Hi Mike,
Yes, I can see from your findings that there is a good phase response for the Post #40 output stage in isolation, but each output half has within it an additional common emitter stage, and without applying additional response slowing 'local stability' precautions these can cause each output half to parasitically oscillate in their own right within a closed loop due to reactive loading.
The best way to see the phase change I mention is to open loop simulate your output stage within a complete amplifier circuit, and then compare its high frequency phase response with that of the same amplifier using an option 'A', 'B' or 'C' output stage.
You will find the compound complementary version needs more stabilisation.
I am not saying that you cannot make a powerful amplifier using a compound complementary architecture, but there is a higher risk of impaired sound reproduction when driving reactive loudspeakers because the additional stabilisation propagation delay is likely to have a greater impact upon NFB loop control.
Hi Mikeks,
I was only lumping pre/driver together in the presented example when I stated that I could not see constant class-A drive in version 'B'; which I illustrated could be the case.
Actually, I was surprised at being able to simulate the output drivers dropping out of class-A without any need to drive beyond 'AF' frequency, or into clipping, but then I did use voltage drive in my illustration.
I found that I did not need to simulate using a virtual loudspeaker load, which can lead to an even greater build up of output base region stored charges within a NFB loop controlled output stage, and which might still lead to class-A dropout when compared to loading resistor values that are okay on the simulator or test bench.
That is - the NFB loop itself can still induce internal *non-linear* operation within what could be a region of completely linear operation using another topology to drive real-world loudspeakers.
The closed loop performance of the 'B' version might be superior .........
but does that 'closed loop', which must necessarily envelop stabilisation components and/or be followed by a series output choke, actually reproduce music better ...........
________________________________________________________________
This discussion also gives rise to the question -
Should not the triple arrangement always be used in a 'follower' output stage ?
I note that MikeB has been trying a double where originally he had a triple in his
http://www.diyaudio.com/forums/attachment.php?s=&postid=584704&stamp=1109419992
Cheers ........... Graham.
Yes, I can see from your findings that there is a good phase response for the Post #40 output stage in isolation, but each output half has within it an additional common emitter stage, and without applying additional response slowing 'local stability' precautions these can cause each output half to parasitically oscillate in their own right within a closed loop due to reactive loading.
The best way to see the phase change I mention is to open loop simulate your output stage within a complete amplifier circuit, and then compare its high frequency phase response with that of the same amplifier using an option 'A', 'B' or 'C' output stage.
You will find the compound complementary version needs more stabilisation.
I am not saying that you cannot make a powerful amplifier using a compound complementary architecture, but there is a higher risk of impaired sound reproduction when driving reactive loudspeakers because the additional stabilisation propagation delay is likely to have a greater impact upon NFB loop control.
Hi Mikeks,
I was only lumping pre/driver together in the presented example when I stated that I could not see constant class-A drive in version 'B'; which I illustrated could be the case.
Actually, I was surprised at being able to simulate the output drivers dropping out of class-A without any need to drive beyond 'AF' frequency, or into clipping, but then I did use voltage drive in my illustration.
I found that I did not need to simulate using a virtual loudspeaker load, which can lead to an even greater build up of output base region stored charges within a NFB loop controlled output stage, and which might still lead to class-A dropout when compared to loading resistor values that are okay on the simulator or test bench.
That is - the NFB loop itself can still induce internal *non-linear* operation within what could be a region of completely linear operation using another topology to drive real-world loudspeakers.
The closed loop performance of the 'B' version might be superior .........
but does that 'closed loop', which must necessarily envelop stabilisation components and/or be followed by a series output choke, actually reproduce music better ...........
________________________________________________________________
This discussion also gives rise to the question -
Should not the triple arrangement always be used in a 'follower' output stage ?
I note that MikeB has been trying a double where originally he had a triple in his
http://www.diyaudio.com/forums/attachment.php?s=&postid=584704&stamp=1109419992
Cheers ........... Graham.
mikeks said:
Hi mikeks,
thanks for the comments. Underbiasing also problem for the crossover distortion of the output device, so I think that everybody set the minimum bias, at least.
Would You please clarify the cross coupling resistor?
Another idea: I think that the base stopper resistors used series with base of the ouput devices (1-10ohms) can help too! And this resistors helps to avoid the oscillation in the driver stage.
And we can define the minimum necessary driver bias for different type of ouput transistors, to keep the drivers in class A.
sajti
May I draw your attention to the output stage of "The Ultimate Suplifier" (see picture) of which the driver (and the output transistor) is also working in class A. For explanation I refer to my website: www.suplifier.nl.
Marc.
http://www.suplifier.nl
Marc.
http://www.suplifier.nl
Attachments
Sorry Upupa Epops, I was not thinking of antisaturation circuits but I don't understand your questions.
Marc.
http://www.suplifier.nl
Marc.
http://www.suplifier.nl
Hi,Graham Maynard said:
Thanks for the analysis. Similiarly a two transistor CFP output
stage alone has a additional common emitter stage hidden within
it. So as a general rule any stage with a CFP would be harder to
stabilize then one of the "A,B, or C" Darllington stages?
The reason for going to the trouble of compounding a stage
would be increased linearity. Lots of local feedback to linearize
the stage. This means that overall less global feedback
would be needed. Similiarly running in Class A.
I was thinking of a design with no global feedback, but lots and
lots of short local feedback loops.
Mike
Hi Mike, ( I am musing here, not lecturing.)
My understanding is that an output stage using any CFP arrangement introduces more hf phase change.
Within a NFB amplifier this means more complex stability circuitry is required; stability componentature that cannot fail to affect music reproduction, especially transient impulses and non-sinewave asymmetrical waveforms.
Is this not why Nelson has made such effort with his balanced/passive class-As, and Susan with her zero global nfb amplifier?
If there is no global feedback then the CFP has advantage because it still responds so very much faster than music/hearing, but if we do not use global NFB then we cannot damp unwanted crossover and dynamic loudspeaker resonances that arise within the audible spectrum (reactions that allow the majority of loudspeakers to colour reproduction more than they would when driven by a 'damping' amplifier) unless we accept class-A inefficiencies; a situation that is not sensible when more than say 25W of output is required.
It is essential that the output terminal potential of an amplifier cannot significantly 'react' to dynamic loudspeaker generated back emfs,
ie. the amplifier must not allow the loudspeaker system to generate, nor the amplifier itself actually generate within its own circuitry (due to NFB loop sensed hf error correction and possibly non-linear current flow in signal path parallel stabilisation components and device capacitances) an error potential at its output terminal that is momentarily out of phase with respect to the input waveform, and this has absolutely nothing to do with sinewave thd measurements observed in phase shifted isolation at the output terminal.
This is because the dynamically generated loudspeaker back emf components can have quite different impulse and resonant current to voltage components to those due to the original energising and ongoing loudspeaker driving waveform, and thus amplifier-loudspeaker reactivity gives rise to new and audible components that can far exceed those normally considered with sinewave - FFT/resistor load examination etc. etc. etc........
Cheers ........... Graham.
My understanding is that an output stage using any CFP arrangement introduces more hf phase change.
Within a NFB amplifier this means more complex stability circuitry is required; stability componentature that cannot fail to affect music reproduction, especially transient impulses and non-sinewave asymmetrical waveforms.
Is this not why Nelson has made such effort with his balanced/passive class-As, and Susan with her zero global nfb amplifier?
If there is no global feedback then the CFP has advantage because it still responds so very much faster than music/hearing, but if we do not use global NFB then we cannot damp unwanted crossover and dynamic loudspeaker resonances that arise within the audible spectrum (reactions that allow the majority of loudspeakers to colour reproduction more than they would when driven by a 'damping' amplifier) unless we accept class-A inefficiencies; a situation that is not sensible when more than say 25W of output is required.
It is essential that the output terminal potential of an amplifier cannot significantly 'react' to dynamic loudspeaker generated back emfs,
ie. the amplifier must not allow the loudspeaker system to generate, nor the amplifier itself actually generate within its own circuitry (due to NFB loop sensed hf error correction and possibly non-linear current flow in signal path parallel stabilisation components and device capacitances) an error potential at its output terminal that is momentarily out of phase with respect to the input waveform, and this has absolutely nothing to do with sinewave thd measurements observed in phase shifted isolation at the output terminal.
This is because the dynamically generated loudspeaker back emf components can have quite different impulse and resonant current to voltage components to those due to the original energising and ongoing loudspeaker driving waveform, and thus amplifier-loudspeaker reactivity gives rise to new and audible components that can far exceed those normally considered with sinewave - FFT/resistor load examination etc. etc. etc........
Cheers ........... Graham.
To Marc : By high overload, when predriver transistors ( and output transistors too ) are fully saturated, very uprise " bias " current. This effect is very distinctive by CE connection. Sometimes I had on table Bryston 4 SST, which use this connection ( combination of both, CC + CE ), where at 1 kHz was all OK ( by clipping ), but at 10 kHz in this case power consumption " jumply " uprise from 4 A ( on AC side ) to 8 A ( ! ), only by " touching " of clipping. In this amp wasn't antisaturation circuity, so I'm asking your experinces 
.
.
Mike,
You were “thinking of a design with no global feedback, but lots and lots of short local feedback loops”. See "The Ultimate Suplifier" on my website. If you consider an emittor follower as a local feedback loop (which I don’t), then The Ultimate Suplifier has many short local feedback loops. Because The Ultimate Amplifier is a bridge tied load (BTL) amplifier, it consists of a left and a right amplifier. If you consider a CFP as a local feedback loop (which I do) then The Ultimate Suplifier has on the upper part of the left amplifier 3 local feedback loops and on the lower part of the left amplifier also 3 local feedback loops. Same for the right amplifier, so totally 12 local feedback loops.
Graham,
“The Ultimate Suplifier” is a balanced input, DC coupled, symmetrical, bridge tied load, class A amplifier with no global feedback using transistors. In a amplifying mode (voltage or current) the transistors are composed in a CFP configuration. The output impedance of a CFP is low so the damping is high. I do agree that the damping factor is lowered down by the output resistors of 0,22 ohm. But, because of the CFP configuration in class A without global feedback, the amplifier will be insensible to the generated back emfs of the loudspeaker. The loudspeaker itself will swing out, but that is something different.
Upupa Epops,
I have no such experience.
Marc.
http://www.suplifier.nl
You were “thinking of a design with no global feedback, but lots and lots of short local feedback loops”. See "The Ultimate Suplifier" on my website. If you consider an emittor follower as a local feedback loop (which I don’t), then The Ultimate Suplifier has many short local feedback loops. Because The Ultimate Amplifier is a bridge tied load (BTL) amplifier, it consists of a left and a right amplifier. If you consider a CFP as a local feedback loop (which I do) then The Ultimate Suplifier has on the upper part of the left amplifier 3 local feedback loops and on the lower part of the left amplifier also 3 local feedback loops. Same for the right amplifier, so totally 12 local feedback loops.
Graham,
“The Ultimate Suplifier” is a balanced input, DC coupled, symmetrical, bridge tied load, class A amplifier with no global feedback using transistors. In a amplifying mode (voltage or current) the transistors are composed in a CFP configuration. The output impedance of a CFP is low so the damping is high. I do agree that the damping factor is lowered down by the output resistors of 0,22 ohm. But, because of the CFP configuration in class A without global feedback, the amplifier will be insensible to the generated back emfs of the loudspeaker. The loudspeaker itself will swing out, but that is something different.
Upupa Epops,
I have no such experience.
Marc.
http://www.suplifier.nl
Hi Mikeks,
Maybe my output stage is not original although I have never seen the schematics of audiolab's amps or Tag Mclaren Audio amplifiers before. But whose ideas are original these days except maybe Susan Parkers non feedback amplifier?
More important, Mikeks, is your opinion on this output stage and on “The Ultimate Suplifier” in general. May I invite you and all others to react on this.
Marc.
Maybe my output stage is not original although I have never seen the schematics of audiolab's amps or Tag Mclaren Audio amplifiers before. But whose ideas are original these days except maybe Susan Parkers non feedback amplifier?
More important, Mikeks, is your opinion on this output stage and on “The Ultimate Suplifier” in general. May I invite you and all others to react on this.
Marc.
Hi Kanwar,
The term "Suplifier" is a contraction of super and amplifier. I used this term for the first time as the title of an article which I wrote in 1989 in a Dutch electronic magazine.
What is your opinion on the circuit of The Ultimate Suplifier?
See http://www.suplifier.nl
Marc.
The term "Suplifier" is a contraction of super and amplifier. I used this term for the first time as the title of an article which I wrote in 1989 in a Dutch electronic magazine.
What is your opinion on the circuit of The Ultimate Suplifier?
See http://www.suplifier.nl
Marc.
Hi Marc,
Thank you for your contribution of the Suplifier.
I will confine my comments to the output stage.
#1. The CFP output pair draws output device current direct from the rail; the purpose of the driver device is to precisely direct the collector efforts of the output device. I think of it as a monkey sitting on the elephant, directing the trunk by pulling at the elephant's ears. Thus, at say 6A output device collector current, the base drive might be 6000/60 or 100mA, and if Vbe of the output device at this current is 1.5V, then an additional 15mA will be drawn through the driver collector resistor, giving a total of 115mA. If you choose a driver with a beta of 80, then the driver's maximum base current will still be less than 1.5mA, so, why do you need to drive it with an emitter follower, an additional phase-inducing stage? I believe the high impedance collector of a VAS running at around 8mA would enable the output stage driver to be driven direct without problem, and lower component count. This would reduce phase shift, and permit simpler, less sonically intrusive lag compensation.
#2. Your circuit is fine for Class A. I have experienced short term instability with CFPs in Class AB. This problem occurs particularly as the negative rail side switches off; short term oscillation can be cured by a cap between base and collector of the driver, but this then kills the music.
#3. I have not checked back through the thread, but recall you used a fully complementary input stage, viz two diff pairs. Is this correct? If so, I would suggest that the resulting complementary current drive to the VAS makes it tough to properly control offset, but more importantly, scotches even order distortion, leaving only odd order. This skews the distortion spectrum, and while it looks great on paper, I'd suggest the sound is better with a single diff pair.
I hope this is useful,
Cheers,
Hugh
Thank you for your contribution of the Suplifier.
I will confine my comments to the output stage.
#1. The CFP output pair draws output device current direct from the rail; the purpose of the driver device is to precisely direct the collector efforts of the output device. I think of it as a monkey sitting on the elephant, directing the trunk by pulling at the elephant's ears. Thus, at say 6A output device collector current, the base drive might be 6000/60 or 100mA, and if Vbe of the output device at this current is 1.5V, then an additional 15mA will be drawn through the driver collector resistor, giving a total of 115mA. If you choose a driver with a beta of 80, then the driver's maximum base current will still be less than 1.5mA, so, why do you need to drive it with an emitter follower, an additional phase-inducing stage? I believe the high impedance collector of a VAS running at around 8mA would enable the output stage driver to be driven direct without problem, and lower component count. This would reduce phase shift, and permit simpler, less sonically intrusive lag compensation.
#2. Your circuit is fine for Class A. I have experienced short term instability with CFPs in Class AB. This problem occurs particularly as the negative rail side switches off; short term oscillation can be cured by a cap between base and collector of the driver, but this then kills the music.
#3. I have not checked back through the thread, but recall you used a fully complementary input stage, viz two diff pairs. Is this correct? If so, I would suggest that the resulting complementary current drive to the VAS makes it tough to properly control offset, but more importantly, scotches even order distortion, leaving only odd order. This skews the distortion spectrum, and while it looks great on paper, I'd suggest the sound is better with a single diff pair.
I hope this is useful,
Cheers,
Hugh
In 1974-8, I designed a range of commercial PA power amplifiers using complementary feedback TRIPLES using the newly released MJ15003/4 - in fact the first use of these in Australia! During the initial design stage, I noticed some small instability which was easily cured with a typ 33ohm resistor in the emitters of the first stage of the triples. Never occurred in production.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.