a quest to build a sine+cos LFO

[charonme]

EDIT: result of this quest in this thread: viewtopic.php?t=215780

I found some schematics for a sine+cos LFO using 2 op-amps - is it possible to control the frequency with a single potentiometer without adding more op-amps or transistors and keep the same output amplitude?





I also found this yusynth lfo with CV, but I didn't figure out how to remove the LM13700 and only keep 2 op-amps for sine+cosine generation

[mskala]

Only if it's a dual-gang pot with good tracking between the two gangs.

[charonme]

Only if it's a dual-gang pot with good tracking between the two gangs.

thanks, could you please point me to a schematic (if it exists)?

[mskala]

I'm not absolutely sure this will work (because when I wrote that I hadn't noticed that this is actually a three-pole circuit), but it would make sense to have R1 and R3 be the variable resistances in the schematic you posted. I'd suggest trying that first.

[charonme]

it seem I need to control all three resistors in this particular circuit for it to work
http://tinyurl.com/yay47fxs
who knows how unstable it would be if the resistances differed slightly

also another problem is it swings rail to rail

[Jarno]

I have this (see PDF attached, disregard VCA's as they do not work properly!), but it is meant for only a single phase output. You could use the remaining unused opamp for the cosine output, but you'd still end up with a quad and not a dual opamp.

[charonme]

thanks
I might have to downgrade my expectations, so perhaps a 3 or 4 op-amp circuit would be OK, but I still want a sine and a cosine LFO output controlled with one pot

[devinw1]

thanks
I might have to downgrade my expectations, so perhaps a 3 or 4 op-amp circuit would be OK, but I still want a sine and a cosine LFO output controlled with one pot

http://www.ti.com/lit/ds/symlink/lm13700.pdf

2 OTAs figure 35, put a pot divider on Vc.

[mskala]

2 OTAs figure 35, put a pot divider on Vc. :tu:

If you're willing to use OTAs instead of op amps as such, that's probably the best thing to do. The circuit is basically a VCF circuit driven into oscillation permanently. As an added bonus, it's easy to add a control voltage input, because the pot is basically just being used to generate a control voltage anyway. In other simple circuits where it'd be functioning as a variable resistor, voltage control would be harder to add.

It's because of this issue that my Fixed Sine Bank module is fixed-frequency. One op amp per output, and I can fit eight of them comfortably in 8HP with through-hole construction. When it comes to analog sine LFOs, just being able to change the frequency makes the circuit a whole lot more complicated.

[charonme]

that's exactly what I wanted to avoid :(

[Jarno]

What is the reason to keep it (perhaps too) simple? Do you want to build a massive amount of them, or do you feel that it is hard to do on stripboard? Cost?

[Mungo]

thanks
I might have to downgrade my expectations, so perhaps a 3 or 4 op-amp circuit would be OK, but I still want a sine and a cosine LFO output controlled with one pot

More detail than you ever needed:
http://sound.whsites.net/articles/sinewave.htm
Trying to do it with a dual opamp package is not going to work if you want a large tuneable range, the twin T will work for less than an octave if only small shifts are needed but larger shifts distort the quadrature wave.

[guest]

how perfect of a sinewave do you need? you could go with a standard relaxtion oscillator, which just needs 1 pot to control frequency, and then do waveshaping afterwards to create the 90degree shift and the sinewaves.

[charonme]

What is the reason to keep it (perhaps too) simple?

my skills are pretty basic and I'd like several of them in a tiny space

how perfect of a sinewave do you need?

seeing that a nice sine probably won't be that simple, a diode-limited triangle quadrature LFO would probably suffice. Doesn't need to be perfectly symmetric too, but it has to be smooth

[fitzgreyve]

Had a simple quadrature LFO many years ago, more or less as diagram below. Very difficult to find anything cuirrenty on these on the web?

THis is essentuially a state variable filter that is oscillating.

one stereo pot required: frequency = 1 / 2 pi R C

I'm not sure I've got the feedback and stabilisation correct (lower left components).

[guest]

figure 5.1.2 in the document mungo linked is pretty similar, and this state variable filter approach is probably the lowest parts count (although requires a dual gang pot).

[charonme]

a dual-gang would be OK if the circuit was stable enough even if the gangs were not exactly the same resistance

in a simulation that 5.1.2 circuit slowly rises from 0v to +-1.34v in a couple of minutes: http://tinyurl.com/y7hpnwaf
I guess the max voltage depends on the diodes? And maybe slightly changing the value of some particular resistor might make it rise from 0 to max faster?

[guest]

R6 and R7 determine the rise time and final amplitude. try increasing R7 to 120k and decreasing R6 to 1k. the distortion looks like to goes down, and it rises a bit faster, and the amplitude is much higher.

[charonme]

lower values do seem to make it slightly faster, but still too slow, especially at lower frequencies (and I need this to be a slow LFO). And if the value is too low it starts to clip. I think R6 and 7 just increase the amplitude and with higher amplitude it rises faster.

Maybe if there was a way to inject some voltage somewhere when it powers up?

[Mungo]

Simulation is not the answer, consider Bob Pease on the matter:
https://www.electronicdesign.com/analog ... how-part-i
Plus his famous statement which has been often quoted out of context:
[video][/video]

[guest]

i was wondering about that. i ran a simulation, and saw the same, long settling time, and it doesnt make a lot of sense to me. there isnt anything in that circuit with such a long time constant. and ive put state variable filters into oscillation before and they didnt take forever to settle either. im wondering what is in their (falstad) model that does this, and how much it reflects reality.

[charonme]

I just breadboarded it and the startup time is about 100s (starting with discharged capacitors) with R6 and R7 = 33k, capacitors = 2uF and 4148 diodes. The final amplitude is +-2.5v. The frequency is 0.4Hz with VR1A + R3 = 200k

with VR1A + R3 = 820k the startup time is 6-8 minutes and f=0.096Hz

[guest]

so, i dont claim to understand this completely, but here is my take on the settling time: this is basically a filter being held at a damping factor of exactly zero. the lower the damping factor, the longer it takes to settle, so this one essentially never settles. the diodes act as feedback, and change the damping factor with output amplitude, to keep it at a certain amplitude. at low ampitudes, it has a negative damping factor, and the amplitude increases; at high amplitudes the damping factor increases and the amplitude goes down. if the difference between the diodes being active or not is a small amount, directly around zero, you will get a very low distortion output that that takes a long time settle, and may not start oscillating to begin with. imagine if the damping factor was 0.001, how long it would take to settle, and essentially this diode feedback tries to keep it near zero, so it spends a fair bit of time at low values like that. but, if you were to make the difference between the diodes being active and inactive very large, then the area around zero damping would span a much larger range, and the filter would settle much quicker. unfortunately, this would also add a lot of distortion to the signal. so there is a tradeoff, you can get fast settling, but it will be more distorted.

[charonme]

thanks, btw what exactly do you mean by distortion and how would I go about inducing it? I'm aware lowering R6 and R7 increases the amplitude and of course that causes clipping when it reaches 12v, but otherwise I haven't noticed any other distortion

also, is there some often used way of charging the capacitors when the circuit powes up that doesn't interfere with the circuit afterwards?

[guest]

so, is the issue that it takes too long to stabilize at startup, or that it takes too long to stabilize when you shift frequency, or both? from my simulations, i think those two times were about the same. but, if its just startup time, then you could do some sort of RC switch with a transistor or something.

as for distortion, take a look at the signal across the diodes. the less this looks like a sinewave, the more distortion there will be. the first integrator stage will show more distortion than the second.