Re: On the need for output coils
Hi Bob,
I notice that you suggest a 1 V input.
Seems to me, that we would want to specify this as a voltage relative to the rated power of the amp. So, one condition might be an input voltage that produces an output voltage that is half power into 8 ohms. We might also want to do 10%, 90%, 110%, and perhaps 120% or 150% to cover all cases.
I find it interesting to also observe the VAS output under overload conditions.
Pete B.
Bob Cordell said:
Hi Bob,
I notice that you suggest a 1 V input.
Seems to me, that we would want to specify this as a voltage relative to the rated power of the amp. So, one condition might be an input voltage that produces an output voltage that is half power into 8 ohms. We might also want to do 10%, 90%, 110%, and perhaps 120% or 150% to cover all cases.
I find it interesting to also observe the VAS output under overload conditions.
Pete B.
Re: Re: On the need for output coils
Hi Pete,
Actually, I think I goofed. I think that what I had in mind for this test was a small-signal check of stability (as a start). I'm pretty sure what I meant to say was apply the squarewave to the input so as to obtain 1V at the output.
Of course, it is also good to do it in a large-signal sense, probably both below and above clipping, to see if instability arises as a result of large-signal conditions, slew rate limiting, or clipping.
Cheers,
Bob
PB2 said:
Hi Pete,
Actually, I think I goofed. I think that what I had in mind for this test was a small-signal check of stability (as a start). I'm pretty sure what I meant to say was apply the squarewave to the input so as to obtain 1V at the output.
Of course, it is also good to do it in a large-signal sense, probably both below and above clipping, to see if instability arises as a result of large-signal conditions, slew rate limiting, or clipping.
Cheers,
Bob
PMA said:
Thanks for providing these informative shots. I'm guessing that where you saw the oscillations, it might be where the amplifier came out of slew rate limiting, but I'm not sure. It might also have been coming out of some kind of current limiting situation. Good work. I wonder what the small-signal shots would have looked like.
Cheers,
Bob
Bob Cordell said:
The small signal shown is about 40-45% of the maximum output amplitude. For smaller output signal, the step response starts to look exponential-like. Probably slew rate limitation or dI/dt limitation (as you have already said) causes peaking of the step response.
Regards,
Pavel
Re: Re: On the need for output coils
Quite so, Pete. One must just warn that sometimes the scope probe can cause an unsuspected influence, even with the 2 odd pF of a X10 probe. I check that by also looking at the output or some other low impedance point on a second trace to notice what changes when the first probe is attached. The VAS can be a sensitive point, also depending on where the probe earth goes. As said before, I was sometimes led astray by perturberances resulting purely from probe placement, especially in high bandwidth designs.
PB2 said:
Quite so, Pete. One must just warn that sometimes the scope probe can cause an unsuspected influence, even with the 2 odd pF of a X10 probe. I check that by also looking at the output or some other low impedance point on a second trace to notice what changes when the first probe is attached. The VAS can be a sensitive point, also depending on where the probe earth goes. As said before, I was sometimes led astray by perturberances resulting purely from probe placement, especially in high bandwidth designs.
PMA, I see you use the same model scope camera I do ;-) Jim Williams has some good things to say about probes in the Linear Technology Application Notes books, and in his various other books and articles. Because I move my probes between different plug-ins, I'm constantly forgetting to re-compensate them, and wondering why my waveform looks so messed up. If probe influence is a problem, maybe an FET probe would be better. I don't know their maximum voltage, but I've seen some articles to build one quite easily. I have an old Tek FET probe, but I've never tried it with audio stuff. That HF oscillation is a bit worrying, as that's what my Tiger amp would do before I sorted out the grounding and stability. It would be very interesting to see if that sort of thing is ever provoked while playing music into a real speaker load. My guess is that it isn't, since your load is so severe. In some other thread I suggested an RF detector, but I think just building a steep 20kHz hipass filter and using the scope would be sufficient. Knock out the audio, crank up the scope gain, and look for things that shouldn't be there. A four pole CLCL filter with a 1K series input resistor and 1K output resistor to ground looks good- going from in to out, 2.3 nF, .957 mH to ground, .957 nF, 2.3 mH to ground. I haven't tried this yet, but it looks good on the simulator.
Something funny happened
Something funny happened, or the spirit of Murphy scrambling things as allways.
On 5/27 Andy_c uploaded to the now defunct "audibility of output coils thread" his derivation for the effective output impedance of a given amplifier with negative back with a given OL output resistance and closed loop bandwidth. I then offered to transcribe to pdf which I did, and exchanged with Andy for proofreading, then nothing happened... I kept wondering why he did not upload ... until it dawned on me he was probably waiting for me to do the same !!!
Whatever, here it is.
Something funny happened, or the spirit of Murphy scrambling things as allways.
On 5/27 Andy_c uploaded to the now defunct "audibility of output coils thread" his derivation for the effective output impedance of a given amplifier with negative back with a given OL output resistance and closed loop bandwidth. I then offered to transcribe to pdf which I did, and exchanged with Andy for proofreading, then nothing happened... I kept wondering why he did not upload ... until it dawned on me he was probably waiting for me to do the same !!!
Whatever, here it is.
I was also waiting to see if the "Audibility of Output Coils" thread would open back up again later.Re: On the need for output coils
I think that one thing that needs to be mentioned about the NEED for output coils, is the fact that a typical amplifiers susceptibility to capacitive loads is directly related to the amplifiers output impedance before the application of global negative and the frequency compensation.
Theoretically, if an amplifier has an output impedance of zero prior to the application of global negative feedback, it will be able to drive any load capacitance without a loss of operating phase margin. In a high power amplifier with lots of output pairs connected in parallel, the output impedance will likely be very low, making a load-isolating coil/network superfluous as far as stability is concerned.
Suppose you have a low- medium power amplifier with only one or two output pairs, with an output resistance before the application of global negative feedback of 0.1 ohms. A 1uF load will incur a phase shift of about 52 degrees at 2MHz.
If such an amplifier has a gain crossover of 200kHz, the effects of a capacitive load will be totally negligible, but if the gain crossover frequency is 2MHz, the phase margin will be reduced significantly.
This example is idealise a bit, because it is unlikely that an output stage with an intrinsic output impedance of 0.1 ohms in the audio band will still be 0.1 ohms at 2MHz, or during high peak current load demands when BJT hfe or MOSFET gm can drop off.
Cheers,
Glen
Bob Cordell said:
I think that one thing that needs to be mentioned about the NEED for output coils, is the fact that a typical amplifiers susceptibility to capacitive loads is directly related to the amplifiers output impedance before the application of global negative and the frequency compensation.
Theoretically, if an amplifier has an output impedance of zero prior to the application of global negative feedback, it will be able to drive any load capacitance without a loss of operating phase margin. In a high power amplifier with lots of output pairs connected in parallel, the output impedance will likely be very low, making a load-isolating coil/network superfluous as far as stability is concerned.
Suppose you have a low- medium power amplifier with only one or two output pairs, with an output resistance before the application of global negative feedback of 0.1 ohms. A 1uF load will incur a phase shift of about 52 degrees at 2MHz.
If such an amplifier has a gain crossover of 200kHz, the effects of a capacitive load will be totally negligible, but if the gain crossover frequency is 2MHz, the phase margin will be reduced significantly.
This example is idealise a bit, because it is unlikely that an output stage with an intrinsic output impedance of 0.1 ohms in the audio band will still be 0.1 ohms at 2MHz, or during high peak current load demands when BJT hfe or MOSFET gm can drop off.
Cheers,
Glen
Re: Re: On the need for output coils
Hi Glen,
This is a very good point, and something that I was thinking along the lines of as well. Here's maybe another way to look at it. Suppose back in the old days we had a 100 watt amplifier with one or two 2 MHz BJT output pairs in only a single-Darlinton configuration. Back then, maybe it needed 4.7 uH and 4.7 ohms.
Now, if we just apply some back-of-the-envelope scaling to it, looking at the open-loop output impedance, we get a different picture. Maybe now we have a triple Darlinton feeding 5 or 10 output pairs in a bigger amplifier. And maybe we are using faster output devices. And maybe we are not using a really high gain crossover frequency. All of those things would seem to push us in the direction of needing either a much smaller coil, or no coil at all.
It is also fair to say that an output stage that maintains its low open-loop output impedance to a higher frequency can probably use a smaller coil, if it needs one at all. This might be the case with output stages using ring emitter transistors.
In all of this, of course, we must still make very sure that there is no opportunity for a local oscillation as a result of the capacitive load as well.
Cheers,
Bob
Hi Glen,
This is a very good point, and something that I was thinking along the lines of as well. Here's maybe another way to look at it. Suppose back in the old days we had a 100 watt amplifier with one or two 2 MHz BJT output pairs in only a single-Darlinton configuration. Back then, maybe it needed 4.7 uH and 4.7 ohms.
Now, if we just apply some back-of-the-envelope scaling to it, looking at the open-loop output impedance, we get a different picture. Maybe now we have a triple Darlinton feeding 5 or 10 output pairs in a bigger amplifier. And maybe we are using faster output devices. And maybe we are not using a really high gain crossover frequency. All of those things would seem to push us in the direction of needing either a much smaller coil, or no coil at all.
It is also fair to say that an output stage that maintains its low open-loop output impedance to a higher frequency can probably use a smaller coil, if it needs one at all. This might be the case with output stages using ring emitter transistors.
In all of this, of course, we must still make very sure that there is no opportunity for a local oscillation as a result of the capacitive load as well.
Cheers,
Bob