I’ve lately been breadboarding new modules before building them, even ones based on designs and PCBs developed by others which presumably have been debugged already. Reasons? One is I think breadboarding helps in understanding the circuit design. Another is testing modifications: Obviously if I have in mind making substantive changes to the design I’m going to want to try them out. Less obviously, I might just be substituting a part. I’ll want to be sure that substitution will work before I solder it in.
Finally, just because a module’s been designed to work the way its creator wants, that doesn’t mean it works the way I want. Maybe I’d prefer different output levels, or different controls sensitivity, or different frequency ranges, or something of that kind. It’s nice to check those things before soldering, and if there’s an issue, then find a fix.
I’m going to be building the MFOS VCO, and I’ve gotten started by breadboarding it.
Mostly I built it per the schematics and BOM. I didn’t bother with the CV1 through CV4 inputs, because the tuning knob is just another CV: if that works, the CV inputs will work.
For U1 I used a TL072. I have the OP275 but I figured I’d rather fry a cheap TL072 if something went wrong.
In the schematic Wilson shows R68 as a 10M resistor, but in a note on the web page he said he tried making it 1M and preferred that. So that’s what I started with; again, see below. This affects the limits of the pulse width control.
U3 is a matched transistor pair, and Wilson used an SSM2210 in a DIP package, but those have been out of production for ten years or so now. Wilson suggested some alternatives, including the SSM2212 in SOIC format, for which he added a footprint on the PCB. That was what I decided to use for the build, but what about breadboarding? I could have just gone with two 2N3904 through hole transistors, but instead I drew up an adapter board for the SSM2212 and a pair of pin headers and had it fabricated. This was the first SMD work I’ve done in nearly a year and only about the fifth SOIC chip I’ve ever soldered. My technique still needs developing but it worked:
Then there was the JFET. Wilson specified PN4391 with NTE457, 2N5457, J112, or similar as substitutes. I had no PN4391 and started off with a 2N5457, but see below.
I got the core working first. Making no attempt made to calibrate the frequency CV, I found the frequency range with coarse tune (fine tune near center) was about 2 Hz to about 13 kHz. It was roughly 1 V/oct. With coarse tune near center frequency was somewhere in the vicinity of middle C.
The linear FM CV worked, though the frequency change with ~4 V was small. There might be some error somewhere but as long as the frequency changes at all I figure it’s “working” and won’t panic until I see how it does on the PCB.
I went on to complete the wave shaping circuitry. All four waveforms (ramp, triangle, sine, pulse) worked. I did only a quick adjustment of the various trimmers, getting just a rough approximation to a sine and seeing some little glitches at the bottom of the triangle (and none at the top where Wilson says to expect them… !?). Good enough for now.
I tested the sync input, using my newly built dodgy bench oscillator. Seems good. Here the controlling oscillator pulse wave is in blue and the MFOS VCO triangle wave is in yellow.
There was a discussion long long ago about improving the sync on the MFOS VCO, and some people said it worked better with a larger input capacitor. But when I say “long long ago” I mean two or three VCO versions ago, and later the sync circuit was changed significantly. Still, I figured it couldn’t hurt to bump the cap up from 1 nF to 10 nF and see what happened. As far as I could tell, nothing did. I put it back to 1 nF.
There was a problem with the pulse width. If I turned the width control almost all the way clockwise the width would narrow to near zero, and beyond that the pulse output would stop, but on the counterclockwise end it would only go down to about 20% width before stopping. This seemed kind of like the problem the R68 mod was supposed to fix, but I tried making it 10M and liked that better. I could get down to near zero width at both ends, although at the top it wouldn’t quite get to the point where the pulse output goes away (a behavior some people dislike, but not Wilson: “I don’t mind this as it lets me set very narrow pulse widths.”) I then tried 3M and liked that even better: On both ends it got down to near zero width and then shut off. I don’t know why the behavior I was seeing was different than Wilson had — maybe it has to do with the ramp and triangle trimmer settings? I didn’t test that — but at least I know how to correct the problem.
Edit to add: Or so I thought yesterday! I looked at it today and it doesn’t get so narrow at one end. Best I can do is maybe 1% to 90%. Granted, Wilson only claims 5% or 10% to 90%, but I think that was with 1M for R68 and I got the impression it improved at both ends with 10M. And I thought that’s what I saw yesterday. But not today. And I did try messing with the ramp and tri trimmers, with little effect. I again switched off between 1M, 3M, and 10M and still liked 3M best. Anyway, maybe breadboard flakiness, maybe I just wasn’t seeing what I thought I was seeing yesterday.
Finally, I tried swapping in various JFETs — anything I could find in my stash — to see which ones worked. I tried 2N5457, 2N5458, J111, J112, and J113. Answer: They all worked, at least in that the oscillator oscillated with them.
Then I took a close look at the core ramp’s falling edge. It looked different with different JFETs. Fall time varied, which I kind of expected, but the voltage it fell to varied too, which I hadn’t. The two most extreme were the J113 (J112 looked just about the same) and 2N5458 (2N5457 looked similar, but not as extreme). J111 was in the middle. Here’s the J113; note the horizontal scale is 2 µs per box:
Fall time is around 1.2 µs. Blowing up the vertical scale I could see it dropped down to about 20 mV. Now here’s the 2N5458:
It takes 3.2 µs to fall, and it only gets down to about 230 mV! I don’t know if either of these is significant. Dropping only to a quarter of a volt seems bad, but in the end isn’t it just an offset and scale? And the offset gets trimmed away. Still, as far as this behavior goes, the J113 clearly does better, whether it matters or not. So I think that’s the JFET I will use, unless I learn some other reason to prefer something else.
Read the rest of the MFOS VCO build series: