Util_lfo stalls out on low frequency end.

[nickajeglin]

As in the title. If you push it below about 2hz, it climbs up to +5v and stays there. This one is beyond me, so if someone with more electronics knowledge can explain why, that would be awesome.

[nickajeglin]

It's gotta be something related to the feedback into the comparator stage, since it never manages to flip. Maybe the integrstor is capping out at the rail before it gets high enough to trigger the comparator, but why only at slow freq?

[nickajeglin]

Actually, the output swings all the way up to the +12v rail when it stalls. I'm now thinking that R3 may be the culprit.

[nickajeglin]

I messed with R3 a lot. I soldered a potentiometer up in there. I still have no idea. Pls halp.

[nickajeglin]

Shit's fucked, I moved onto a different ckt structure that I didn't ask permission to distribute. If you can explain what's happening here, get in touch, cause I'm still confused. For the moment, I'm putting this on hold though.

[nickajeglin]

Was working on another thing, and I suddenly remembered that phase inversion exists. I have a strong feeling that this is the issue. I'll get back on this when I have some more time and investigate.

[stevemandl]

I'd put money on current leakage across C1 causing this problem. At 1Hz, the capacitive current across C1 is very low. (I=Cdv/dt, 12vpp @1Hz => 24v/s, I= 3.3n*24~80nA) You should expect at least 10 nA leaking around C1 through flux residue and possibly much more if someone is hand soldering an 0805. Have you tried increasing C1 and decreasing the corresponding resistors to keep the same time constant? Replacing with these values should increase stability and keep the same range : C1=330nF, RV10, RV11, R4=10k, R3=2.2. With these values 1uA < I(C1) < 10mA, which should work just fine.

[duckythescientist]

Phase inversion, at least for the TL072, is only an issue if the inputs are outside of the voltage rails which isn't the case here. I agree with stevemandl's assessment that leakage current is overpowering the actual charging/discharging current of C1. I just YOLO'd the kicad_pcb file into OSH Park, so I'll throw one together in a couple weeks and report back.

[duckythescientist]

I put one together as originally specced (1Meg), one where I used 100k pots (C1=33nF, RV10=RV11=R4=100k, R3=22), and one where I used 10k pots (C1=330nF, RV10=RV11=R4=10k, R3=2.2). All three of them still oscillate when turned all the way slow. I still mostly agree with stevemandl. Another source of leakage is the input bias current to the op-amp.

A different possible fix could be increasing R3 to make the slowest setting less slow.

I did however find weird behavior that for the slow settings, the triangle wave became asymmetric. My 1Meg had version faster rise times, and my 100k version had faster fall times. I swapped the op-amps between the builds, and the behavior swapped too. I'm attributing this to the TL072's input bias current being asymmetric.

My general takeaway is that this is good enough for around 20 seconds per cycle and faster, but to consistently go slower, you'll probably need a different circuit.

Here's the waveform of the original 1Meg version:

And here it is after swapping the op-amp with the 100k version:

Here's the original 100k version (10s per division instead of 20s on the other images):

And here it is after the swap:

And here's the 10k for completeness:

[nickajeglin]

@duckythescientist @stevemandl

You guys are rockstars, thanks so much for the help. Leakage on the cap isn't something that I knew was a possibility. I will try changing components as you recommend Steve. Does overheating the cap while hand soldering increase leakage? That might explain why Ducky's would oscillate down to the low limit but mine would stall out. I was pretty inexperienced at hand soldering SMD parts when I made my original run, so I absolutely could have cooked them.

It's been a pretty crazy 18 months, so I've been pretty AWOL here. Haven't had much time or motivation to work on synth stuff lately.