Phase Response and Stability?

[barefoot]

I've designed a class AB buffer than has some fairly nice properties: low distortion, and smooth frequency and phase response roll-off above 10MHz. Now I want to put it inside of a feedback loop, but I'm getting strange results from the simulator.

Here's an example of what I get using the buffer with a NE5532 (the only op amp model I have in my demo software) in a non-inverting configuration. The gain is 12dB and I've adjusted the value of a parallel feedback capacitor to yield the smoothest high frequency roll off with no bumps. But I can never get rid of that positive phase response.

A positive phase response is bad, no? I've tried altering the feedback capacitor as well as the Miller cap, but that just creates more or less of a peak around 7Mhz. The phase response is always positive, no matter what combination I try. The load value doesn't really change things either. Is this a normal and/or stable phase response?

Thanks,
Thomas

 

[Kev]

can you be sure the simulator is correct here ?

you will have to make it to find out. I suspect it's not quite right ... and even if there is a step up at 10Mhz :roll: I'm not sure it will be of concern for your filter/driver.

.... what version of LEAP are you currently on ?

 

[Ian MacGregor]

Keeping in mind that I have just (well almost, final exam is on Wed) completed a class that went over feedback basics: I believe that a positive phase shift will only make things unstable if the gain @ the frequencies that are positively shifted is greater than 1 (or 0dB).

Right?

Also, I'm curious about the application of using an active element in a feedback loop. What is the advantage??

Ian

 

[barefoot]

Hi Kev. I'm running LEAP 5.0, but I'm using Micro-Cap for this amplifier simulation.

I think high frequency stability is always a concern. Even though it's out of the system bandwidth it can still be excited by RF sources and the like. Then you can wind up having your output saturated with HF oscillations.

Anyhow, I just figured out a solution. I removed the two Miller caps across Q5 and Q6 in the buffer and replaced them with four caps across the emitters and collectors of Q7 and Q8. For some reason this increases the bandwidth of the buffer out past 100Mhz. Then I assume this increased bandwidth helps keep the phase roll off negative.

If I had to venture a guess I would say the new found stability might have something to do with the buffer corner frequency being pushed out past the second pole of the op amp open loop gain curve? Maybe a slower op amp would add stability ass well?

 

[barefoot]

[quote author="Ian MacGregor"]Keeping in mind that I have just (well almost, final exam is on Wed) completed a class that went over feedback basics: I believe that a positive phase shift will only make things unstable if the gain @ the frequencies that are positively shifted is greater than 1 (or 0dB).

Right?[/quote] I'm not sure about this, but I did come up with a solution that I described above.

Also, I'm curious about the application of using an active element in a feedback loop. What is the advantage??

Current gain is the major advantage of using a buffer. You can drive a bigger load. Putting the buffer inside the feedback loop yields higher linearity than if it were outside the loop. Another advantage with this circuit specifically is the buffer has a 2x gain. So the op amp stage doesn't need to swing so close to the rails during wide output swings.

 

[bcarso]

You might want to start with the open loop freq and phase response. That is, disconnect the feedback network; put a teeny d.c. voltage at the input of the op amp and adjust until you roughly zero out the output of the buffer (a volt or so is close enough). Then put in a tiny a.c. signal (like a uV or so) in series with the d.c. zeroing signal and look at freq and phase. Strive for the overall gain to be less than unity before the phase shift has accumulated to 180 degrees beyond where it started at low frequencies. Look at the op amp output itself to see how much of the contribution is from it to begin with, and thus how much is being contributed by the buffer.

Most of the time the op amp is working hard to behave like that by itself, so that it will be unity-closed-loop-gain stable. So don't be surprised that even a seemingly fast buffer contributes enough additional phase shift to preclude a simple resistive feedback network with low closed-loop gain being stable.

 

[barefoot]

[quote author="bcarso"]
Most of the time the op amp is working hard to behave like that by itself, so that it will be unity-closed-loop-gain stable. So don't be surprised that even a seemingly fast buffer contributes enough additional phase shift to preclude a simple resistive feedback network with low closed-loop gain being stable.[/quote]
So, does the positive phase indeed cause instability even though the amplitude response isn't peaking? Why?

And am I right in guessing that my new found negative phase response is a result of the wider buffer bandwidth?

Do you think it might help to add a high gain internal feedback loop across the op amp and limit its bandwidth with a parallel feedback cap?

 

[barefoot]

Here are some update graphs.

The green curves result from changing the Miller caps in the buffer like I described above. It also helped to raise the gain a little to 15dB (still workable for my application). And there's no capacitor in the global feedback loop. It only worsened the response peaking near 7Mhz.

Then I added that bandwidth limited internal feedback loop around the op amp. You can see the results in the Red curves. Would you say the Red curves are better?

 

[bcarso]

Can you run a transient response sim? That's really the most revealing---hit it with a square wave.

 

[kvintus]

final exam is on Wed

Good luck!

/Anders

 

[Kev]

[quote author="barefoot"]Hi Kev. I'm running LEAP 5.0, but I'm using Micro-Cap for this amplifier simulation.[/quote]yep
I'm still on an old DOS LEAP
just don't do enough of this to warrant an upgrade

I think high frequency stability is always a concern. Even though it's out of the system bandwidth it can still be excited by RF sources and the like.

yep
can't the usual methods handle this though ?

Maybe a slower op amp would add stability ass well?

yes
I admit I do tend to use this method
I don't feel the need to go beyond the LM833 often

 

[bcarso]

Higher bandwidth is not always a unalloyed blessing---although sometimes it works better to reduce RF sensitivity, somewhat couter to intuition, if it means the interfering signals are just treated as another signal rather than something to be demodulated. OTOH this may have to propagate eventually to some downstream equipment that may well be susceptible itself.

I had a powered speaker design that, for reasons of signal to noise, cost, and distortion had a bunch of discretes in the front end run at fairly hefty currents. It ended up being the first ~cheap product that I knew off that was virtually immune to cell phone radiation in the vicinity.

As far as what to do to the overall circuit, that is, op amp + buffer, if it is unstable, usually a little forced high frequency gain via a judiciously placed external loop RC network will do the trick. This pushes down the loop gain where the phase shift is excessive, while having little effect on the desired closed-loop response. Noise at high frequencies will be higher but will be out-of-band, and as long as it doesn't itself get demodulated downstream will not be audible.

 

[SonsOfThunder]

[quote author="Kev"]I'm still on an old DOS LEAP...just don't do enough of this to warrant an upgrade[/quote]
And I don't think you will be really getting anything except a different interface for your $. Yea, there are some more features, but the models in LEAP are dead on and always have been. Chris had been promising a Win interface since ~1993! I quit paying any attention about 98, I see it took until 2k. Hope you don't get him on the phone for tech support either...his wife is quite congenial however.

[quote author="barefoot"] Maybe a slower op amp would add stability as well?[/quote]
My resident mentor says: "Don't use anything faster than you need".

HTH!
Charlie