Random LFO Experimenting & CMOS Mixing

[Ohioisonfire]

Hello Again,

I have been playing recently with "Random" LFOs in the vein of the Geofex "Pseudorandom LFO" or Ken Stone's "Psycho LFO". Both of these circuits mix the square wave outputs and then RC filter to smooth the waveform. For my purpose I would rather mix the triangle waveform that can be found on the input pin of the inverter for a more linear ramp up and down. I know that sending the triangle wave straight into other circuitry can cause issues with the oscillator if it is not buffered. Can we get away with mixing them together into an inverting op-amp?

This seems to simulate well in LTspice and I will be breadboarding shortly, just wanted to see if I was missing something in theory here. Also very open to other circuits/ideas on random LFOs.

This is cross posted on a couple of the other DIY pedal communities.

See below for rough paint edit of the idea. Please ignore specific values and circuit peripherals.

[ElectricDruid]

I'd have thought it'd be ok as long as the mixer resistor values are "large", whatever that means (560K in your schematic). They terminate at a virtual ground, so they will affect the oscillators, but as long as they don't stop it from working, it's all good, right? It's supposed to be random, so a bit of a tweak here or there isn't the end of the world...
I'd certainly give it a try as drawn, or perhaps try just one oscillator into an inverting op-amp like that and see what component values I could get away with and what I have to do to compensate.

[Edward357Johnson]

Your approach of mixing triangle waveforms into an inverting op-amp for a more linear ramp up and down should work well. Since the triangle wave can cause issues if not buffered, using an inverting op-amp as a mixer is a good idea, as the op-amp can serve as a buffer. If your simulations in LTspice are promising, it indicates the concept is sound. Just be mindful of potential phase issues and ensure proper power supply decoupling to avoid any unwanted oscillations or noise.

[amptramp]

Power supply decoupling is critical because oscillators will tend to sync into each other where edges may shift to occur at the same time even if the frequencies are widely separate.  It could get bad enough that you would need a separate IC package for each oscillator.  CMOS is known for its low current drain when the output is either high or low, but I have measured 8 mA per section when biased toward the  middle of the voltage range.  With one IC, this would be a total of 32 mA.

[ElectricDruid]

In general I'd agree Ron, but in the specific case of CMOS oscillators there's plenty of examples of using a ton of oscillators on the same chip without there (apparently) being any problems with them locking together. The TR808, DR-110 and many other drum machines are examples - they all use massed oscillators for "metallic noise" sounds to create cymbals or whatever. That wouldn't work if they all locked together. So I guess it must be ok, despite the obvious risks?

[PRR]

oscillators on the same chip

In the psycho LFO they are not just same chip, they are cross-connecting the R-C timing networks. They are sure to run different than without the mix network.

hmmmmm.... if the -IN at IC2A is dead-nuts zero impedance, then it is a static load and may not interact much.

560k is not "real large" compared to 220k timing resistor but sure will affect the 1Meg pot at "10" (slow).

Is that 1Meg the "overall speed" used to adjust rate to samba, salsa, or gavotte? Seems odd that it turns to faster (less Ohms) than the fast ends of the fixed pots.

But hey! 40 years old last year! Gotta respect the classics.

[Ksander]

Your approach of mixing triangle waveforms into an inverting op-amp for a more linear ramp up and down should work well. Since the triangle wave can cause issues if not buffered, using an inverting op-amp as a mixer is a good idea, as the op-amp can serve as a buffer. If your simulations in LTspice are promising, it indicates the concept is sound. Just be mindful of potential phase issues and ensure proper power supply decoupling to avoid any unwanted oscillations or noise.

Another AI post? Why tho?!

[PRR]

Another AI post? Why tho?!

Forums get several to dozens of AIs per day. Sometimes they are placeholders, the poster waits a day and then edits the message or the signature with a spam payload. Most get removed quickly.

These spammers are still getting their act together, from selling blue man-pills to off-lease servers.
"I have an interesting project idea... I'm planning to integrate my DIY stompbox with a PeeLiant ML230 server (Octa Core 2.666GHz Xeon) to create a powerful audio processing setup. ....Can anyone recommend robust audio processing software that can run on the PeeLiant ML350 .....any tips on optimizing the PeeLiant ML230....?"  {Specs/model munged to not give a free back-link to these scum.}

In the .signature was a link to a used-server website, open to a listing for off-price pre-abused OctaCore 2.666GHz Xeons.

[Ohioisonfire]

I'd have thought it'd be ok as long as the mixer resistor values are "large", whatever that means (560K in your schematic). They terminate at a virtual ground, so they will affect the oscillators, but as long as they don't stop it from working, it's all good, right? It's supposed to be random, so a bit of a tweak here or there isn't the end of the world...
I'd certainly give it a try as drawn, or perhaps try just one oscillator into an inverting op-amp like that and see what component values I could get away with and what I have to do to compensate.

Appreciate your input! Breadboard is the next step...

Power supply decoupling is critical because oscillators will tend to sync into each other where edges may shift to occur at the same time even if the frequencies are widely separate.  It could get bad enough that you would need a separate IC package for each oscillator.

Am I looking at the "standard" CMOS decoupling here or is additional decoupling needed? Definitely want this thing to be random and glitchy, but don't want syncing at all..

560k is not "real large" compared to 220k timing resistor but sure will affect the 1Meg pot at "10" (slow).

Is that 1Meg the "overall speed" used to adjust rate to samba, salsa, or gavotte? Seems odd that it turns to faster (less Ohms) than the fast ends of the fixed pots.

But hey! 40 years old last year! Gotta respect the classics.

That 1Meg pot can be ignored for my purposes. I will have the oscillators all fixed, and will use all 6 of them, and have a much more interesting idea for controlling "overall speed".

Calling this a classic makes me think again, is there a better or simpler way to do this in analog? It's a lot of parts for what it does..

Another AI post? Why tho?!

How do you know that post was AI? Fairly generic text from an account with only 1 post? I didn't even consider that a possibility..

[duck_arse]

How do you know that post was AI? Fairly generic text from an account with only 1 post? I didn't even consider that a possibility

ohh, the irony. the more ai posts we see, the more we learn to recognise ai posts. and - we aren't about to tell everyone what it is we are seeing before we are deleting, it's not for US to be teaching THEM.



I did a many osc thing w/ cmos invertors. I had a series supply resistor to the cmos, and a single transistor phase shift osc on the same supply, so as the pso drew current, the cmos supply dipped, and the oscillators were all slow modulated. how deep the mod was was controlled by how big the series supply resistor was. but still only small, like 47R.

[ElectricDruid]

Calling this a classic makes me think again, is there a better or simpler way to do this in analog? It's a lot of parts for what it does..

No it's not! 4 oscillators with one chip and an RC for each one?! It doesn't get much more bare-bones than that. Certainly there's no other analog way that would be less parts (that'd be only *one* passive part per-osc!). I could write something similar in PIC code and get it on a single 8-pin DIP (and then get carried away adding extra features... but that doesn't fit your "analog" requirement. That'd be about the only way to get it smaller.