Alternative approaches to sweeping variable-state filters

[Mark Hammer]

Many of the better-known auto-wahs employ a variable-state filter configuration, generally involving three op-amps, and varying the input resistance to two of them, using opto-isolators.  This includes the Mu-Tron III and Lovetone Meatball, and all their assorted love-children.

I was musing the other day, and remembered a suggestion from Craig Anderton waaaaayyyyyyy back, and found in issue #8 of DEVICE, on page 13 ( http://hammer.ampage.org/files/Device1-8.PDF ).  In it, he notes that the focal/corner/cutoff frequency of the filter can be adjusted in two ways.  One is the series resistance, as mentioned (and here one would normally stick an LDR of some sort in parallel with a fixed resistor), but the other is to use a voltage divider on the output of the preceding stage (as shown in the mod).  Anderton notes two principle advantages: 1) the much greater sweep range available (1000:1 vs 100:1), and 2) the greater latitude in what value of pot needes to be used.

So I'm thinking, maybe the approach to producing a variable resistance that is adopted for the Dr. Q, Bassballs, and a myriad of other swept bandpass filters could be implemented with the standard variable-state configuration.  So, let us say that the two stages in question each feed a section of a 100k dual-ganged pot, but there is a bipolar transistor from wiper to ground in parallel with the ground leg of the pot section, driven by a simple envelope follower.  The dual-ganged pot would allow for some tuning of the initial; start-frequency of the sweep.

Would that work?

[R.G.]

I'll think about it some. I think it would work fine, unless there is some fishhook hiding in there.

The typical state variable filter uses inverting amplifiers and integrators. Way down at the bottom of the pile of how these filters work is the idea that the input signal is a *current* into the inverting input. The series resistor approach changes this current by forcing a signal voltage before the series resistor to flow through the (variable) series resistor. In the real world, the series resistor as a pot can only go as low as the contact resistance on the end of the sweep at one end and as high (that is, as low a current) as the total resistance on the other end of the rotation.

A shunt pot with the signal off the wiper to a current input looks like a variable voltage as divided by the pot setting, which is then converted to a current into the inverting input by whatever resistances the pot connection and any other resistances make it. The advantage that the shunt pot/voltage divider approach has is that the pot could, if perfect, take the input signal all the way to zero, reducing the input current to nearly zero. This is the equivalent of a series resistor going to infinity as far as how much current goes into the input, so the available ratio of full current to lowest current is indeed much larger.

My handwaving before saying it's universally applicable is based on not knowing what the rest of Anderton's circuit is. I have used this approach on the speed control of LFOs in an integrator/Schmitt trigger oscillator for the freedom it gives in letting you use a linear pot and taper it instead of using a reverse-log speed pot. It also, as Craig notes, makes the exact pot value matter less. I'd have to do some math before agreeing that it can be almost anything. I've always approached this design by using the pot to divide the voltage, then using a resistor in series with the wiper that's greater than the pot value to ensure that the pot's taper is what makes the variation, and not to rely on the Thevenin equivalent resistance of the pot as part of the control curve.

That being said; a series resistor feeding a shunt BJT divider at the input of an opamp acts more like a current divider, but this particular connection has the advantage that the BJT sees only millivolts of signal across it, literally the error voltage in the opamp loop. That means the normal limitations of the BJT as a variable resistor are minimized. The signal is so small at that point it acts linear.

There may be some quirks of the BJT used as a variable resistor, including offsets and feedthroug; however, they do seem to be OK in the pedals you mentioned, so it may be fine.

Good place to look.

On the other hand, I've had really good luck with the LM13700 plus a dual opamp as an SVF.

[Blitz Krieg]

using a 10k pot along with the mod puts the range back where it was in the original (with 100k)?