I like that analysis, explains some things I was wondering about, like why the triangle was so perfectly shaped and maxed out at the same time.

Has anyone seen this ciruit before I started publishing it? I'd like to think I invented it, but then, I thought I'd invented the diode bridge when I was in high school.

 

I like that analysis, explains some things I was wondering about, like why the triangle was so perfectly shaped and maxed out at the same time.

Has anyone seen this ciruit before I started publishing it? I'd like to think I invented it, but then, I thought I'd invented the diode bridge when I was in high school.

Bill,

Sorry. This circuit topology has been around for a while. It even pre-dates opamps.

hgmjr

 

I'm typing it all up right now. I'll have it ready in 15-20min.
I love this stuff!

 

I love this stuff too. I look forward to seeing the analysis from another persective.

hgmjr

 

Even multivibrators are predictable, use similar RC formula's, and the feedback on this oscillator is linear, even if the comparator isn't. I would be interested in how you would explain it too.

 

Bill,

Sorry. This circuit topology has been around for a while. It even pre-dates opamps.

hgmjr

The oscillator? What is it's name? I guess I'll have to live with the disappointment.

 

I like that analysis, explains some things I was wondering about, like why the triangle was so perfectly shaped and maxed out at the same time.

Has anyone seen this ciruit before I started publishing it? I'd like to think I invented it, but then, I thought I'd invented the diode bridge when I was in high school.

Bill,

Even though you did not originate the circuit, it is still a feather in your cap to have come up with it on your own without previously encountering the circuit.

Kudos,
hgmjr

 

Okay so that took way longer than I thought. But here it is. It is a bit involved.

 

Thanks for taking the time to put your explanation in words. It is going to take a while to digest your write-up. I do recognize the basic method being used.

hgmjr

 

Hard to say this stuff without a whiteboard

 

Based on your final equation and selecting the opamp conditions of your choosing, what would you estimate the frequency of Bill's circuit to be. For simplicity's sake can we assume whatever opamp you choose is rail to rail?

hgmjr

 

hgmjr,

Frankly, I don't know. And that's my point. The analysis just shows that isn't well-controlled by the linear elements. Simulation has shown 5-15Hz with different op-amps.

But, this is the kind of thing that you characterize rather than predict. They don't exactly spec the non-linear aspects of op-amps very well. I don't know how repeatably they can manufacture them. Or how they respond to temperature changes.

I like this circuit because it has some pretty interesting aspects. But I would never use it in an application. If I ever taught an electronics class, I would give it to them, just to show them the importance of using the linear elements to stabilize a circuit's operation.

 

I bet that one of the conditions that makes this circuit oscillate is the propagation delay that is present in each of the opamps. Without this inherent delay, the circuit would not be able to achieve the 360 degree delay needed to satisfy one of the main criteria for oscillation.

Further, I would be willing to bet that the difference in the delays associated with the various simulation models is a significant contributor to the difference in oscillation frequency observed in the simulations.

hgmjr

 

Now the question is how do you make this circuit have a well defined frequency?

 

I bet that one of the conditions that makes this circuit oscillate is the propagation delay that is present in each of the opamps. Without this inherent delay, the circuit would not be able to achieve the 360 degree delay needed to satisfy one of the main criteria for oscillation.

Further, I would be willing to bet that the difference in the delays associated with the various simulation models is a significant contributor to the difference in oscillation frequency observed in the simulations.

hgmjr

Like I've said before, this is a non-linear circuit. Those criteria (360deg, gain>1) do not apply. Think 555 not colpitts or hartley, etc.

I would say that the circuit oscillates because it is harder to achieve zero than not. Let's say it is not oscillating. Everyone knows that an integrator will drift over time. This will kick it on.

 

OK, being a practical hands on sort lets try an experiment. I'll breadboard two circuits using identical caps and resistors, each with two different op amps. I'll use the clamping diodes and without. One of the op amps is a LM324, you pick the other one (but it has to be a standard op amp, hopefully one I can acquire and not spend a fortune), and I'll get back to you with the frequencies. If my oscope still works I'll look at the waveforms, but that is iffy as it has sat around for a very long time. We'll be able to check the predicted value using hgmjr analysis vs. real measure values. If you want I can even use a variable power supply. The design must be 1Khz or less. It is my belief that the frequencies will stay the same through out and be within acceptable tolerances to predicted values. I plan on using a single ended power supply with a voltage divider for ground.

Basically a technicians approach to designing circuits. I agree it is a multivibrator design, but disagree initial starting condition make that much difference after it has stabilized. The triangle wave forms and pseudo sine wave are just bonus features, much like the sawtooth is on a 555 oscillator.

What say you?

 

Good idea, Bill. We should tack on a couple of other things as well, if you don't mind. You should use opamps that you can simulate. That way you can also check the simulation results. I'm just curious as to how well their models are. Also, getting at least 2 of each op amp would show if the results are curious. I'll even send you the parts if you like. (Yes, I'm that curious ). If they're from Mouser, you should have them by Tuesday with UPS Ground.

I did want to clear something up, though. When I say initial starting condition, I'm not talking about at power up. If you follow the analysis, you will see that I'm talking about the value of the ramp output at state switchover. The analysis just shows that whatever it is this cycle, it will stay that next cycle.
So, there is nothing to set the original value explicitly. And there is nothing in the linear analysis to change it.

 

The LM324 is one of the slowest opamps ever designed (a long time ago).
The oscillator is a "phase-shift" oscillator but it has only two RC circuits instead of the standard of three. The third RC network is the delay of a slow opamp.

 

OK, being a practical hands on sort lets try an experiment. I'll breadboard two circuits using identical caps and resistors, each with two different op amps. I'll use the clamping diodes and without. One of the op amps is a LM324, you pick the other one (but it has to be a standard op amp, hopefully one I can acquire and not spend a fortune), and I'll get back to you with the frequencies. If my oscope still works I'll look at the waveforms, but that is iffy as it has sat around for a very long time. We'll be able to check the predicted value using hgmjr analysis vs. real measure values. If you want I can even use a variable power supply. The design must be 1Khz or less. It is my belief that the frequencies will stay the same through out and be within acceptable tolerances to predicted values. I plan on using a single ended power supply with a voltage divider for ground.

Basically a technicians approach to designing circuits. I agree it is a multivibrator design, but disagree initial starting condition make that much difference after it has stabilized. The triangle wave forms and pseudo sine wave are just bonus features, much like the sawtooth is on a 555 oscillator.

What say you?

Bill,

I think it is worth a shot. I will be happy to assist you anyway I can.

hgmjr

 

Howdy Guru, you're getting in late, but your getting there. Actually the 2 integrators both have a 90 degree phase shift, and the comparator has a 180, no delays from the op amps needed.

I have a plentiful supply of LM324's. What other op amp would you suggest? I also have some 1458's but I suspect they are too similar for the experiment. I can also try voltages using something like 6, 9, 12, and 15 volts. I'll use an DVM in RMS scale if the oscope doesn't work, and probably even if it does, it will be interesting to see if a digital camera can make decent pictures. This oscope is ancient, an HP130 I believe. It will have to be unburied in the garage, something akin to an archaeological dig.

I'll be using a protoboard to do the breadboarding, and when I said identical caps and resistors, I meant reusing the same parts.

This thread is turning out longer than I had anticipated.