chopping; there's a few spins on this theme.

The signal is three guitar notes on the upper strings.
The chopping signal is 1kHz.  I haven't tried to harmonize it with the guitar.



Zoomed into mid-section:

[Mr. Lime]

The problem if you mix divider outputs is that some switching of the logic levels will coincide perfectly and not quite give the expected audible result of the different frequency cycles in the total output. However, if you mix octave related outputs you will hear something of each, but bear in mind the "chopper" only responds to logic levels.
For example, if logic high is 9v and low is 0v, then 2 counter outputs simply mixed together will at times produce 4.5v which the CMOS switch just won't know what to do with.
You could add pulse generators to turn the square waves into shortened pulses which don't overlap, but it can be tricky to tune that since the pulse widths will be fixed whatever the period of the counter outputs are and only sound right over a limited musical range.

That's what I was expecting somehow.
So there are some possibilities left;
I could add a BJT as chopper which can be fed with more than one octave related outputs.
How about another 4016 with 4 switches and wire them parallel to the chopper feeding each with a different output (see schematic)?
Is there anything else to consider when mixing the choppers?
This would probably be the easiest way using the envelope and multiple octave outputs, correct?

[anotherjim]

Extra switches "maybe" doable.
What you will get is that the high period of the lowest octave will dominate and mask the higher ones off. When the lowest octave goes low, the next octave that is high will take over. Only when all of the octaves connected are low will you get anything happening.
I'd sketch a timing diagram - the best you might get is an output which has the frequency of the lowest selected octave, but a different pulse width that isn't 50% duty. That will change the harmonic balance, but it can't sound like separate "voices". The closest you might get to properly separating them is send some through phase shifters or chorus delay, if that's the hope.

[anotherjim]

Found a diagram...

You can see all outputs are low following the first clock. Once QA goes high, there is a long period when at least one Q output is high - until the 16th count and it all goes low again. That will sound like the lowest frequency output with a really narrow pulse width. As you take one or more Q outputs out of circuit, the width and frequency of pulses increase, introducing either wider pulse widths and/or brief mixtures of harmonics which will sound ok provided the lowest frequency isn't so low that we begin to hear the pulses as clicks.

[Mr. Lime]

Thanks, well that makes perfect sense.
I saw something similar in the phonic taxidermist circuit so i thought that might work..
http://www.mediafire.com/download/a9dgz3gzxje/voicevandal.zip
I think the sound demos of it show that the frequency bands are overlapping and won't give articulate octaves as well..
The Barton Waveshaper does a pretty good job in mixing the octaves so maybe I have to think the other way around and ditch IC9B.

Are there other ways to keep the envelope of the guitar signal for the harmonized output?
Maybe something like an envelope triggered vactrol instead of the chopper?
Probably that's the intention of the Zorgtaver here: http://www.zorgeffects.com/images/Manuels/Zorgtaver_Schematic.pdf

[Rob Strand]

Are there other ways to keep the envelope of the guitar signal for the harmonized output?
Maybe something like an envelope triggered vactrol instead of the chopper?

Virtually any modulation technique can be used.
The cool thing about the EMM circuit is it achieves the envelope with a simple circuit.
You can get a bipolar output a number of ways:
1) Have a negative peak detector as well.  Then on alternate cycles of the digital chopper input switch between the two envelopes to produce a bipolar signal.  That means you would add another gate like IC9c and R6, except the "new" IC9c is controlled by the inverted control signal of original IC9c.
Technically this produces a more accurate envelope as it tracks the assymetrical envelope of the guitar - see my previous plot.
2) A simpler way to generate the negative envelope swings is to use the positive envelope then invert that envelope and switch between the positive envelope and the synthesized negative envelope using the same switching scheme as (1).
3) Same result as 2 but moves the analog inversion to the output.  This one feeds the positive envelope into the existing circuit.  However on the opposite cycle of the chopper input you feed the output a non-inverting mixer instead of the current inverting input.
4) Same result of 2 except this time you use a circuit with a switch gain of + or -.  Basically this circuit with the inputs tied and R4 is made open and closed.    That would be my least favourite option as it puts the most demands on the opamp.  (It also loads down the peak detector.)
https://www.allaboutcircuits.com/uploads/thumbnails/Inverting-Op-Amp-Resistor-Calculator.png

Here's the type of waveform you would get out of (2), (3), (4).   Notice how it uses the positive envelope and the negative envelope is not followed (as it would in case (1)).

Probably that's the intention of the Zorgtaver here: http://www.zorgeffects.com/images/Manuels/Zorgtaver_Schematic.pdf

Notice the second order low pass filter to remove the harmonics.

[Mr. Lime]

Thanks for the plots!

I'm not sure if I got the bipolar signal thing going on right..
Can't the triggering signal for the "new IC9C" be taken from IC4B for a inverted control signal or do I have to add another inverter?

Mixer for the octave related signals plus envelope triggered vactrol is fine here?

[anotherjim]

Can't the triggering signal for the "new IC9C" be taken from IC4B for a inverted control signal or do I have to add another inverter?

I've no idea where that idea is going, but in principle, you can use existing IC4 pin 4 or it's inverse on pin10 for other switch control if they have the action you want.

The mixer/attenuator should be ok. I'm not sure if you may have problems with leakage when it should be off since it goes "off" only to whatever the attenuation might happen to be with the LDR dark resistance.

The caps coupling the octaves to the mixer pots could usefully be reduced for some high pass filtering to accentuate the treble of the higher octaves. A simple overall low pass filter could be a cap across the mixer amp feedback resistor. If more low pass filter was needed, you can "split" the 100k mix summing resistors in two and put caps from the middle to 0v. That might be particularly beneficial on the lower octaves.

[Mr. Lime]

The mixer/attenuator should be ok. I'm not sure if you may have problems with leakage when it should be off since it goes "off" only to whatever the attenuation might happen to be with the LDR dark resistance.

Yeah I might try two LDRs in series to get a total resistance of ~2M when the LED is dark.

The caps coupling the octaves to the mixer pots could usefully be reduced for some high pass filtering to accentuate the treble of the higher octaves. A simple overall low pass filter could be a cap across the mixer amp feedback resistor. If more low pass filter was needed, you can "split" the 100k mix summing resistors in two and put caps from the middle to 0v. That might be particularly beneficial on the lower octaves.

Good point! There's an article on Geofex about a PLL and divider and it's recommended to filter the squarewaves to get smoothed edges of the squarewaves. http://www.geofex.com/Article_Folders/Freq%20Mult-Div.pdf
But a 220p acorss the feedback R sure is a good idea. Have to try caps from split resistors as well if needed. 

I've no idea where that idea is going, but in principle, you can use existing IC4 pin 4 or it's inverse on pin10 for other switch control if they have the action you want.

I got that one wrong interpreted, I was actually looking at the threshold detector..
Is guess a negative envelope detector isn't beneficial for a LED/LDR combo?

I thought about a LFO input at the 4046. Maybe it's a good idea to use IC9B for it's gate so that the LFO is disconnected when not playing..