XR-2206 VCO

[expanoncolin]

My concern is that the VCO is probably the single most sensitive part of the analog synth. Other less stable VCO's can be Synched to a good one. But one of them needs to be real good if you're doing music. I've even toyed with idea of digital synthesis using microprocessor (haven't ruled this idea out yet).

I have thought of doing a DCO-core VCO too.  Seems like it would be legitimate to maybe start with a digitally generated ultra-precise sawtooth, then use the typical waveshaping circuitry to convert that to sine, triangle, "analog" saw, etc.  I bet you could get plenty of analog "fatness" with a digital core in that way.

-Colin

[ve3wwg]

For the digital VCO, I would just generate a square wave.

From a square wave you can:

• shape that to what you want (as you said)
• use that as a synch to an analog VCO.

A design requirement is that it must respond as if it were analog as the control voltage changes. This is critical for portamento effects. You don't want to hear discrete digital jumps in the frequency but rather a nice flowing of pitch change as the CV changes.

[expanoncolin]

How would you convert a square to anything else in an amplitude-independent way?  The conversion circuitry would have to know the frequency of the square wave, because nothing about the square's shape can indicate where in the wave period you are.  Saw to try is easy, http://www.musicfromouterspace.com/analogsynth_new/SAWTOTRICONVERT/SAWTOTRICONVERT.php and square is easy too (comparator), so is sine after you have the triangle (with LM13700 clipping).  To respond smoothly to control voltage changes, you'd just need a highly accurate ADC and some interpolation in your code.

-Colin

[ve3wwg]

How would you convert a square to anything else in an amplitude-independent way? 

The one way is to simply use the square wave as the synch signal. Just like the XR-VCO accepts a synch input, you can then generate your sine, triangle and ramps from that. So that was the 2nd bullet item I outlined.

The conversion circuitry would have to know the frequency of the square wave, because nothing about the square's shape can indicate where in the wave period you are. 

A square wave is still composed of a fundamental frequency. You can know when it starts using it's rising edge (at 0%) and you know when it is 50% when the negative going edge occurs (assuming 50% duty cycle). Between those two points, you can infer the time period. So a triangle wave could be rising (charging) until the negative going edge is received, at which point it then reverses and has a negative slope after that.

Of course amplitude would be a problem without some correction by averaging. But AGC type circuits can smooth that out over the course of a few cycles.

I'll agree that square might not be the most convenient starting point for processing a signal into other forms. But I'm mostly thinking in terms of synch-ing a VCO that already generates wave forms. The digital aspect just helps to make it frequency acurate.

[expanoncolin]

The one way is to simply use the square wave as the synch signal. Just like the XR-VCO accepts a synch input, you can then generate your sine, triangle and ramps from that. So that was the 2nd bullet item I outlined.

But sync signals are normally "hard sync", e.g., they will interrupt and reset the waveform being generated, creating new waveforms.  Unless the XR-VCO works differently?

Of course amplitude would be a problem without some correction by averaging. But AGC type circuits can smooth that out over the course of a few cycles.

I see this as a pretty big problem - you don't want to have the attack of your note "click" or fade in.  I could see a system where the DCO's brain was also outputting a voltage control to amplify the waveforms in a calibrated, frequency-dependent way... but I think it's easiest to start with saw, which is why you see so many saw-core VCOs and fewer square-core.

-Colin

[ve3wwg]

But sync signals are normally "hard sync", e.g., they will interrupt and reset the waveform being generated, creating new waveforms.  Unless the XR-VCO works differently?

I suppose it depends upon the VCO, but the XR-VCO does this rather gracefully. It's clever how it works actually. The synch signal actuates a "selection" to a different RC pair (actually just R in this case). If the osc is in synch, there is very little adjustment in play. When it's off, the synch causes a portion of the "other" R to be used, to in effect speed up the osc.

I've tried it out using the square wave from a signal generator. It works rather well. However, it can lock onto a harmonic if the two are way out. But I think this is acceptable.

In the digital with synched VCO case, you'd move the analog and digital VCOs in concert, so that both are always in the right vicinity. Then you lock the analog one to the pitch of the digital one with the synch input.

Of course amplitude would be a problem without some correction by averaging. But AGC type circuits can smooth that out over the course of a few cycles.

I see this as a pretty big problem - you don't want to have the attack of your note "click" or fade in.  I could see a system where the DCO's brain was also outputting a voltage control to amplify the waveforms in a calibrated, frequency-dependent way... but I think it's easiest to start with saw, which is why you see so many saw-core VCOs and fewer square-core.

Certainly there is no reason you couldn't generate both and completely eliminate the need for the analog VCO (unless you wanted fatness). You could easily generate the square, saw, triangle and even a pseudo sine.  But for the saw/triangle, I would lean to the idea of a constant current source controlled by the CPU, to generate the wave. With an appropriate DAC, you should be able to create the correct constant current charge rate for a saw/triangle wave (and discharge).

I think you're right-- generate it all up front. 

The one thing I want to avoid however, is that I don't want any sampled output. The sine wave might be the exception, though there are several ways to attack that too if approximations are acceptable.

[davebranton]

Hi there,

I just wanted to add to this discussion slightly, I just finished building an XR2206-based signal generator and it my circuit the 2206 runs off 24 volts without any problem at any frequency.

I've pretty much used the exact same circuit that's on the data sheet, with the addition of a set of 4 caps that can be selected in for timing. The circuit will go from less than 1Hz to 2MHz and the output looks pretty clean.

So whatever was causing your noise issue, I don't think it was the XR.

Hope that helps

-dave

[ve3wwg]

I don't know what further to say that:

• The VCO here is not the same circuit as the data sheet
• The same problem existed with 5 different chips
• Someone else on the net had the same problem on a breadboard (I can't find that link now)
• and his solution simply was to reduce voltage

The circuit diagram of this VCO is available here:

http://www.experimentalistsanonymous.com/ve3wwg/doku.php?id=synth_vco#xr-2206_circuit

To kill two birds with one stone, I used a simple RC decoupling method of reducing the voltage. So the question remains- was it the decoupling or the lower voltage that fixed it?

In my opinion, the solution would seem to be that the RC decoupling from the power line. But, the post that I cannot locate at the moment, claimed he changed the PS voltage and found that the problem occurred near 15 volts.  It is possible that going higher than 15 volts may have also have fixed it. I don't recall if he had that experience or not.

Note that there is a special circuit in this VCO going from pin 11 around to pin 1 of the XR2206 chip in this VCO design. Diodes D4 and D5 along with the voltage appearing at the cathode of D4 allow the Skew control to do some wonderful things to the waveform. There is some discussion of Skew here:

http://www.birthofasynth.com/Thomas_Henry/Pages/XR-VCO.html

Those who have built it from the PCB don't seem to experience the problem (if they have, they're not talking about it). So layout is probably a big factor (I have some long wire runs to and from the front panel). 

At this point, it is all academic since all is working very well now. I'm now only waiting for the tempco resistor, to keep it more temperature stable.

[Jarno]

Could very well be layout related, I made two, using the same layout, both running +/-15volts, both well behaved.

I used the bugbrand layout, with PC mount pots.

[Girts23]

Hi, collegues,
I just completed XR-2206 VCO and experienced EXACTLY SAME PROBLEMS. I tripple checked control circuits and they worked well, so thought that a generator chip is not working properly and ordered few more XR-2206s. But then I found this post and solved problems exactly according to your recommendations. Thank you!

[ve3wwg]

Very interesting indeed!

The circuit you built, was it on a pcb or did you hand
wire it like I did? Just curious.

[pollo]

this is a rather candid coincidence!

i was working in a version of the xr-vco of my own, and since PTC resistors are uncommon if not inexistent around here -and the customs charges and paperwork are not only expensive, but rather tedious!-, i implemented a version with master's Rene Schmitz CA3086 tempco.

would it be too shameless from my part to include my version in this thead?... anyway, here's the schematic, i've included some of the issues master Stites has noted in his site, mostly regarding diode orientations:

http://postimage.org/image/n2qj550s3/

disregard most of the opamps: i think i found LF442's around here, so these would replace the 1458 as required by master Henry; and the TL07x's that are not part of the TC circuit should be a TL074, for compactness sake.

also, note that this is a schematic only, i haven't tested it, or breadboarded or created a prototype out of it; just thought it would be nice to make a great sounding VCO with some parts at arms length. any revision to the board, advice, insult or flaming will be very welcomed.

[ve3wwg]

Interesting circuit. The added circuitry of U2B and U2A seem like added complexity, when the LF442 in place of the LM1458 would have been a lot simpler.

I found that you can still get the LM394 from Futurelec Electronics for $1.80 USD:

 http://www.futurlec.com/cgi-bin/search/search.cgi?Q=LM394

The LF442 is also available there cheap:

 http://www.futurlec.com/cgi-bin/search/search.cgi?Q=LF442

I finally just got my 2K TempCo resistors (R7) in the mail yesterday, after months of waiting (Magic Smoke).  At least they were kind enough to throw in a couple extra for my wait.  So now I have everything I need to finally finish this VCO with temperature stability. I also have a 10-turn pot waiting to replace the cheap trimmer I had used for the calibration. These are two things I would advise not scrimping on.

There must be other sources for these tempo resistors (they got them from somewhere), but knowing what to order is very tricky with these things. Anyway, I have 6 now, which might keep me going until retirement.

[pollo]

U1A and U1B are part of the 3086 TC circuit; that is, in order to replace the 2K PTC, so they should have not much to do with the LF442 -however, making complexity a necessity  . for comparation, see master Schmitz's VCO 4, between the v/oct trimmer and the expo pair/hi freq trimmer, and right above the waveshaping stage: 

http://www.uni-bonn.de/~uzs159/vco4.html

i'd use the 4.7 NTC, garden variety resistor method for temperature compensation (between the input summing stage and the expo pair), but given their fairy unstable nature i like to leave these for differential-paired VCLFOs and my 4069 VCOs (look amongst Schmitz and Takeda's designs for these).

by the way, the 442's are already at my disposition, according to Master Henry's demands  . and trust me, customs in chile are a pain: unless you're a private company dedicated to electronics, a physics/engineering/electronics college department, or something else like that you either have to lie (risking a hefthy fine); or pay for paperwork, a "customs managing services office" (some sort of weird enterprises dating from The General's dictatorship) and the multiple taxes themselves, which do not justify importation from serious companies, let alone the shipping rates. but we as chilean hobbists are a resourceful bunch, so added complexity is a minor nuisance.

which might keep me going until retirement.

from modular synths? hah, i tend to find this hard to believe.

[ve3wwg]

That sure sounds like a difficult situation- almost like a Russian owning a modem in the 80's. Hopefully things improve over time.