Something New from Doc Sketchy

[Dr. Sketch-n-Etch]

If you would like a quantizer module in 5U behind a 1U panel with an input jack, an attenuating pot, and four output jacks, then I will. Cost: $125 US.

Sounds great! Can I put on my 4U list?

Well, the board is 3.1" x 6.8", so perhaps a little long for 4U, but I believe it would fit. However, it would require a pretty wide panel, and there is very little on the panel (one knob and a few jacks). For 4U, it would be better to mount the PCB somewhere else in the case, and just bring the input and output leads to a very narrow panel, or even incorporate the quantizer function in some other module, and stack the board on top of other one.

[KSS]

@Dr. Sketch-n-Etch
Fully realize that you've achieved elegant simplicity. But since it's flash conversion, how many channels do you think could be reliably supported with the addition of input output switching and S/H's ?

[Dr. Sketch-n-Etch]

@Dr. Sketch-n-Etch
Fully realize that you've achieved elegant simplicity. But since it's flash conversion, how many channels do you think could be reliably supported with the addition of input output switching and S/H's ?

I'm not sure I entirely understand the question, but if the switching were clean and done completely before the sample was held, then as many as you'd like. The "propagation delay" through the quantizer is the delay of six opamps and a CMOS gate, so about 7 microseconds or less. Hence, you could conceivably sample a new voltage every millisecond and have loads of time to spare.

Does that answer your question?

[Dr. Sketch-n-Etch]

Here is the completed module. I baked the panel lastnight (15 minutes at each temperature: 1) Warm, oven door open, 2) Warm, oven door closed, 3) 200F, 4) 250F, 5) 300F, 6) 350F, and 7) 400F, so one hour and 45 minutes total bake time -- the bake was perfect (which is also a function of giving the printed Lazertran a good going over with the heat gun before transferring). Add 15 minutes of prep, and it's a two-hour process to make a baked-on Lazertran panel). I drilled it this morning, which was a snap. Then I made the stooge bracket, which took about 15 minutes. Wiring took another 10 minutes or so. It works (I tested it). The only thing is that I neglected to include a Panel Ground pin on the layout, so I soldered a wire to one of the ground jumpers and took that to the nearest jack, then from there to the pot. I'll fix that for future builds.

[KSS]

Yes it does! Thank you.
--------------------
Nice looking module.

[Moog$FooL$]

nice build!!

[Dr. Sketch-n-Etch]

nice build!!

Why, thank you, Moog$Fool$. I had fun doing it -- I had to move the wife's van out of the garage so I could get to my bandsaw to cut the stooge bracket -- a typical day in Sketchy Labs -- you know the drill.

[Rex Coil 7]

Wow. Dave, I've arrived at a point where I can no longer muster words. Fit, finish, functionality, workmanship, sensible layouts. You've certainly left your mark. Inspirational in every regard. Demonstrable pride and passion.

[Dr. Sketch-n-Etch]

Wow. Dave, I've arrived at a point where I can no longer muster words. Fit, finish, functionality, workmanship, sensible layouts. You've certainly left your mark. Inspirational in every regard. Demonstrable pride and passion.

Well, thank you kindly, Mr. Coil. As you know, I'm even more blown away by the craaaazy stuff that you get up to as well. Of all the people out there in the Muffyverse, I believe that you'd be the most fun to collaborate with (and maybe to jam with).

But... you just know that those single-wire 2-pin connectors are probably gonna shake loose eventually. Of all the builds I've done, these seem to be the best candidates for actually soldering wires to the PCB.

[Prunesquallor]

If you would like a quantizer module in 5U behind a 1U panel with an input jack, an attenuating pot, and four output jacks, then I will. Cost: $125 US.

Sounds great! Can I put on my 4U list?

Well, the board is 3.1" x 6.8", so perhaps a little long for 4U, but I believe it would fit. However, it would require a pretty wide panel, and there is very little on the panel (one knob and a few jacks). For 4U, it would be better to mount the PCB somewhere else in the case, and just bring the input and output leads to a very narrow panel, or even incorporate the quantizer function in some other module, and stack the board on top of other one.

Gotcha!

[chuckg]

Awright, this is pretty cool. I'm still trying to get my noodle around one thing: the gray code. I know what gray code is, but not seeing how it happens.

Seems to me each ABC stage essentially doubles the incoming voltage, and compares the output to some fixed reference. Would not the row of outputs from these comparisons already be in plain old standard binary?

[cygmu]

Awright, this is pretty cool. I'm still trying to get my noodle around one thing: the gray code. I know what gray code is, but not seeing how it happens.

Seems to me each ABC stage essentially doubles the incoming voltage, and compares the output to some fixed reference. Would not the row of outputs from these comparisons already be in plain old standard binary?

No, because the rectification folds the input of each stage over. If you don't do the rectification then you get plain binary.

This is quite a nice article on the folding ADC idea.
https://www.analog.com/media/en/trainin ... MT-025.pdf

[Dr. Sketch-n-Etch]

Yes, MT-025 is what I referred to when I was designing my circuit. It's a very good article. As cygmu said, the folding gives the Gray code. If the ramp is just doubled so that you have a series of voltage ramps at doubled frequencies (like a sawtooth wave), then this gives regular binary. The problem with that is that the sudden jump from +5.333V to -5.333V can create problems, whereas the reversal of direction at the peaks is much better behaved. Plus, you can't really drive an R-2R ladder properly from op-amp comparators. The CMOS gates convey exactly 0V and 5.333V into the ladder, so even without the need to convert Gray to Binary, you'd still want to put the comparator outputs into CMOS buffers.

Here are the relevant figures from the paper. The first shows just doubling the voltages, and resetting them to the bottom when the top is reached. Putting these ramps into comparators gives regular Binary code:

The second shows doubling and rectifying the voltages. Putting these triangles into comparators gives Gray code. That's what I'm doing.

I did something very similar in my interpolating scanner projects.

[Dr. Sketch-n-Etch]

Incidentally, I am also going to normal the +5.333V source into the input jack, so that if nothing is plugged in, the quantizer will generate stepped constant voltages, like a one-step sequencer. I don't know why I didn't think to do that before.

[Rex Coil 7]

Incidentally, I am also going to normal the +5.333V source into the input jack, so that if nothing is plugged in, the quantizer will generate stepped constant voltages, like a one-step sequencer. I don't know why I didn't think to do that before.

... hose head ...

[Dr. Sketch-n-Etch]

So, I did it and it worked. I now have the final layout all ready to rock, if anybody wants a board. $25 + shipping, or $75 built and tested, or $125 panelized.

Also, I'm working on dodecatonic (chromatic), octatonic, heptatonic (diatonic), and hexatonic (whole-tone) versions of the quantizer. For these, I'm separating the quantizer into a 3-bit octave quantizer, and a 4-bit chromatic quantizer. The octave quantizer will give 8 octaves in 1-V steps, and the chromatic quantizer will break each octave into 12, 8, 7 or 6 equal steps. This is relatively easily accomplished simply by manipulating the reference voltage levels to set the voltage where the 15 to 16 transition would occur (1.333V for chromatic, 2.000V for octatonic, 2.286V for diatonic, 2.667 for whole-tone). Then, if the octave quantizer output is subtracted from the incoming voltage, I'll always get a control voltage to the chromatic quantizer which is a 0-to-1-V ramp. Each of the four versions will use the same R-2R ladder which generates chromatic steps of 83.333mV. Once the proper number of notes-per-octave is achieved, the logic to the ladder will be altered to give various scales (whole-tone for the 6-step; major, ascending melodic minor, and harmonic minor for the 7-step, diminished for the 8-step, each selected by a rotary switch which drives various logic and analog switching chips).

For example, the heptatonic quantizer will generate the first 7 binary numbers in the 0-to-1-V ramp, and these have to be altered to other numbers with digital logic to achieve a major scale:

0 = 0000 --> logic --> 0 = 0000
1 = 0001 --> logic --> 2 = 0010
2 = 0010 --> logic --> 4 = 0100
3 = 0011 --> logic --> 5 = 0101
4 = 0100 --> logic --> 7 = 0111
5 = 0101 --> logic --> 9 = 1001
6 = 0110 --> logic --> 11 = 1011
7 = 0111 --> logic --> 12 = 1100

This system should never reach state 7, but if it does, then it will briefly generate an octave, before the octave quantizer jumps 1V and simultaneously resets this system back to state 0 via the voltage subtractor. Since there is the possibility of slight voltage mismatch between the two quantizers, the logic for generating an octave at step 7 should be in place since the octave transition may occur after the chromatic quantizer has already reached state 7. If this happens, the transition should be seamless, although the slew rates of the opamps may create a situation where the output jumps one octave for a few microseconds. Clever use of a small capacitor somewhere should prevent this, though.

I expect to have this design done later this coming week. I just hope the logic doesn't turn out to be too excessive. Of course, to get a melodic minor from a major, one only has to change the 2 --> 4 transition to a 2 --> 3 transition, so this is just one tiny additional logic element. This will give me a chance to dust off the ol' Karnaugh maps an' shit. Of course, it really doesn't matter how complex the logic is, since I'm doing all of this for fun anyway.

[Rex Coil 7]

So, I did it and it worked..... Also, I'm working on dodecatonic (chromatic), octatonic, heptatonic (diatonic), and hexatonic (whole-tone) versions of the quantizer.... This will give me a chance to dust off the ol' Karnaugh maps an' shit. Of course, it really doesn't matter how complex the logic is, since I'm doing all of this for fun anyway.

Ok, it shall be named the Hiptocatic Dude-A-Tonic ol' Karnaugh Maps an' Shit .. that whole thing ..

I mean with all of the ~nics~ and ~tics~ in there ...... right?


... what? ... it works. .... or something.

[KSS]

For these, I'm separating the quantizer into a 3-bit octave quantizer, and a 4-bit chromatic quantizer. The octave quantizer will give 8 octaves in 1-V steps, and the chromatic quantizer will break each octave into 12, 8, 7 or 6 equal steps.

This sounds a *lot* like the one I did long ago and mentioned earlier.
Same 8V for octaves and independent 4bit 12 of 16 for the semitones. I used 4051's to select taps in resistor strings rather than R/2R. <-- mistake in earlier post. It was essentially Flash ADC to 4051's giving a digital input to a typical analog KBD. Like you I messed with manipulating the logic to get different scales but eventually settled on simple semitones. Did try 12 switches to mute different taps in the ladders. Was a nice mod.
Initial ADC was PAIAs flash converter. Back then -pre midi and most things digital- it didn't matter that the ADC wasn't very linear -without extensive resistor tailoring- because the voltage in wasn't being compared to or used with other things like is more common today. Just needed to get a sequencer to output chromatic steps from arbitrary CVs. For that it worked well.

As I wrote before much more complicated than your design. Looking forward to seeing your no doubt more elegant design!

Edit:

For example, the heptatonic quantizer will generate the first 7 binary numbers in the 0-to-1-V ramp, and these have to be altered to other numbers with digital logic to achieve a major scale:

0 = 0000 --> logic --> 0 = 0000
1 = 0001 --> logic --> 2 = 0010
2 = 0010 --> logic --> 4 = 0100
3 = 0011 --> logic --> 5 = 0101
4 = 0100 --> logic --> 7 = 0111
5 = 0101 --> logic --> 9 = 1001
6 = 0110 --> logic --> 11 = 1011
7 = 0111 --> logic --> 12 = 1100

This is the exact problem I needed to solve for my Triadex MUSE clone "Euterpe". Went with the tapped resistor string again over using logic.

[KSS]

But... you just know that those single-wire 2-pin connectors are probably gonna shake loose eventually. Of all the builds I've done, these seem to be the best candidates for actually soldering wires to the PCB.

Or just remove the white plastic body. Without the second contact -used or not- It's adding nothing -besides "pretty"- except mass and leverage to the single pin contact. Putting the 2nd contact -unused- in place will help stabilize the affair. If you haven't already.

I'd just use a single pin HDR and a single crimped hi-pressure molex contact. Who sez we have to have a plastic contact holder? A single pin-contact like this *can* be trusted and is still removable for service.

[Dr. Sketch-n-Etch]

But... you just know that those single-wire 2-pin connectors are probably gonna shake loose eventually. Of all the builds I've done, these seem to be the best candidates for actually soldering wires to the PCB.

Or just remove the white plastic body. Without the second contact -used or not- It's adding nothing -besides "pretty"- except mass and leverage to the single pin contact. Putting the 2nd contact -unused- in place will help stabilize the affair. If you haven't already.

I'd just use a single pin HDR and a single crimped hi-pressure molex contact. Who sez we have to have a plastic contact holder? A single pin-contact like this *can* be trusted and is still removable for service.

Actually, I altered the board by adding GND to the top unused pin, and +5V to the bottom unused pin. Both were needed on the panel (GND for panel ground, and +5V for the normal to the INPUT jack).

[KSS]

[Dr. Sketch-n-Etch]

For these, I'm separating the quantizer into a 3-bit octave quantizer, and a 4-bit chromatic quantizer. The octave quantizer will give 8 octaves in 1-V steps, and the chromatic quantizer will break each octave into 12, 8, 7 or 6 equal steps.

This sounds a *lot* like the one I did long ago and mentioned earlier.
Same 8V for octaves and independent 4bit 12 of 16 for the semitones. I used 4051's to select taps in resistor strings rather than R/2R. <-- mistake in earlier post. It was essentially Flash ADC to 4051's giving a digital input to a typical analog KBD. Like you I messed with manipulating the logic to get different scales but eventually settled on simple semitones. Did try 12 switches to mute different taps in the ladders. Was a nice mod.
Initial ADC was PAIAs flash converter. Back then -pre midi and most things digital- it didn't matter that the ADC wasn't very linear -without extensive resistor tailoring- because the voltage in wasn't being compared to or used with other things like is more common today. Just needed to get a sequencer to output chromatic steps from arbitrary CVs. For that it worked well.

As I wrote before much more complicated than your design. Looking forward to seeing your no doubt more elegant design!

Edit:

For example, the heptatonic quantizer will generate the first 7 binary numbers in the 0-to-1-V ramp, and these have to be altered to other numbers with digital logic to achieve a major scale:

0 = 0000 --> logic --> 0 = 0000
1 = 0001 --> logic --> 2 = 0010
2 = 0010 --> logic --> 4 = 0100
3 = 0011 --> logic --> 5 = 0101
4 = 0100 --> logic --> 7 = 0111
5 = 0101 --> logic --> 9 = 1001
6 = 0110 --> logic --> 11 = 1011
7 = 0111 --> logic --> 12 = 1100

This is the exact problem I needed to solve for my Triadex MUSE clone "Euterpe". Went with the tapped resistor string again over using logic.

I think I'm going to be able to do mine with a CD4512, a DG408 or 409, and some diodes. There really isn't that much logic, but the "non-logical" aspect of the logic recommends a "data selector logic" approach. Since the major, melodic minor, and harmonic minor scales only differ from each other by one note, they can all use the same diode ROM with the 3 and 6 positions switched.

[Dr. Sketch-n-Etch]

I just wanted to announce that I've been having a hard time getting up the motivation to put the finishing touches on this quantizer design. I've been very busy with school work, and, more significantly, I have a lot of undone school tasks hanging over my head. In that situation, it is always hard for me to get motivated to do other things. Maybe next week. I have to write a recommendation letter, finish a big research proposal, finish a journal article, make some adjustments to a short course that I have to deliver on Oct 22, set a midterm exam for Oct 28, and review a couple of things. After that, my to-do list will be pretty blank and I'll have lots of mental space for electronic design again.

[Thorsday]

I just wanted to announce that I've been having a hard time getting up the motivation to put the finishing touches on this quantizer design. I've been very busy with school work, and, more significantly, I have a lot of undone school tasks hanging over my head. In that situation, it is always hard for me to get motivated to do other things. Maybe next week. I have to write a recommendation letter, finish a big research proposal, finish a journal article, make some adjustments to a short course that I have to deliver on Oct 22, set a midterm exam for Oct 28, and review a couple of things. After that, my to-do list will be pretty blank and I'll have lots of mental space for electronic design again.

You're still prolific by mere mortal standards, Doc. Don't forget rest and relaxation!

[KSS]