VCF Circuit Tests

[gstopp at fibermux.com]

     Hi all,
     
     Being badly bitten by the building bug, I've turned my energies from 
     VCO's to VCF's this week and have come up with two working filters to 
     add to my growing pile of pending projects. The first VCF is a 
     discrete 4-pole Moog ladder, and the second is a state variable as 
     published in the Electronotes Preferred Circuits Collection.
     
     *********************************************************************
     
     The Moog Ladder VCF:
     
     The Moog ladder VCF is built entirely of discrete components. I lifted 
     the design straight from the service manual for my Minimoog, which is 
     as far as I can tell a really old one. The VCF drawing is 1446 Rev. C, 
     dated 6/14/71. I used all 2N3904's, plus one 2N3906 at the CV summing 
     node. I picked the Minimoog filter over the Moog modular design 
     because it uses identical bipolar power supply voltages rather than 
     the asymmetric +12v/-6v supplies in the modular. The circuit works 
     fine at +/- 10, +/- 12, and +/- 15 volts. All resistor values are the 
     same as the original (except for a 91K on the range trim rather than 
     120K), and all the capacitor values are the same except the 
     electrolytic values were rounded up to the nearest next available 
     voltage and capacitance values. There may be a couple other resistor 
     values to change once I try keyboard tracking at 1v\octave, since I'll 
     probably be using +\- 15v supplies.
     
     This VCF sounds great. It has the characteristic 4-pole "warm" sound 
     when no resonance is used, with the decrease in overall signal volume 
     when resonance is increased, all the way up to the smooth whistle of 
     self-oscillation.
     
     One mistake I did make during testing - I applied the resonance output 
     from the output op-amp structure to the top of the resonance pot, and 
     the wiper of the pot to the resonance input at the bottom of the 
     transistor ladder, with the bottom of the resonance pot grounded. In 
     this state the filter acted strangely and the waveform at the output 
     was very strange. I double-checked the schematic - the resonance pot 
     is connected as a rheostat, not a voltage divider! I un-grounded the 
     bottom of the resonance pot and all was fixed.
     
     I would recommend this circuit to any home-builder - it's a definite 
     classic and it is not really hard to build (no power pins to hook up 
     on any of the components!).
     
     *********************************************************************
     
     The Electronotes State Variable VCF:
     
     From the Preferred Circuits Collection, VCF Option 1 from EN #71 page 
     16.
     
     This VCF uses an op-amp CV summer driving an exponential current 
     source made from a 2N3904 and an AD821 matched PNP pair. Each PNP 
     transistor in the AD821 drives its own filter pole in the filter guts. 
     The filter guts has a CA3140 signal input summer, feeding into an 
     integrator made from a CA3080 and a CA3140 (the first pole). This then 
     feeds another identical integrator (the second pole). The output of 
     this goes back around into the signal input summer, completing the 
     state variable loop. The highpass output is taken from the output of 
     the signal input summer. The bandpass output is taken from the output 
     of the first integrator's CA3140. The lowpass output is taken from the 
     output of the second integrator's CA3140. Each output has its own 
     buffer op-amp to protect the CA3140s from output loading. A notch 
     output is created by summing the lowpass output with the highpass 
     output. The resonance pot is a voltage divider fed by the bandpass 
     output, with the wiper of the pot feeding back into the signal input 
     summer. An interesting point is that maximum feedback equals the 
     lowest resonance, and minimum feedback or none at all equals 
     self-oscillation, which is the opposite from the 4-pole filter where 
     higher feedback means higher resonance.
     
     The CA3140's are used due to their extremely high input impedance, 
     which allows the filter's cutoff to reach well down into the sub-sonic 
     region, making it useful for control voltage processing (it's D.C. 
     coupled). This filter also has phase-lead circuitry (a cap in the 
     dividers at the inputs of the CA3080's) which contribute to the 
     stability of the filter at very high resonance and high frequency 
     settings. The CA3140's are also important in their predictable 
     phase/frequency response. The highest cutoff frequency is around 18 
     Khz.
     
     This VCF sounds great also. The high resonance settings are 
     particularly "sproingey" sounding. The lowpass output has a slightly 
     brighter sound than a 4-pole VCF. The bandpass sounds like a wah-wah 
     pedal, and the highpass sounds, well, like a highpass filter. The 
     notch output sounds like a simple phase shifter such as an MXR Phase 
     90. The best phase shifter sound is when the Q is turned all the way 
     down. If the Q is increased to high settings, the phase shifter effect 
     goes away, probably since the notch gets really narrow.
     
     As with all state variable filters I have played with, there is a fine 
     line between really high resonance settings and oscillation, and 
     self-oscillation does not happen smoothly but rather gets "triggered" 
     by something (like a large input transient or rapid change of cutoff 
     frequency). Once the filter gets "tripped" into self oscillation it 
     slams hard against the power supply rails and therefore is loud and 
     clipped (i.e. obnoxious). Perhaps some experimentation should be done 
     to provide some kind of limiter to the Q feedback to maintain a nice 
     sine wave.
     
     Anyway this VCF is quite easy to build and performs great. While the 
     original design uses a tempco resistor and an AD821 matched pair 
     (which the Analog Devices MAT-03 will replace) in the exponential 
     converter, I simply used a normal 5% resistor instead of the tempco 
     and a pair of 2N3906's for the matched pair. In VCF applications this 
     is much less critical than for VCO's and so I'll probably leave it 
     this way.
     
     That's it for now,
     
     - Gene