Hi Jaka
The idea is even older than the Carver hybrid amp. It is used almost everywhere in feedback control.
Because you don't have to compensate for the output filter's phase-shift up to the VHF range it is wise to use a resistor in series with the feedback cap.
It is theoretically possible to get an almost perfect squarewave output after the filter. I.e. not only doesn't it have overshoot, also the rise-and fall times can be very short if the feedback loop is adjusted properly (they can't be zero of course, the reason for this is explained next).
But there is a huge disadvantage: In order to get steep transitions after the filter, the transition before the filter has to be even steeper.
That means such an amp is very susceptible to TIM.
So it is always better to restrict input signal slew-rate to reasonable values by the use of an input lowpass. Furthermore you can get into aliasing problems when too much RF content is getting into your amp.
Regards
Charles
The idea is even older than the Carver hybrid amp. It is used almost everywhere in feedback control.
Because you don't have to compensate for the output filter's phase-shift up to the VHF range it is wise to use a resistor in series with the feedback cap.
It is theoretically possible to get an almost perfect squarewave output after the filter. I.e. not only doesn't it have overshoot, also the rise-and fall times can be very short if the feedback loop is adjusted properly (they can't be zero of course, the reason for this is explained next).
But there is a huge disadvantage: In order to get steep transitions after the filter, the transition before the filter has to be even steeper.
That means such an amp is very susceptible to TIM.
So it is always better to restrict input signal slew-rate to reasonable values by the use of an input lowpass. Furthermore you can get into aliasing problems when too much RF content is getting into your amp.
Regards
Charles
Hi Charles,
I am not sure you are correct here. You can take my circuit and simulate it with and without the feedback cap. You will see that it has the same two pole frequency response, only the resonance is gone. There is no zero introduced at resonant frequency to compensate for two pole phase shift. It more likely feeds back image of filter capacitor resonant current into the summing junction of the modulator thus canceling the resonance. You can see a lenghtier explanation in this Carver patent. I think they would not patent it without being novel.
Also output slew rate of the switching amplifier is fundamentaly limited by output filter inductance and supply voltage and no feedback manipulation will make it any faster.
Regarding input filter we are on the same side, I only thought it is not necessary for ringing compensation.
Best regards,
Jaka Racman
I am not sure you are correct here. You can take my circuit and simulate it with and without the feedback cap. You will see that it has the same two pole frequency response, only the resonance is gone. There is no zero introduced at resonant frequency to compensate for two pole phase shift. It more likely feeds back image of filter capacitor resonant current into the summing junction of the modulator thus canceling the resonance. You can see a lenghtier explanation in this Carver patent. I think they would not patent it without being novel.
Also output slew rate of the switching amplifier is fundamentaly limited by output filter inductance and supply voltage and no feedback manipulation will make it any faster.
Regarding input filter we are on the same side, I only thought it is not necessary for ringing compensation.
Best regards,
Jaka Racman
What I am talking of is a so called PID controller. It is used to control (and damp) arrangements with 2nd order behaviour. The capacitor in question is responsible for the "D" in PID (though a proper PID would have another one at it's input as well).
As far as slew-rate goes: It is dependant on frequency-response of the output filter O.K.
If you compensate for he response of the output filter (what a properly designed feedback circuit could do to some extent) then the output signal would be exactly as the input signal.
Or more exactly: It would try to follow the input signal as exactly as possible.
If we now feed our arrangement a square-wave it will desperately struggle to achieve this signal at the output. Since an ideal squarewave has already infinitely steep slopes (and still very steep ones for a "real-world" squarewave) this wouldn't be achievable and the whole thing runs into a transient overload situation.
That's one reason for adding a low-pass at the input of a class-d amp.
Regards
Charles
As far as slew-rate goes: It is dependant on frequency-response of the output filter O.K.
If you compensate for he response of the output filter (what a properly designed feedback circuit could do to some extent) then the output signal would be exactly as the input signal.
Or more exactly: It would try to follow the input signal as exactly as possible.
If we now feed our arrangement a square-wave it will desperately struggle to achieve this signal at the output. Since an ideal squarewave has already infinitely steep slopes (and still very steep ones for a "real-world" squarewave) this wouldn't be achievable and the whole thing runs into a transient overload situation.
That's one reason for adding a low-pass at the input of a class-d amp.
Regards
Charles
Well, well, well- this "input filter" is not so simple...Tonight (actually all this night) i was doing measurement 8938 based amp by PC sound card (0.004%THD), as "input filter" used 1st order LP 30khz, BTW it's reduced THD in 2 times! Correlation with MC7 simulations is good for switching frequency etc, but regarding THD is more bad. Bandwidth for measurement are 22khz:
THD at 1khz, IMD 18khz&19khz, 0db are almost clipping.
0db THD-0.1%, IMD-0.07%
-3db THD-0.07%, IMD-0.04%
-15db THD-0.015%, IMD-0.02%
THD is almost independent of frequency . I have sensing that for this "input filter" very fine adjusting is required (in dvx-s200 board it contained 5res&3cap) .
Best regards Ivan.
edited:
in my post #34 DP are PD (like PID without I).
THD at 1khz, IMD 18khz&19khz, 0db are almost clipping.
0db THD-0.1%, IMD-0.07%
-3db THD-0.07%, IMD-0.04%
-15db THD-0.015%, IMD-0.02%
THD is almost independent of frequency . I have sensing that for this "input filter" very fine adjusting is required (in dvx-s200 board it contained 5res&3cap) .
Best regards Ivan.
edited:
in my post #34 DP are PD (like PID without I).
Thanks Jaka Racman and Ivan for your honorable mention. The phase compensation used to prevent overshoot I discovered while trying to solve an overshoot problem in my switching power supply. Those types of things get rediscovered all the time, and I am probably the last to do so. When I discovered the synchronous rectifier, I was unaware that anyone else had discovered it until someone else mentioned it. I practically literally live in a cave.
I noticed that lower distortion often results when the compensation capacitor is as small as possible. But since I am lazy, I may just build an amp without much input filtering and add more phase compensation. Besides, it is important for the amp, without input filtering, to pass the square wave test into an inductive load to be sure that in real life with real music it does not ever exhibit any strange behavior.
I noticed that lower distortion often results when the compensation capacitor is as small as possible. But since I am lazy, I may just build an amp without much input filtering and add more phase compensation. Besides, it is important for the amp, without input filtering, to pass the square wave test into an inductive load to be sure that in real life with real music it does not ever exhibit any strange behavior.
Since there will not be any fast current changes in an inductive load, this same feature might also be tested without any load at all (which is also a very critical load condition for some switching amps). Have you tried to do this test without any load ? How does it correlate with the inductive-load measurement/behaviour ?
Those types of things get rediscovered all the time, and I am probably the last to do so. When I discovered the synchronous rectifier, I was unaware that anyone else had discovered it until someone else mentioned it. I practically literally live in a cave.
That does only mean that you will not be the one who is commercially successful with a specific idea (this is not even guaranteed for the one who owns the patent !).
But it is always a sign of skill and intelligence when you independantly come to an already known solution.
Regards
Charles
Thx Charles. After some additional thought and after benefiting from the input of your lucid insight, I think an even better way to test an amplifier's behavior into a load may be to make it clip a sine wave. This test can be rather convenient since a 60 (50) hz frequency may be able to provide useful results. Simply increasing the amplitude increases the steepness the rise and fall slopes.
This point is very good. I felt there was something missing from my last post. The amp as a closed system is very relevant. Resonance of the output filter is a crucial factor to watch in switching amps, even more so than the speaker inductance which actually has a series damping resistance. I simulated a switching amp according to your suggestion and noticed my oversight. Then the dominance of the output filter was apparent.
This point is very good. I felt there was something missing from my last post. The amp as a closed system is very relevant. Resonance of the output filter is a crucial factor to watch in switching amps, even more so than the speaker inductance which actually has a series damping resistance. I simulated a switching amp according to your suggestion and noticed my oversight. Then the dominance of the output filter was apparent.
It is still not 0.002%, but were dramatically improved THD & IMD performance 8938 amp
by resistors value reducing at 10 times (obviously, that caps was increase).
Almost at the clipping.. THD@1KHZ;4OHM;400KHZ are 0.045%, 5khz are 0.033%,
7khz are 0.028%, 10khz are 0.009%(22khz bandwidth and only 2
harmonic like AES17),
for -3db 1khz are 0.028%, 5khz are 0.02%, 7khz are 0.02, 10khz 0.015%.
IMD at 18khz vs 19khz for -3db are 0.02%, for -6db are 0.015%, for -10db are 0.013% too.
My previous choice of value for passive components
has been high sensitive for EMI products from itself.
And now i was trying listening it vs my old linear mosfet amp (200ma biased AB@0.003-0.005%THD;IMD=?).
Well.. different in sound does exist. If you just once used any "modulating" musical studio effect processing(LFO is OFF) like chorus, phaser etc, then you must can understand me exactly. I didn't know it earlier, but now in my old amplifier it seems there are 3-5 % of this effect.. New amp is haven't it. This is more possible to listen on the rich of harmonics sounds like human voices "AAH", violin etc, but this is almost not audible at like noise distribution -Hi hat, snare etc.
Edited:
left channel is new , right is old amp.. after some time it reversed right to left etc.
by resistors value reducing at 10 times (obviously, that caps was increase).
Almost at the clipping.. THD@1KHZ;4OHM;400KHZ are 0.045%, 5khz are 0.033%,
7khz are 0.028%, 10khz are 0.009%(22khz bandwidth and only 2
harmonic like AES17),
for -3db 1khz are 0.028%, 5khz are 0.02%, 7khz are 0.02, 10khz 0.015%.
IMD at 18khz vs 19khz for -3db are 0.02%, for -6db are 0.015%, for -10db are 0.013% too.
My previous choice of value for passive components
has been high sensitive for EMI products from itself.
And now i was trying listening it vs my old linear mosfet amp (200ma biased AB@0.003-0.005%THD;IMD=?).
Well.. different in sound does exist. If you just once used any "modulating" musical studio effect processing(LFO is OFF) like chorus, phaser etc, then you must can understand me exactly. I didn't know it earlier, but now in my old amplifier it seems there are 3-5 % of this effect.. New amp is haven't it. This is more possible to listen on the rich of harmonics sounds like human voices "AAH", violin etc, but this is almost not audible at like noise distribution -Hi hat, snare etc.
Edited:
left channel is new , right is old amp.. after some time it reversed right to left etc.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.