I built a relatively simple circuit - twelve phase shift stages in series, 3 quad op amps, each stage as shown - on perf board, with fairly good layout practice. Power busses were between the DIP IC pin rows, with bypass caps at every op amp (3 of them) and electrolytics at one end, where the power came in. Each stage has short leads, especially the op amp negative inputs, which have 2 0.1 inch wires to two resistors. The input was from another op amp, and the output went to a 100n cap and 100k resistor to ground, for testing purposes. There is no global feedback, yet.

I powered it up, and there was no output for 1v audio input. DC all tested OK, and audio signal tracing showed that the fourth stage sounded a little funny (it shouldn't have any sound until mixed with the input, as it's just fixed phase shift with no gain or loss), the fifth stage was weak and ugly, and the sixth stage and beyond had no audio to speak of. Touching a finger probe to the first, second, or third stage output pin had no effect (which is normal, it's an op amp output and supposedly low impedance). After that, it generated that 'swirly' hissy sound of RF oscillations beating, with worse noise with each further stage. I don't have a scope without dragging a computer into the shop and using its sound card, so I haven't actually eyeballed the oscillations.

Replacing the three TL0xx with pretty much any other op amp in my junk box fixes the problem, including 'difficult' ones like the 5532 (in dual-quad adapters). So it was an easy fix, and not a wiring or layout problem. But I still don't understand it.

What about the phase shifter stage makes it so hard to drive? Or conversely, what is weak about the TL0xx output stage that can't do this job? And why only after four stages in series? Why do the first ones work? Or are they just not exhibiting big enough symptoms?

 

I built a relatively simple circuit - twelve phase shift stages in series, 3 quad op amps, each stage as shown - on perf board, with fairly good layout practice. Power busses were between the DIP IC pin rows, with bypass caps at every op amp (3 of them) and electrolytics at one end, where the power came in. Each stage has short leads, especially the op amp negative inputs, which have 2 0.1 inch wires to two resistors. The input was from another op amp, and the output went to a 100n cap and 100k resistor to ground, for testing purposes. There is no global feedback, yet.

I powered it up, and there was no output for 1v audio input. DC all tested OK, and audio signal tracing showed that the fourth stage sounded a little funny (it shouldn't have any sound until mixed with the input, as it's just fixed phase shift with no gain or loss), the fifth stage was weak and ugly, and the sixth stage and beyond had no audio to speak of. Touching a finger probe to the first, second, or third stage output pin had no effect (which is normal, it's an op amp output and supposedly low impedance). After that, it generated that 'swirly' hissy sound of RF oscillations beating, with worse noise with each further stage. I don't have a scope without dragging a computer into the shop and using its sound card, so I haven't actually eyeballed the oscillations.

Replacing the three TL0xx with pretty much any other op amp in my junk box fixes the problem, including 'difficult' ones like the 5532 (in dual-quad adapters). So it was an easy fix, and not a wiring or layout problem. But I still don't understand it.

What about the phase shifter stage makes it so hard to drive? Or conversely, what is weak about the TL0xx output stage that can't do this job? And why only after four stages in series? Why do the first ones work? Or are they just not exhibiting big enough symptoms?View attachment 194828

UPDATE: This post contains erroneous info and is suitable only for the garbage bin. Sorry to all...
So, you DC coupled 12 stages in-to-out-to-in-to-out... and you are surprised that operation was not perfect? Each opamp has an input offset voltage, and each stage amplifies that input offset by 2 (imagine a voltage source connected in series with the + input of each amp)...how much is 2 to the 12th power? A lot. The same thing happens with the AC noise of each stage. By the time noise of the first stage has passed through the succeeding stages, it's likely to be very high. Recognize that your circuit is not merely shifting phase; it is also amplifying. There are other possible culprits as well including susceptiblity to noise on the power supply rails and the inductance of long wires to power and ground all 12 stages. A star layout (rather than a linear string of amps) would be better but is impossible to easily achieve with breadboard systems. Make a note to yourself: Don't forget about DC and noise effects, just because you are only interested in audio frequency AC.

 

each stage amplifies that input offset by 2

Each stages internal offset is amplified by +2, but the offset from previous stages is amplified by -1.
That's also true for the intrinsic noise.

 

Your schematic is missing the power supply voltages.

 

Each stages internal offset is amplified by +2, but the offset from previous stages is amplified by -1.
That's also true for the intrinsic noise.

Indeed! I stand corrected. @crutschow However, would not the internal/inherent noise of each stage add (not directly, but as RMS) to the noise component of succeeding stages? This would be considerably less than I was hypothesizing due to cascading gains but worse than one stage.

 

So, you DC coupled 12 stages in-to-out-to-in-to-out... and you are surprised that operation was not perfect? Each opamp has an input offset voltage, and each stage amplifies that input offset by 2 (imagine a voltage source connected in series with the + input of each amp)...how much is 2 to the 12th power? A lot. The same thing happens with the AC noise of each stage. By the time noise of the first stage has passed through the succeeding stages, it's likely to be very high. Recognize that your circuit is not merely shifting phase; it is also amplifying. There are other possible culprits as well including susceptiblity to noise on the power supply rails and the inductance of long wires to power and ground all 12 stages. A star layout (rather than a linear string of amps) would be better but is impossible to easily achieve with breadboard systems. Make a note to yourself: Don't forget about DC and noise effects, just because you are only interested in audio frequency AC.

The problem with your theory is that the circuit works perfectly well - stable, no significant DC offset, as low noise and distortion as you'd expect for a dozen low-gain op amps in series, and no oscillations even with ugly loads on more than one op amp output. The failure only occurs with TL0xx series op amps (at least of all the op amps that I own).

It doesn't have DC problems, nor do many other circuits just like it. The offset voltage is trivial. It's a few millivolts, they're FET opamps so bias current doesn't really contribute to offset, and the offset gets cancelled out stage-to-stage because they're unity-gain inverters for DC, (only the stage that generates the offset has a gain of +2 for that offset, it's still +1 for any input voltage). At any rate, in practice it's got maybe 10-20mv of offset, and unity gain for DC all the way through (with functioning op amps).

Aside from that, dozens of successful devices have used the circuit, and more noisy and challenging variations (such as varying the phase shift resistors over a wide range at up to 30Hz), with no DC problems. Likewise for noise - it's a real, functioning circuit that's been in use for decades with no such problems. A "phase shifter" in my own music gear has ten such stages in series, with photoresistors varying over more than a 100:1 range, AND has about 95% global feedback around the circuit if you turn the regen knob up, without any instability problems. It's worked fine since I built it in the 80s, it's not any noisier than any other ten op amps in series (like in a mixing desk, for example), it just doesn't use TL0xx op amps.

Here's the schematic of a commercial device that's been in production more or less since the seventies, with no such problems reported or attributed to it. It's where I got the idea from, although I'm adding fixed stages to a different device rather than building something based on this one. Unfortunately, it's a PDF that I can't repost.
http://www.generalguitargadgets.com/pdf/ggg_p100_sc.pdf

The problems is that TL0xx (TL084, TL074, TL072, TL064) don't work in the circuit, while literally every other op amp I tried, from NE 5532s to 1458s to modern rail-to-rail amps work just fine. Being a unity gain audio circuit, you can even get away with 741s or 324s. It's been done in commercial gear. I don't have any TL022s, or I would have tried them, as the original used them at least part of the time.

Being quad op amps, there's only three of them on the board, and star vs. linear is more or less irrelevant. The power supply leads are six inches long, max.

And what I want to know is what specifically about the performance of the TL0xx series op amp would have allowed me to predict this misbehavior BEFORE building the circuit. Because, as I said, it works fine with other op amps. It's worked fine for many builders and manufacturers. And I should have been able to figure out from the datasheet and the schematic that it wouldn't work with these op amps. But I don't see how.

 

The problem with your theory is that the circuit works perfectly well - stable, no significant DC offset, as low noise and distortion as you'd expect for a dozen low-gain op amps in series, and no oscillations even with ugly loads on more than one op amp output. The failure only occurs with TL0xx series op amps (at least of all the op amps that I own).

It doesn't have DC problems, nor do many other circuits just like it. The offset voltage is trivial. It's a few millivolts, they're FET opamps so bias current doesn't really contribute to offset, and the offset gets cancelled out stage-to-stage because they're unity-gain inverters for DC, (only the stage that generates the offset has a gain of +2 for that offset, it's still +1 for any input voltage). At any rate, in practice it's got maybe 10-20mv of offset, and unity gain for DC all the way through (with functioning op amps).

Aside from that, dozens of successful devices have used the circuit, and more noisy and challenging variations (such as varying the phase shift resistors over a wide range at up to 30Hz), with no DC problems. Likewise for noise - it's a real, functioning circuit that's been in use for decades with no such problems. A "phase shifter" in my own music gear has ten such stages in series, with photoresistors varying over more than a 100:1 range, AND has about 95% global feedback around the circuit if you turn the regen knob up, without any instability problems. It's worked fine since I built it in the 80s, it's not any noisier than any other ten op amps in series (like in a mixing desk, for example), it just doesn't use TL0xx op amps.

Here's the schematic of a commercial device that's been in production more or less since the seventies, with no such problems reported or attributed to it. It's where I got the idea from, although I'm adding fixed stages to a different device rather than building something based on this one. Unfortunately, it's a PDF that I can't repost.
http://www.generalguitargadgets.com/pdf/ggg_p100_sc.pdf

The problems is that TL0xx (TL084, TL074, TL072, TL064) don't work in the circuit, while literally every other op amp I tried, from NE 5532s to 1458s to modern rail-to-rail amps work just fine. Being a unity gain audio circuit, you can even get away with 741s or 324s. It's been done in commercial gear. I don't have any TL022s, or I would have tried them, as the original used them at least part of the time.

Being quad op amps, there's only three of them on the board, and star vs. linear is more or less irrelevant. The power supply leads are six inches long, max.

And what I want to know is what specifically about the performance of the TL0xx series op amp would have allowed me to predict this misbehavior BEFORE building the circuit. Because, as I said, it works fine with other op amps. It's worked fine for many builders and manufacturers. And I should have been able to figure out from the datasheet and the schematic that it wouldn't work with these op amps. But I don't see how.

I cannot disagree with anything you say. As I think you know, @crutschow brought my error to my attention previous to your response. I considered simply deleting my post but decided that what was done, was done. I do thank you for both your information and especially for your even-handed presentation.

 

Your schematic is missing the power supply voltages.

It is. Does it matter? Assuming the part works at that voltage, which it does in hundreds of commercial products, which op-amp parameters vary significantly with supply voltage, aside from output voltage limits? In really old ones, gain would vary, but I think modern designs took care of that.

BTW it was tested on 12v bipolar supplies, a 12v split supply, and an 8v split supply. Given that the maximum signal level is about 1v, every op amp I tried worked, even the 5532, which is just BARELY tolerant of 4v supplies, and really shouldn't be run that way.

 

For what it's worth, my simulator says the input impedance of the stage is 22k, purely resistive, all the way up to like 5MHz or something, where it drops to half that. Unfortunately, this doesn't explain anything or make it easier to understand.

In the simulator, I can also get the TL0xx models to ring by hanging a moderately large cap, like 2.2n to 100n, between the op amp output and ground. But they just ring, not oscillate, and there's nothing like that sort of cap in the circuit.

 

The TL0xx family have FET input stages whereas the 5532 and 1458 have bjt input stages. The difference in behaviour around the turn-on threshold of those stages perhaps accounts for the effect you are seeing?

 

I asked about the power supply voltage because all the Jfet-input opamps that cause your problem have the bad "input phase reversal problem" (where the output suddenly goes as high as it can if an input voltage becomes within the not-allowed negative voltage range which is about 4V above the negative supply).

The problem occurs only with the old Jfet-input opamps but not with other opamps that have a much lower input impedance. Then maybe the 20k input resistor values should be reduced when Jfet-input opamps are used.

 

Assuming the part works at that voltage, which it does in hundreds of commercial products,

And ... what are those products? 12 phase shifters in a row smells like a music synthesizer something. Or not?

To your question - not all unity-gain-stable opamps are created equally stable. The answer might be in the gain/phase plots of the different devices.

ak

 

What about the phase shifter stage makes it so hard to drive?

? a biasing error

 

There are a number of things that must be known that are not covered:
o What is your rail voltage to your OpAmp +Vcc and -Vcc?
o What frequency are you pushing through the OpAmp stages?
o Are you checking your Thermal junction temperatures (calculating for each stage)?

Inputs of an Op Amp are NOT tied to the outputs. It is NOT a direct correlation of energy. Output is a circuit that is influenced by comparison of the inputs. As such, impedance on the output is really a function of how much current that output Vcc can deliver within the max specifications of the OpAmp itself, as influence by the result of the input comparison.

Your problem has nothing to do with the output current being weak. It's all about input signaling. I agree with another poster that likely the problem is related to the JFETs in the TLxxx series OpAmps.

Regarding no o-scope: You should not be trying to do a circuit like this without an O-scope. Just not. Most people in electronics do not have a precise enough understanding of exactly how voltage, current, and impedance work, to be able to recognize what's going on simply by a few simple indicators. A scope will give you _knowns_ not guesses, and could save you untold dizzying mental gyrations, and time.

Rf oscillation on the output- That's what an OpAmp does- it oscillates on the output if any feedback loop is used. Usually it's so fast, and at a frequency that is unnoticed, or negligible, and the magnitude is very small because of the frequency.

Could we see the whole schematic of at least how the OA stages are connected.... perhaps there is an issue in the circuit.