Wednesday, September 30, 2009

CGS Bi-N-Tic Filter

Well, the Super Psycho LFO rebuild / hot rodding is on hold for a couple of weeks. The issue: Each of the six oscillators has a 1M pot in the circuit that controls the osc rate. After thinking of various ways of combining the pot in series or parallel with a Vactrol's light-dependent resistor, I've decided that I'm not happy with any of them. There's just no good way to do it such that there isn't some value of the pot setting that makes the Vactrol ineffective, and vice versa. The right way to do it is to take the pot out of the oscillator circuit and have just the Vactrol in the circuit; then make the pot be part of a voltage divider that will supply an offset voltage, which is added to the external control voltage via an opamp voltage-adder circuit, and the sum of the voltages drives the Vactrol's LED. Problem: Although I can make a 0-5V voltage divider using the existing 1M pots, it's a bit iffy as far as staying well clear of the opamp's specified offset and bias currents. I'd feel more comfortable using a 100K or 50K pot, so it can supply a bit more current. But I don't have any. I'm in on a group buy at Muff's place, but it's going to take a few weeks for that to come together. So until that happens, the Super Psycho is in a wait state again.

In the meantime I have something else to play with:

This is a kit for a Ken Stone / CGS 57 Bi-N-Tic switched-capacitor filter. The kit is from the wonderful folks at Bridechamber and includes all parts. A close-up of the circuit board:

And the panel:

The kit incorporates nearly all of the mods and hot rodding that Ken calls out as options in his notes on this module. (The control that Ken labels "DAMPER" in his notes is called "RESONANCE" on this panel.) Switched-capacitor filters are very unusual in the synth world. I'm not quite sure why that is; they have a reputation for being noisy, but I think that rep might be the result of some early designs of the switched-capacitor concept that used mechanical commutators and crummy caps. I'm curious to build this and see how it sounds. I am going to add one mod: a switch to take the most significant bit of the digital count that goes into the multiplexor which selects one of the eight caps, and move it from the 4-bit to the 8-bit of the counter. That will have the effect of dividing the eight caps into two banks; instead of scanning all eight caps in order, it will scan over a bank of four caps twice, then switch to the other bank. I'm not quite sure what this will actually do; I think it will give the filter two resonance peaks. Anyway, it will be interesting to find out.

Sunday, September 20, 2009

The Wrong Vactrols

So I was all set to prototype the control voltage input circuits for the Super Psycho LFO rebuild. The basic idea is to feed the control voltage to the LED half of a Vactrol, and put the light-dependent resistor (LDR) half of the Vactrol in series with the rate pot of the specific oscillator (so there would be 6 Vactrols total). I have some VTL5C7's that I've had laying around from a group buy I participated in on Synth-DIY several years ago. I've never done anything with them.

So I went to the data sheet to find out how much current would be required to make the LDR vary between about 0 and 100K, which seems like a good operating range. (You don't want to get into the high resistances because the data sheet says the response is not specified above 1M or so.) And I saw these curves:

Note in particular the turn-off curve. If I want to bias it to operate between, say, 1K and 100K, then the turn-off time from full on is about half a second! Obviously, if you feed the input something like a 10 Hz square wave, the output is going to flatline. Never mind modulating at audio frequencies.

I know this is a fundamental characteristic of LDRs. However, Perkin-Elmer does make different types. On the same data sheet are the specs for the VTL5C6. Its turn-on and turn-off curves look like this:

Note the difference in the time scale. This one has a turn-off from 1K to 100K of 2 milliseconds. That's more like it! Trouble is, I don't have any of these. My usual go-to parts source, Mouser, doesn't carry Perkin-Elmer. Google showed me that Allied carries them, and I checked and they have hundreds in stock. So that's on order, and I expect them Wednesday or Thursday. Until then...

Sunday, September 13, 2009

Rebuilding a Super Psycho, Part 2

OK, I know it's been a while, but I ran into some employment difficulties. But that's all resolved now. Anyway, I've been working on the Super Psycho rebuild. The first step, as I documented in the previous post, has consisted of moving everything from the circuit board that was falling apart to a new board. Below is a photo of what I've gotten done:

A few things to note. I replaced all of the resistors along the top of the board with new ones from my stock, mainly because desoldering the old ones was going to be such a pain; that was an area of the old board where I'd had to add a lot of jumpers and kluges. Most of the capacitors were salvaged from the old board; I tested them all first. The two green 100 uF caps at the left are new from my stock because the old board was an earlier revision and it didn't have these. There's also a 10 uF cap above and below these; I replaced the one at the bottom because the original was borderline (measured 8.1 uF vs. a specified tolerance of 20%).

I've installed MTA-100 headers everywhere that panel wiring will attach, a la Dotcom. This should help considerably with the soldering mess that the old board had along the top edge, where 42 individual wires were soldered in. Look, it's much cleaner now:

(This photo was taken prior to washing, which is why it appears to have a lot of excess flux; it does. I used organic solder for this part.)

I wound up replacing the transistors and the IC sockets. Really, trying to salvage sockets just isn't worth the trouble; I destroyed one of them during the desoldering, and even though I got the other two out, they had distortions that made them very difficult to insert into the new board. Sockets don't cost much. I also decided to replace all of the transistors after two of them had the base leads break off while I was stuffing them. The replacements aren't the same type; the originals as specified by CGS were BC557, while the replacements are 2N4403. (Why those? Because that's what I had on hand.) All they do in this circuit is drive the LED indicators; almost any PNP type should work for that.

So the basic board is almost ready to go. There are four 470K resistors that go on the board just to the left of the 8-pin socket at bottom center in the first photo. The originals were butchered and I didn't try to salvage them. These resistors form a passive mixer for the four oscillators on the board that don't have the switchable waveform. I want to take those signals off the board at this point, so I need to get new resistors that I can stand up and have a long lead to solder a wire to. This is in pursuit of one of the two mods that I've decided to make: adding a second output bus with mix capability. Below is a block diagram of what I have in mind:

There are six of these, one for each individual oscillator. I'll pick off the four non-waveform-selectable ones from the board as stated above, and I'll intercept the other two at the waveform selection switch common terminal on the panel. The signal will come into this block at left. Each will have a pot to mix it to the B bus. There will also be a switch that allows the signal to be removed from the A bus (the on-board bus). To save some panel space, I decided to use pots with pull-out on-off switch capability to implement the A bus switching. I wasn't going to do that originally because of the cost, but I found some inexpensive ones from All Electronics. Only problem: The switch is SPST and the "on" state is when the knob is pulled out. I want the opposite sense, so to create it, I'm going to invert the switch signal and then use it to switch a bilateral switch that will feed the signal back to the A bus.

The other thing I'm doing, which I haven't drawn yet, is adding control voltage capability. I'll do this by putting a vactrol in series with the resistor that determines the frequency for each oscillator. There will be two CV inputs, an A and a B, and for each oscillator there will be a switch that switches it to the A input, the B input, or neither. I haven't decided yet if I'll make any attempt to compensate for the non-linear response of the vactrol.

I'll build all of the new circuitry on a piece of stripboard. I'll make an auxiliary panel to hold the bus mixing pots, the A/B CV input switches, and the extra jacks. I haven't figured out yet how I'm going to do this. One possibility is getting a blank MOTM-format panel with studs on the back for a board mounting bracket; that would allow me to mount the stripboard to the auxiliary panel without any screws coming through the front. Another possibility is using JB-Weld to glue standoffs to the rear of a panel and mount the stripboard that way, Dotcom-style. The third option is simply to mount the stripboard on the bottom inside surface of the case, behind the panels. However, that would make it harder to move the combined modules later.

I had to order some more 470K resistors from Mouser to replace the ones I messed up. I also had to order additional 100 nF (0.1 uF) polyester box caps; they are used for decoupling, and the rev B board uses more of them than the old board did. I've got these now and I'll be installing them next week.