traderbam said:
Hi Brian,
I liked your suggested set of constraints, but I was also concerned that we needed to do the comparisons on a complete amplifier because some of the other techniques that we wanted to compare HEC to were techniques that would be applied earlier than the output stage, even if they include the output stage.
I think we also want approaches that include the global NFB, since some of the inner loops may affect stability of the outer loop.
In saying this, I am not dismissing the value of looking at techniques that can just be put around a unity gain output stage only.
All of this just underlines how difficult it may be to come up with an apples-apples comparison that most here are comfortable with.
Here's what I would do if I could spend the time right now. As a start, I'd design and simulate three amplifiers based on the same very good input/VAS and output stage. One would just use HEC to get low distortion. The second would use TMC. The third would use NDFL. Later we could explore other variants. I would constrain them to all use a single pair of output devices of the same type biased at the same idle current. I'd probably also arbitrarily constrain them all to have the same global feedback loop gain crossover frequency of 1 MHz.
There are many different variations, but I think once a couple of initial variations are done, the path will be clearer.
Cheers,
Bob
Bob Cordell said:
I certainly call that one arbitrary - with nested loop structures like your feedback EC or others recently mentioned: TMC, OMC, NDFL the outer global feedback loop T doesn't necessarily predict stability properties
This is clearly true of your amp in my sim (previous page), the gain intercept with both output stage enclosing feedback loops cut gives feedback gain 0 dB intercept frequency of 4.3 MHz and 78 dgrees phase margin
jcx said:
I agree completely; the global feedback gain crossover is only one parameter to keep an eye on. Those inner loops in any of the techniques can really get you, and must be watched closely.
However, I think it is still useful to have a common global NFB loop gain crossover for the apples-apples comparisons. Else the guy who pushes it to 2 MHz might get a 6 dB advantage without proving much about the underlying differences among the techniques we are trying to evaluate.
Cheers,
Bob
Limiting the global loop bandwidth. That is sort of restrictive...hmm...why not dispense with it altogether if it is going to be fixed anyhow? Have a dc servo if necessary but concentrate on measuring the OS EC proper.
traderbam said:
Hi Brian,
That's definitely an option, and maybe an exercise that also makes sense. But I still think it best, at least to get started, to evaluate these approaches against each other in a complete amplifier.
Cheers,
Bob
goal-driven design
Bob,
You've already suggested THD20 at 50W average into 8-ohms resistive load as a measure. What other performance measures would you like to see? If any?
Bob,
You've already suggested THD20 at 50W average into 8-ohms resistive load as a measure. What other performance measures would you like to see? If any?
Re: goal-driven design
Hi Brian,
Good question. If I could only pick one more, it would probably be 19+20 kHz CCIF with spectral analysis into 4 ohms. That would line us up pretty well with measurements that JA takes in his reviews.
The other category would be stability measures. At minimum, maybe 100 kHz square wave into 8 ohms at 1 V p-p and something close to clipping.
Cheers,
Bob
traderbam said:
Hi Brian,
Good question. If I could only pick one more, it would probably be 19+20 kHz CCIF with spectral analysis into 4 ohms. That would line us up pretty well with measurements that JA takes in his reviews.
The other category would be stability measures. At minimum, maybe 100 kHz square wave into 8 ohms at 1 V p-p and something close to clipping.
Cheers,
Bob
traderbam said:
I was thinking we should show simulated 100 kHz squarewaves at the output of the amplifier being simulated, both small signal and large signal. The 1 V p-p cited would be the small-signal test at the output. This would allow us to visually assess and compare stability in addition to other metrics that we might be able to show, such as gain and phase margin of the outer loop.
If the amplifier has a closed loop bandwidth on the order of 1 MHz, 100 kHz square waves should be a reasonable test.
Another interesting stability test is to assess the gain margin by seeing how stable the amplifier is when its closed loop gain is reduced by 6 dB.
Cheers,
Bob
Bob Cordell said:
Hi Bob,
May I add to this that it is also important to look at the phase and gain margin of other loops, in particular the ones with a high Ft, like the Miller NFB loop.
Also, one could inspect the stability margins with a capacitive load at the output of say 1....50nF (of course before the Zobel network).
Last but not least, don't forget to simulate the parasitic inductances!
Cheers, Edmond.
Theoretical question
Why does stability matter?
Is the premise not that if circuit A has -10dB THD20 compared with circuit B that circuit A is better regardless of the relative stability margin?
Or is there another premise implied here that says a THD20 figure is somehow not correct or not adequate and that some measure of stability must be used to qualify the THD figure? If so, what is that qualification?
It would seem that the system stability measures suggested are somewhat ambiguous compared to the THD measures. There is a selection of interesting things to look at but not much in the way of specific metrics. For example, with a 100kHz squarewave test what overshoot or settling time is acceptable? What phase margin is acceptable and how should it be measured on the closed loop system as a whole? Should reactive loading be used and what value should it be?
Where's John Curl?
John, what performance measures do you use to evaluate a Parasound amplifier and what limits do you consider acceptable?
Why does stability matter?
Is the premise not that if circuit A has -10dB THD20 compared with circuit B that circuit A is better regardless of the relative stability margin?
Or is there another premise implied here that says a THD20 figure is somehow not correct or not adequate and that some measure of stability must be used to qualify the THD figure? If so, what is that qualification?
It would seem that the system stability measures suggested are somewhat ambiguous compared to the THD measures. There is a selection of interesting things to look at but not much in the way of specific metrics. For example, with a 100kHz squarewave test what overshoot or settling time is acceptable? What phase margin is acceptable and how should it be measured on the closed loop system as a whole? Should reactive loading be used and what value should it be?
Where's John Curl?
John, what performance measures do you use to evaluate a Parasound amplifier and what limits do you consider acceptable?
Re: Theoretical question
Hi Brian,
The real figure of merit is phase-marigin/distortion (or stability/distortion). It's too easy to design an amp with ultra low THD at the expense of the stability and handling of reactive loads.
Cheers, Edmond.
traderbam said:
Hi Brian,
The real figure of merit is phase-marigin/distortion (or stability/distortion). It's too easy to design an amp with ultra low THD at the expense of the stability and handling of reactive loads.
Cheers, Edmond.
Edmond Stuart said:
Hi Edmond,
These are all very good points, and I agree completely.
Cheers,
Bob
Re: Theoretical question
Hi Brian,
Stability margin certainly matters in any fair comparison. Consider two amplifiers, made with technique A and B, and each performaing at 0.001 % THD-20. They both cross at 1 MHz. Amplifier A has 90 degrees of phase margin and 10 dB of gain margin. Amplifier B has 30 degrees of phase margin and 3 dB of gain margin. A could obviously be pushed to a somewhat higher gain crossover, then achieving lower THD-20. This would seem to indicate that in this particular case, Technique A was better.
This is not to say that stability margin is easy to assess and quantify, especially when multiple loops are involved.
The square wave tests can be very revealing, especially when the loop is closed to a gain 6 dB lower. However, the results are not particularly quantitative. One can always say that the peak overshoot not be greater than 10% and that there not be more than one or two decaying cycles of ringing discernable, for example.
Cheers,
Bob
traderbam said:
Hi Brian,
Stability margin certainly matters in any fair comparison. Consider two amplifiers, made with technique A and B, and each performaing at 0.001 % THD-20. They both cross at 1 MHz. Amplifier A has 90 degrees of phase margin and 10 dB of gain margin. Amplifier B has 30 degrees of phase margin and 3 dB of gain margin. A could obviously be pushed to a somewhat higher gain crossover, then achieving lower THD-20. This would seem to indicate that in this particular case, Technique A was better.
This is not to say that stability margin is easy to assess and quantify, especially when multiple loops are involved.
The square wave tests can be very revealing, especially when the loop is closed to a gain 6 dB lower. However, the results are not particularly quantitative. One can always say that the peak overshoot not be greater than 10% and that there not be more than one or two decaying cycles of ringing discernable, for example.
Cheers,
Bob
Re: PGP (Pretty Good Poweramp)
Congrats to Edmond! This will take some time (for me) to digest but is seems a very complete and well thought-out project!
Jan Didden
syn08 said:
Congrats to Edmond! This will take some time (for me) to digest but is seems a very complete and well thought-out project!
Jan Didden
Re: Re: PGP (Pretty Good Poweramp)
Thank you Jan. Don't forget Ovidiu (syn08). He has built it, designed the PCB and did all the measurements. An awful lot of work.
Take your time to grab all the details and feel free to contact me in case of any questions.
Cheers, Edmond.
janneman said:
Thank you Jan. Don't forget Ovidiu (syn08). He has built it, designed the PCB and did all the measurements. An awful lot of work.
Take your time to grab all the details and feel free to contact me in case of any questions.
Cheers, Edmond.