so what means blameless?
so what means blameless?
These are my personal beliefs or thesis:
Obvious?
Sure!
Something more?
so what means blameless?
These are my personal beliefs or thesis:
- no overdrives in NFB-circuitry
- rather low gains of each stage
- use class A if possible, this reduces crossover dists and minimalyse "memory"
- use PP circuitry, this reduces both odd and even nonlinearities
- try to not use emmiter or source followers - they not sound so good...
- use low number of amplifying stages
- for most circuits better are local nfbies than global
- poor designed strong nfbies gives glassy sound
- for strong signals use jfets, for weak bjts
Obvious?
Sure!
Something more?
"Blameless" means the designer couldn't figure out why it doesn't sound better. 
"rather low gains of each stage" - why?
"use PP circuitry, this reduces both odd and even nonlinearities" - are you sure?
"try to not use emmiter or source followers - they not sound so good..." - why?
"use low number of amplifying stages" - why?
"poor designed strong nfbies gives glassy sound" - true.
"for strong signals use jfets, for weak bjts" - why?
"rather low gains of each stage" - why?
"use PP circuitry, this reduces both odd and even nonlinearities" - are you sure?
"try to not use emmiter or source followers - they not sound so good..." - why?
"use low number of amplifying stages" - why?
"poor designed strong nfbies gives glassy sound" - true.
"for strong signals use jfets, for weak bjts" - why?
"rather low gains of each stage" - why?
when one uses a few gain stages the strong 2V (I preffer to connect CD via a potentiometer turned close to maximum direct to the amp) signal form an output of cd may be clipped in any further stage.
"use PP circuitry, this reduces both odd and even nonlinearities" - are you sure?
yes both but mainly the second
"try to not use emmiter or source followers - they not sound so good..." - why?
don't know why? EFs sound worse just in comparsion with common source/emmiter circuits
"use low number of amplifying stages" - why?
I think that an ideal is a first stage and then output transistors. The sound is distored only by 2 devices
"for strong signals use jfets, for weak bjts" - why?
because of the characteristic. for the signals from MC/MM cartridge bjts gives high gain and low dist. for the signals from the cd player jfets are more "robust" and their use is simpler.
when one uses a few gain stages the strong 2V (I preffer to connect CD via a potentiometer turned close to maximum direct to the amp) signal form an output of cd may be clipped in any further stage.
"use PP circuitry, this reduces both odd and even nonlinearities" - are you sure?
yes both but mainly the second
"try to not use emmiter or source followers - they not sound so good..." - why?
don't know why? EFs sound worse just in comparsion with common source/emmiter circuits
"use low number of amplifying stages" - why?
I think that an ideal is a first stage and then output transistors. The sound is distored only by 2 devices
"for strong signals use jfets, for weak bjts" - why?
because of the characteristic. for the signals from MC/MM cartridge bjts gives high gain and low dist. for the signals from the cd player jfets are more "robust" and their use is simpler.
Referring back to the memory distortion stuff (http://peufeu.free.fr/audio/memory/): what about using a tube VAS in an otherwise solid state amplifier? Say a triode loaded with one of Gary Pimm's MOSFET CCS? And what about the input stage? Pros/cons?
Prune said:
using a triode you do not have to use VAS. There is only the input stage which is VAS as one.
from a single ecc83 loaded with a resistor you can get 70V~ peak undistored (thd 1%) signal to drive the output current devices.
probably there is nothing better...
but how to couple it direct?
Blameless Class B
I've been reading this thread with some interest. I believe a few comments are called for.
The differential input pair is a voltage to current stage. The VAS is, by virtue of the feedback capacitor, a current to voltage stage with an output impedance that is reduced with increasing frequency. At low frequency the situation is more complicated.
Someone, very early in the thread, hit the nail on the head. Bipolar devices have relatively high output admitance, and this effect can feed back to the VAS at lower frequencies where the output impedance of the VAS is reduced.
Most of Self's testing was done with a resistive load, so that feedback characteristics were well defined. An earlier post correctly identified that more work must be done to understand the effect of output loading. The only work I saw from Self on this subject was doubling the output transistors. While this may solve quantitative issues, it does not appear to address any qualitative issues resulting from the significant output admittance of the circuit.
If one only cares about the sound, then one should use whatever amplifier sounds best. But I am quick to defend those who are willing to take pains to relate sonic qualities to measurable phenomena. Otherwise, we are left with religion, not science. As science, Self has done the kind of science I agree with. Finding a parameter that matters (in this case distortion) and working through a series of experiments to measure and control it.
If it still sounds awful, then we should look for WHAT ELSE is wrong.
With a resistive load the blameless class B does in fact reduce distortion. But more modelling of the output stage with a music source and reactive speaker load obviously must be done. It is possible that MOSFETs provide better sonic results, in spite of higher distortion, owing to lower output admittance. This would vindicate both sides of this unseemly (and unscientific) debate. If square-law devices sound better (there is overwhelming anecdotal evidence to suggest that it does) then we outght to try to understand why.
Instead of arguing between subjectivist "Sounds Bad" and objectivist "Looks Good" we should assume that we are all telling the truth, and use that knowledge to understand what OTHER parmeter we are not measuring, and therefore, have left uncontrolled.
Since everything from room acoustics back to the VAS stage has some causal connection in the 'wrong' direction, my suspicion is that it will be impossible to build a truly 'blameless' amplifier until we have a better understanding of the effects of speaker impedance nonlinearities. Simply put, the speaker impedance *must* be modelled with the amplifier. Using an 8 ohm resistor in place of a wildly reactive (and probably nonlinear) load is wildly optimistic.
Any thoughts?
I've been reading this thread with some interest. I believe a few comments are called for.
The differential input pair is a voltage to current stage. The VAS is, by virtue of the feedback capacitor, a current to voltage stage with an output impedance that is reduced with increasing frequency. At low frequency the situation is more complicated.
Someone, very early in the thread, hit the nail on the head. Bipolar devices have relatively high output admitance, and this effect can feed back to the VAS at lower frequencies where the output impedance of the VAS is reduced.
Most of Self's testing was done with a resistive load, so that feedback characteristics were well defined. An earlier post correctly identified that more work must be done to understand the effect of output loading. The only work I saw from Self on this subject was doubling the output transistors. While this may solve quantitative issues, it does not appear to address any qualitative issues resulting from the significant output admittance of the circuit.
If one only cares about the sound, then one should use whatever amplifier sounds best. But I am quick to defend those who are willing to take pains to relate sonic qualities to measurable phenomena. Otherwise, we are left with religion, not science. As science, Self has done the kind of science I agree with. Finding a parameter that matters (in this case distortion) and working through a series of experiments to measure and control it.
If it still sounds awful, then we should look for WHAT ELSE is wrong.
With a resistive load the blameless class B does in fact reduce distortion. But more modelling of the output stage with a music source and reactive speaker load obviously must be done. It is possible that MOSFETs provide better sonic results, in spite of higher distortion, owing to lower output admittance. This would vindicate both sides of this unseemly (and unscientific) debate. If square-law devices sound better (there is overwhelming anecdotal evidence to suggest that it does) then we outght to try to understand why.
Instead of arguing between subjectivist "Sounds Bad" and objectivist "Looks Good" we should assume that we are all telling the truth, and use that knowledge to understand what OTHER parmeter we are not measuring, and therefore, have left uncontrolled.
Since everything from room acoustics back to the VAS stage has some causal connection in the 'wrong' direction, my suspicion is that it will be impossible to build a truly 'blameless' amplifier until we have a better understanding of the effects of speaker impedance nonlinearities. Simply put, the speaker impedance *must* be modelled with the amplifier. Using an 8 ohm resistor in place of a wildly reactive (and probably nonlinear) load is wildly optimistic.
Any thoughts?
Hi upstart.
Neatly reasoned.
You ask - What other parameter are we not measuring ?
Maybe it is the way we have been measuring that has been wrong.
Does anyone yet routinely measured THD and do a Fourier analysis when a loudspeaker is the load, either on the bench or via a computer simulator ?
Even if we did, would it tell us anything about first cycle amplifier activity in response to a suddenly changing waveform voltage driving an energy storing loudspeaker when some back EMF generating effects arise within microseconds, and others do not settle until after maybe two or three cycles of the enegising waveform have passed.
Steady sinewave THD measurements and Fourier based calculations miss all of these.
What did we do before computers made diyAudio a worldwide reality - we used sinewave nulling equipment to isolate the distortion component that an amplifier output was carrying, and we became experienced at recognising which error traces would sound bad.
It is easy to do this with a computer too, and with the additional use of a virtual loudspeaker load with crossover we can get some worthwhile results.
Say at 10kHz -10V.RMS amplifier output set up a second 'nulling' generator in series with the output measuring voltmeter until an even crossover phase an amplitude nulled residual is obtained. You need to run several traces as the settings are 'fine tuned'.
THEN run from t=0 minus system propagation delay period and study first cycle errors, noting how with some amps they can be appreciable and stabilise after three or four cycles.
___________________________________________________
If you do this on a Blameless type of amplifier the first half cycle error at 15 to 45 uS is poorer than -40dB due to its output choke; depending on bias and devices it stabilises at approx -50dB. Take the choke away and the first cycle error disappears going straight to the -50dB level, which is much worse than when a resistor is the amplifier load.
This -50dB error is due to Miller Effect induced by the VAS connected C.dom, take this away and the distortion drops by another 40dB !!!!
___________________________________________________
Cheers .......... Graham.
Neatly reasoned.
You ask - What other parameter are we not measuring ?
Maybe it is the way we have been measuring that has been wrong.
Does anyone yet routinely measured THD and do a Fourier analysis when a loudspeaker is the load, either on the bench or via a computer simulator ?
Even if we did, would it tell us anything about first cycle amplifier activity in response to a suddenly changing waveform voltage driving an energy storing loudspeaker when some back EMF generating effects arise within microseconds, and others do not settle until after maybe two or three cycles of the enegising waveform have passed.
Steady sinewave THD measurements and Fourier based calculations miss all of these.
What did we do before computers made diyAudio a worldwide reality - we used sinewave nulling equipment to isolate the distortion component that an amplifier output was carrying, and we became experienced at recognising which error traces would sound bad.
It is easy to do this with a computer too, and with the additional use of a virtual loudspeaker load with crossover we can get some worthwhile results.
Say at 10kHz -10V.RMS amplifier output set up a second 'nulling' generator in series with the output measuring voltmeter until an even crossover phase an amplitude nulled residual is obtained. You need to run several traces as the settings are 'fine tuned'.
THEN run from t=0 minus system propagation delay period and study first cycle errors, noting how with some amps they can be appreciable and stabilise after three or four cycles.
___________________________________________________
If you do this on a Blameless type of amplifier the first half cycle error at 15 to 45 uS is poorer than -40dB due to its output choke; depending on bias and devices it stabilises at approx -50dB. Take the choke away and the first cycle error disappears going straight to the -50dB level, which is much worse than when a resistor is the amplifier load.
This -50dB error is due to Miller Effect induced by the VAS connected C.dom, take this away and the distortion drops by another 40dB !!!!
___________________________________________________
Cheers .......... Graham.
Re: Blameless Class B
good post, like the next
the religion and science is starting to equaling!
like the sounding and measuring
Regarding earlier thoughts on this thread: maybe this is the reason that I do not like the global nfb because this moves the speakers "wild voltages and currents" back to the input. this spoils both the sound and the "real speaker" measurments.
the solution might be realy parraleling the output devices and taking the nfb signal from an "ideal" load resistor of the first pair of the devices. the rest devices are steering the speakers = both sound and thd are good (The ALTMANN nnfb amp conception)
upstart said:
good post, like the next
the religion and science is starting to equaling!
like the sounding and measuring
Regarding earlier thoughts on this thread: maybe this is the reason that I do not like the global nfb because this moves the speakers "wild voltages and currents" back to the input. this spoils both the sound and the "real speaker" measurments.
the solution might be realy parraleling the output devices and taking the nfb signal from an "ideal" load resistor of the first pair of the devices. the rest devices are steering the speakers = both sound and thd are good (The ALTMANN nnfb amp conception)
One topology for this may be to have the main voltage-gain and driver stage enclosed in an overall NFB loop, but have the output emitter/source follower stage not being part of the overall NFB loop.
The output stage could instead have a 'Malcolm Hawksford'-like error-correction mechanism around it to give it the best possible linearity.
The output stage could instead have a 'Malcolm Hawksford'-like error-correction mechanism around it to give it the best possible linearity.
Exchanging individual notions of "blameless", is on par with debating "optimal", or "perfect". The original purpose for the concept by Self, was to establish a departure point for a standard Lin-type topology. From that one could alter a single element at a time and measure/observe the result. This is a useful concept. Otherwise you get in to situations where you try to compare alternatives (for example "Current Source A" vs. Current Source B") but get ambiguous results because other topological features are different as well.
I wish Self has named it differently, avoiding any qualitative adjectives so as to avoid misunderstanding. Perhaps "The Self Baseline Amp" would have been better. I've the impression that many interpret the name to mean it was put forth as some kind of optimal or perfect design. That misses the point since Self uses it to examine possible improvements. It seems obvious, to me at least, that something that is put forth as a starting point to seek improvement would not have been considered ideal in any way.
If one disagrees with the choice of baseline, describe your own and conduct your own investigations. I can easily concieve that for some purposes a push-pull Class-A baseline would be mor suitable; in other cases a single ended configuration might serve that purpose.
Personally, I don't see much point in constructing a "blameless" as described by Self, just for listening, but if one is going to investigate alternate circuit configurations, I don't see how you can avoid using a baseline if you hope to reach a verifiable, repeatable conclusion. If you are then going to comunicate those conclusions to others, you will be more successful if you explicitly describe the baseline.
hehe, that's Pavel, a very special man
I can recognize you, if you enter any forum,whatever the name you use, no matter the country flag or the name origin.... the way you call people to change ideas, a very intelligent and educated way to ask a challenge, that will result wonderfull to the "knowledge vacuum cleaners... sucking informs like me"...you have a passionate way to feed the people to put their ideas out.
Of course, 5 or 6, that could not keept their secrets hidden, because you make them appear, will do not like this....but, you will have more 9994 other people clapping when you turn you "question mark panzer 75mm cannon" to us.
Or they oppen their mouth, to tell the hidden secrets...or... they will loose in arguments...hehe..... no way to scape.... if run, the bear catch you....if you stand.... the bear cutt you in half slices...hehe.
It is good to have you here Pave.
You will make those mans put their hidden things out!
that's very good!
regards,
Carlos
I can recognize you, if you enter any forum,whatever the name you use, no matter the country flag or the name origin.... the way you call people to change ideas, a very intelligent and educated way to ask a challenge, that will result wonderfull to the "knowledge vacuum cleaners... sucking informs like me"...you have a passionate way to feed the people to put their ideas out.
Of course, 5 or 6, that could not keept their secrets hidden, because you make them appear, will do not like this....but, you will have more 9994 other people clapping when you turn you "question mark panzer 75mm cannon" to us.
Or they oppen their mouth, to tell the hidden secrets...or... they will loose in arguments...hehe..... no way to scape.... if run, the bear catch you....if you stand.... the bear cutt you in half slices...hehe.
It is good to have you here Pave.
You will make those mans put their hidden things out!
that's very good!
regards,
Carlos
Re: Sorry, maybe not the same Sam... the other one was...
Hehe. You mean Rock Bottom? http://www.felixthecat.com/friends-rock_bottom.htm
Hehe. You mean Rock Bottom? http://www.felixthecat.com/friends-rock_bottom.htm
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.