VCOs for home-built modulars - update

[gstopp at fibermux.com]

     Howdy howdy,
     
     Vacation's over, back to the salt mines....
     
     NATIONAL APP NOTE #299 VCO:
     
     I gave up on this one. Too funky.
     
     ELECTRONOTES VCOs:
     
     Went to the eee-lectronic store today and picked up some CA3140 
     MOSFET-input op-amps, then stuck 'em into the Elecronotes VCOs as the 
     integrator buffers.
     
     The sawtooth VCO runs real nice but won't go below about 9 hertz no 
     matter what I do. Looks like a good audio range VCO however. High end 
     is around 60 khz. The sawtooth waveform is perfect.
     
     The tri-square VCO now oscillates from 1/15 hertz to 58 khz in one 
     sweep of the course frequency pot. The waveforms are picture perfect 
     all the way except for the sawtooth, which due to the way that it is 
     derived gets funny lookin' starting at about 18-20 khz. This however 
     is no big deal and it's real pur-dee from there on down to the lowest 
     frequencies.
     
     THE CHROMA VCOs:
     
     I got the Exar 4151's in the mail (thanks Ric!) and made a dual 
     sawtooth VCO per the Chroma schematic. This dude has a sweep range 
     from around 60 hertz (try to go lower and oscillation stops) to 
     just above 30 khz. Waveshape ranges from a very jittery 
     rounded-off rising saw at the low end, to a fairly normal-looking 
     saw in mid-frequencies, to a half-sawtooth half-square-wave mutant 
     thing around 15 khz, to a train of narrow pulses around 25-30 khz 
     before they dwindle away into thin spikes. Now the jitter at the 
     low end is probably due to the VCOs glitching each other's power 
     rails since I have their CV inputs ganged together and no bypass 
     caps at all (yah I know bad design practice). Also the Chroma docs 
     talk about waveshape being funny and I guess they don't think it's 
     a big deal since it's a subtractive machine anyway. No tracking 
     tests were done.
     
     What do I think about this design? It's so similar to the 
     Electronotes sawtooth VCO that the slightly lower parts count does 
     not justify the drop in performance and waveshape. This VCO may be 
     good enough for the bowels of a polysynth, but I'm after 
     studio-quality modular specs here. Besides if I'm gonna spend the 
     time, energy, and all that to sit down and burn my fingers with a 
     soldering iron making some analog monster from hell, at least I 
     want a QUALITY WAVEFORM analog monster from hell.
     
     Besides I personally believe that waveform precision from 
     sub-audio to hypersonic has an effect on the sound. Yes this is 
     subjective and therefore I can get all cranky and argumentative 
     about it because hey - after all, the less evidence you have for 
     something the stronger the personal belief, right? Look at 
     religion. I mean if it were casually proveable on the test bench 
     by any Tom, Dick, or Harry with an O-scope then who would be 
     arguing?
     
     CONCLUSIONS:
     
     That being said, my vote falls to the Electronotes triangle-square 
     VCO. To recap, this VCO consists of a TL082 dual op-amp acting as 
     the CV input summer with a tempco resistor element and exponential 
     converter reference amp, driving a volts/octave trimpot into an 
     Analog Devices MAT-03 precision matched PNP pair, which drives a 
     CA3080 integrator. The cap in the integrator is buffered with a 
     CA3140 ultra-high input impedance MOSFET op-amp, which then goes 
     into a Schmidt trigger made from another CA3080. The output of this 
     Schmidt trigger feeds back into the integrator 3080 to act as the 
     triangle direction control. The 3140 buffer is itself buffered with 
     another half-TL082 amp, to form the triangle output buffer. The 
     Schmidt trigger 3080 output drives a 748 comparator to form the 
     square wave output buffer. The triangle and square also go to the 
     sawtooth function generator, which uses an invert/don't invert 
     method to make a sawtooth wave. Another 748 is used to create a 
     pulse wave, with another half-TL082 as a pulse width control summer 
     for pulse width modulation. A "Symmetrized Ramp Modulation" feature 
     is implemented so that the pulse width control voltages also mutate 
     the sawtooth shape for a unique timbral modulation. The builder can 
     make the PWM and saw shape independent to comform more to "normal" 
     VCO design practices if desired.
     
     Component discussions:
     
     The MAT-03 may be replaced with hand-matched 2N3906's. I have not 
     tried this and therefore can't dictate a procedure to anybody at 
     this time. However there is a procedure somewhere in Electronotes 
     and I will try to look for it. Doing this will get you around the 
     $8.50 or so that these parts cost. Personally I would rather use 
     the expensive bugger because frankly it's worth it to me.
     
     The tempco resistor can be replaced with a normal resistor during 
     construction. Eventual swap-out is recommended because then the 
     VCO will be totally perfect in all respects and heck that's a good 
     feeling.
     
     The CA3080 is readily available.
     
     The CA3140 has really high input impedance. This means that when 
     used as a buffer on the integrator cap the VCO will go down to 
     really low frequencies. You can use something else but it sure is 
     neat to have a real low bottom end for modulation purposes.
     
     The 748 is an old, old, uncompensated 741. You may wonder what such 
     an old non-ideal device is doing here but when you make a Schmidt 
     trigger with hysteresis out of it I dare you to see the vertical 
     edges on the square wave it produces with a scope. Once again here 
     goes my personal belief that the extra extra huge slew rate which 
     makes for the extra extra sharp edges on the square waves makes a 
     really profound hollow-sounding square wave. Just my opinion.
     
     The TL082s are real common and better than 1458s. Besides I steal 
     them from work.
     
     It may seem redundant to buffer the buffer to bring out the 
     triangle wave. This is done since the 3140's output is used 
     elsewhere in the VCO and won't drive a short to ground through a 
     1K resistor and so it should not be available to the outside 
     world. See the act of plugging and unplugging patch cords in 
     modular systems causes temporary shorts in the tip to ground and 
     if you did this with the raw triangle output the VCO guts could 
     get glitched. My motto is "buffer everything".
     
     The parts I will order and then offer for resale to anybody who 
     wants them are:
     
     MAT-03 Precision Matched PNP Pair - Analog Devices
     CA3080 Transconductance op-amp    - Harris, National (LM3080) 
     CA3140 MOSFET input op-amp        - Harris
     MC1748 Uncompensated 741 op-amp   - Motorola 
     MPF102 Small signal FET           - Motorola
     
     And of course the tempco resistors, for which I am still awaiting 
     samples for testing. I will be ordering 2 K ohms +3300 ppm/C' since 
     that is the part specified in the EN schematics. I know that some of 
     you out there might like 1.87 K ohm values because those are 
     replacement parts for ARP VCOs among others, but I'm not fixing ARPs 
     here, I'm building new modules.
     
     Okay bye.