So since I’ve built my since whenever I’m using it the oscillators slowly rise in pitch from the heat building up, it usually takes a couple minutes to go up a half step so I was thinking it’s the air inside the case that’s getting hot and has nowhere to go. I’m thinking that I could add a computer case fan and a hole for ventilation to maybe keep the inside cool because I have some leftover 12V case fans. I don’t believe this happens with my megadrone but those oscillators work differently so maybe it wont affect them. I’m going to try to get a video before I do it.
Which VCO design(s) are you using? Sounds like a rather large temperature sensitivity. Fans might help but also may introduce noise.
3 Likes
If it is mainly within the first few minutes of turning on the synth, why not tune it after that? It’s quite natural for any instrument to take a few minutes to warm up. Unless you have stuff running significantly hot, I doubt a fan will do much.
4 Likes
Ok heres a video of whats happening, it starts detuning quickly but after a while it slows down, this was about an hour and a half of recording
The panels are a couple degrees (F) hotter than the rest of the synth, around 76-77 on the oscillators and around 75 everywhere else. Having it off they are all around 74
I see how that can be annoying, but I highly doubt fans will do much here.
A fan works by continually removing the hot air from the surfaces, giving a larger driving force (= temperature difference) for heat transfer from the surface to the air. In this case, there is hardly a driving force, so no hot air to be removed.
I think your best option would be the wait for it to reach a thermal equilibrium , then tune it.
But from the video it looks like it hasn’t reached equilibrium after an hour and a half. That surprises me. Usually VCOs will stabilize fairly well within a few minutes. Maybe there’s something else (non temperature) going on here.
Silicon circuits are temperature sensitive per design; there’s a kT/q factor that keeps appearing in equations…
…so you often see 26 mV (or 19.2) popping up, but the exact value varies with temperature in the real circuit.
However, oscillators are usually designed to compensate for this, either by using components with opposite temperature characteristics in the control voltage circuitry (e.g. tempco resistors) or having different active components pull in opposite directions, and place them in direct contact or on the same silicon (the 3340 uses the latter approach). But there will always be some drift.
(for VCAs and other circuits, there’s usually less compensation and more “deal with it, it’s analog gear”)
But I’m not convinced fans will do much; the thermal effects happen at the junction level, long before the heat reaches the component surface and starts heating up the ambient air.
(you could heat up the entire oscillator (cf OCXO oscillators, or precision circuits with built-in heaters) but that may be a bit trickier to experiment with 
)
2 Likes
Well I’m not so surprised by it taking a long time. Theoretically, it will never reach an equilibrium, just get infinitesimally closer to it. I’m more surprised by how sensitive it appears to be to these small changes. Maybe there’s an unsuitable component in there? Ceramic timing cap maybe?
1 Like
Well, it’ll reach a point where the difference from equilibrium is smaller than fluctuations, at which point the frequency will go randomly up and down. But here there’s definitely an upward trend throughout the hour and a half, slower during the last hour, but definitely trending. I’ve never used one of Sam’s oscillators or indeed any 3340 based so I’m not sure what to expect, but two or three semitones for a 2–3°F temperature increase seems very large.
right? It’s remarkeble to me that all 3 follow the same curve. I don’t have Sam’s VCOs either, but two Frequency Central 3340 oscillators. I must admit I’ve never checked the temperature stability of those. Matter of fact, I hardly ever tune them at all…
1 Like
On paper, the 3340 should be as good as analogue circuitry gets – datasheet says “nearly perfect” and gives a max tempco cancellation error of ±150 ppm (typical 0). But this is the circuit compensating for the thermal effects I mentioned earlier, not temperature drift in external components.
I’d check
- the frequency capacitor, as Tim mentioned above (pin 11 to ground)
- the scale resistor (pin 14 to ground, should be 1.8k)
- the scale adjust resistors (pin 1 and pin 2 to negative supply at pin 3, should typically be near 29.5k and 5.6k)
1 Like
As Tim also points out, all three VCOs show very similar behavior, which suggests it’s not a random bad component but something systematic. The CEM3340 datasheet does say “a low leakage, low tempco capacitor, such as mica, should be used for C_F”. The BOM specifies ceramic. Mica is rather large and rather spendy but I’d think polypropylene would be good. But… The observed drift is, what, about 1% of the frequency? That would imply about maybe 0.5% per K, or 5x10^5 ppm/K — which looks to me like it’s at least 100x worse than what even a cheap ceramic should show, but maybe I’ve miscalculated or misunderstood…
1 Like
High capacity ceramics can be very temperature dependent; class 2 has 0.17%/K on average over the full temperature range, class 3 as much as 1.7%/K over a shorter range, but as pointed out in the comments in the linked article it’s not very linear, so even the better ones exceeding 0.5%/K doesn’t seem that unlikely.
I just wanna say i was only measuring the temperature on the panels facing outside so there could be some other things going on inside, probably getting hotter in many places
ok so i took a module off and put a fan in front of the synth and it slowed the rising down quite a lot, then as soon as I turned the fan off i could see in ableton that the frequency is slowly rising again. If I turn the fan back on the frequency quickly starts to go back down to where it was before (and the waveform seems to be fine, unless the interference would only affect pitch because there are subtle pitch sways from time to time)
Also if its any useful info i have done this
Ok so going further I found that the issue is with C6 the 1nf capacitor, I took a vco out and blowing a fan on it brings the pitch down very quickly and i figured out the bringing a bag of ice near (not touching) that capacitor causes the pitch to go down VERY quickly (yes im being safe not to get things wet i just needed a quick temperature change.) So the C6 cap is the culprit
This is the one connected to pin 11 of the 3340 but im not sure where i would get a better one, the 1nf i got came from tayda along with the rest of my components
1 Like
I don’t see a temperature coefficient specified for any of the polyester film capacitors Tayda carries, but I think 600 ppm/K is typical (see this). Mouser carries 1 nF polypropylene film caps with coefficient -200 ppm/K.
(I updated this to recommend a film cap for C6.)
1 Like
Thanks for your help to all of you! I’m going to order some better capacitors and I’ll let you guys know how it goes.
4 Likes
@lookmumnocomputer maybe you should considering specifying the timing cap as a film cap in the BOM. That should improve temperature stability, as well as exponential tracking
2 Likes