Up to 3 signal inputs can be applied to circuit points AIN1 thru AIN3. The .1uF input caps into 1 meg input resistors reduces the high pass effect that a smaller input resistor would have since pin 2 of U3-A is a virtual ground. Square waves of relatively low frequency will not differentiate too badly thus preserving their low frequency content. Signal levels of +/-5V are expected. If you have higher signal levels then reduce the value of R19 to insure that U3-A is not clipping when you feed in your signals. The opposite would be true as well (for lower levels increase the value of R19 to get adequate signal to noise ratio). U3-A acts as an active mixer with a gain of .30 (with values shown). The output of U3-A is fed into the filter via R20. The high pass output appears at the output of U3-A also and is fed to the HP output via R14 2K.

The suggested panel layout shows jacks and three pots used as adjustable voltage dividers (level controls) to provide an input mixer for the unit. The jacks and mixer pots are not shown on the schematic.

This VCF uses LM13700 transconductance amplifiers as voltage-controlled integrators. There are two of them in a chain and they operate in the same way. U2-B's transconductance is controlled by current flowing from ground via Q3 and R16 into U2-B pin 16. It acts like a voltage-controlled resistor that in conjunction with C2 is an RC filter. Since the output of U2-B is a current the signal is actually integrated onto C2. The filtered (actually integrated) signal is buffered by U4-B and fed to the next stage (a portion is also fed back to the input of U2-B via R23). R15 is used to bias the LM13700 linearizing diodes on (which is advertised to reduce distortion through the amp). R26 compensates for the positive offset applied via R15 (yes it goes to -12 but its applied to the inverting input). The goal is to keep the signal path of the system as close to operating about ground as possible. In practice you will see as high as +/-200 to +/-300 mV of offset at any of U4's outputs but that's fine. If things were operating near the rails we would have a problem.

Here's my LAME-O explanation of how this works in laymans terms (the only terms I know). In order to obtain the state variable function notice that the low pass signal (at pin 1 of U4-A) is fed back to the inverting summer (U3-A and associated components) via R32 49.9K resistor. This in essence causes the low pass signal to be subtracted from the original signal resulting in high pass output at U3-A pin 1. The band pass output sounds like it consists of the difference between the output of the first filter pole and the low pass output. This drawing helps me to visualize this concept.

If this is so off base that it offends any math-savvy electronics genius types please feel free to submit your explanation. If you want a far more detailed and exact description of the theory of State Variable Filters check out Rayal Johnson's excellent paper Programmable State-Variable Filter Design For a Feedback Systems Web-Based Laboratory. Oh and go dig up your big-ass HP graphing calculator... cause you're gonna need it.

Each filter section contributes 6dB/octave filtering (thus the two together result in 12dB/octave). The cut-off frequency control voltage inputs (CV1, CV2 and CV3) are applied to U1-A via the 100K input resistors R9, R10, and R12. Control voltage of between -5V and +10V are expected. Initial Cutoff Frequency control (R4) is used to set the initial cut-off frequency. R1 and R5 limit the range of voltage available at the wiper of R4 in the extreme positions. The summed control voltages are inverted by U1-A (inverting amp gain of .02) and fed to trimmer R2. The 20mV per volt output from U1-A can be trimmed to the requisite 18mV/volt by R2 that drives logging transistor Q1 to achieve a 1V/oct response in filter cut-off frequency. Current through Q1 is mirrored by Q2 that controls the current flowing from emitter to collector in Q3 and Q4. These transistors (Q3, Q4) control the current flowing from the "amp bias input" pins of the LM13700s to ground via current limiting resistors (R16 and R18). The result is that both integrators are controlled simultaneously. Since both are tuned to the same cut-off frequency very little signal above the cut-off frequency gets through the low pass portion. The low pass output is critical to the operation of the other filter modes (as described above). U3-C (and associated components) buffers and amplifies the low pass output. U3-B (and associated components) buffers and amplifies the band pass output. U3-A (and associated components) is the input mixer and doubles as the high pass output buffer. All filter outputs are protected against shorts by 2K resistors. The band pass output is is inverted and given a gain of 3 by U3-D which feeds the resonance/feedback circuit (Point RA).

I added a 100K pot (R58 100K trimmer) to allow for high frequency scale trimming. It is important that the first few octaves be tuned while the trimmer is adjusted to have the least effect on the cut off frequency. You can tell which direction to turn the trimmer by setting the filter to oscillate (resonance all the way up) at about 1 to 2 KHz. Turn the trimmer. If the frequency starts to go up then turn the trimmer the other way completely. Later during calibration of the filter's scale you will add some high frequency compensation as needed for octaves above about 1KHz.