If there was any amplifier that is ideal and that would really deliver
output = input * gainfactor
then I could fully agree with you.
But there is no such amp around. And one has to keep in mind that even the most objective subject's ears are subjective !
The most clever definition by the objectivist Douglas Self was that of "blameless" as being a piece of equipment whose adverse effects are not zero but below limits of perception.
But I am definitely not sure if he defined the limits of perception correctly ! Chances are great that no one is able to, so who is to judge which amp has lower performance than the other one ?
Keep in mind that auditory masking for instance is a function of level, frequency and time (i.e. multi-dimensional). So there are physiological reasons why SET amps for instance might not just be some kind of nostalgy.
Regards
Charles
Hi Glen,
There are many aspects to consider in designing/making an amplifier.
Going back to post#176 I still do not know which 'compromises' you have in mind and which you would maybe like me to mention out of the blue.
Please realise (as I have written before, though maybe not here) that I rate how a NFB amplifier behaves in relation to loudspeaker generated back-EMF (especially damping phase coherence under the influence of leading current wrt amplified signal output voltage) as being one of the most important capabilities in relation to audio reproduction. Is this the compromise you were thinking of ?
This already covers driving a capacitive loudspeaker with just a few feet of LS cable, though not just sticking a capacitor across the output terminals which is likely to either instantly cause oscillation and heat/destroy the capacitor (actually burnt my fingers once without blowing my amplifier) or blow amplifier fuses. Of course some amplifiers blow their output devices before the fuse fails.
Cheers ........... Graham.
There are many aspects to consider in designing/making an amplifier.
Going back to post#176 I still do not know which 'compromises' you have in mind and which you would maybe like me to mention out of the blue.
Please realise (as I have written before, though maybe not here) that I rate how a NFB amplifier behaves in relation to loudspeaker generated back-EMF (especially damping phase coherence under the influence of leading current wrt amplified signal output voltage) as being one of the most important capabilities in relation to audio reproduction. Is this the compromise you were thinking of ?
This already covers driving a capacitive loudspeaker with just a few feet of LS cable, though not just sticking a capacitor across the output terminals which is likely to either instantly cause oscillation and heat/destroy the capacitor (actually burnt my fingers once without blowing my amplifier) or blow amplifier fuses. Of course some amplifiers blow their output devices before the fuse fails.
Cheers ........... Graham.
Hi Charles,
Yes indeed >> "If there was any amplifier that is ideal and that would really deliver" <> "but there is no such amp around" <<
and thus I see it as so futile that anyone should get hot under the collar and personally derisive about other's opinions and findings.
Cheers .......... Graham.
Yes indeed >> "If there was any amplifier that is ideal and that would really deliver" <> "but there is no such amp around" <<
and thus I see it as so futile that anyone should get hot under the collar and personally derisive about other's opinions and findings.
Cheers .......... Graham.
Re: I have listened that telefunken radio you said.
Carlos,
I am well aware of Graham's physical burden, and frankly admire his drive to work and pursue his ideas in the face of it, when probably many people could have slipped into depression.
Yet, we all must accept this does not give a free pass to assert unwarranted claims.
One thing is the personal appreciation one may have for an individual, and a very different thing is the value one credits to his/her propositions. They belong necessarily to different realms, unfair as some may find it. Nature is neither bad nor good, simply has no mind, simply is.
Graham fails to adhere to the rules as mentioned in my post, which determine how science progress. He does not (for he physically cannot) support his theories with experimental evidence, and he does not work out critique in a convincing way.
Rodolfo
PS. you came short by 21 years ....
destroyer X said:
Carlos,
I am well aware of Graham's physical burden, and frankly admire his drive to work and pursue his ideas in the face of it, when probably many people could have slipped into depression.
Yet, we all must accept this does not give a free pass to assert unwarranted claims.
One thing is the personal appreciation one may have for an individual, and a very different thing is the value one credits to his/her propositions. They belong necessarily to different realms, unfair as some may find it. Nature is neither bad nor good, simply has no mind, simply is.
Graham fails to adhere to the rules as mentioned in my post, which determine how science progress. He does not (for he physically cannot) support his theories with experimental evidence, and he does not work out critique in a convincing way.
Rodolfo
PS. you came short by 21 years ....
That is fairly enough stated Rodolfo, but I have faced challenges from those who have not only failed to investigate in ways I have suggested they will obtain findings, but who appear to refuse to do them because (as a result of established steady-sine-based and time isolated examination theory) they already believe in advance that nothing new will be revealed !!!
And when I simulate output abberations due to stabilisation of a NFB amplifier, as in the NFB thread, no one comments at all.
Not yea - not nay - only 'shoot the messenger' derision !!!
Sad really.
Cheers ........ Graham.
And when I simulate output abberations due to stabilisation of a NFB amplifier, as in the NFB thread, no one comments at all.
Not yea - not nay - only 'shoot the messenger' derision !!!
Sad really.
Cheers ........ Graham.
I'd place the same weight on what an amplifier does when presented with a fast rise/fall square wave, as I would to its response when fed from an RF generator. Last time I checked, time domain and frequency domain were equivalent. The information is interesting, and I don't think people pay enough attention to RF problems, but it's important to keep things in perspective to an audio signal. I'm not as eloquent as Ingrast, but the only logical way to get where we want to go is a combination of listening and testing. It's too easy to get lost/confused/deceived with listening tests, and too easy to build things with sonic flaws if you rely entirely on the test bench. I can find any sonic difference, if one exists, between two pieces of equipment, but I simply don't have a comprehensive enough list of tests to say that, in a vacuum, no difference exists, without backing it up with listening test. That has to include tests by others, since sadly, by the time you have enough experience to really grasp the technical end (if we ever do), your hearing won't be what it was when you were a teenager in the days before ipods.
Conrad Hoffman said:
This is the point, only I would make the distinction that listening is one more way to catch candidate leads for improvement, but measurement is the sanity check.
Again, without a more dissected way of specifying what means "listening" we are not going to make progress.
Rodolfo
Hi Conrad,
You wrote >>Last time I checked, time domain and frequency domain were equivalent. <<
YES - BUT - when current flow is phase shifted wrt source voltage, then voltage propagation through a reactive circuit cannot remain coherent in time.
The straight line response below is the voltage group delay at R due to a series output choke and steady sine excitation, which is of course what everyone measures and bases their understanding upon !!!
The widely variable voltage group delay is that at the (typical)loudspeaker terminal due to another series output choke (both driven by a perfect source), yet this includes a negative delay component caused by leading back-EMF generated by the loudspeaker.
This may be viewed as an expectable reactive response - BUT -
The first cycle of an amplifier's output starts out driving a loudspeaker system that has yet to deveop its back-EMF. Where anyone says it is unacceptable to have a suddenly starting sinewave then they simultaneously deny that an output choke can introduce loudspeaker voltage variation with frequency, and in music performance time, during the first 2 or 3 cycles of any hf waveform change such as transients, high frequencies, harmonic relationships !!!
This is significant because the varying voltage error momentarily is superimposed upon the entire waveform, and not just the component which generates it.
No series output choke means no frequency selective propagation delay variation in time between changing high frequencies, or energetic mids.
Why does everyone get at me, rather than respond about this affectation ?
Once this is understood it is but a small step to see how this distortion of the relationship of voltage in time applies equally to NFB amplifiers which, due to internal stabilisation networks within a global loop, render an entire amplifier inductive with inconstant response to loudspeaker generated back-EMF !!!
Yes the amplifier inductance value is small, but stage current inadequacies and class-AB crossovers exacerbate any effect.
Hi Rodolfo,
Is this computer generated (irrefutable) measurement not the sort of thing you seek in relation to output choke usage ?
Cheers ............ Graham.
You wrote >>Last time I checked, time domain and frequency domain were equivalent. <<
YES - BUT - when current flow is phase shifted wrt source voltage, then voltage propagation through a reactive circuit cannot remain coherent in time.
The straight line response below is the voltage group delay at R due to a series output choke and steady sine excitation, which is of course what everyone measures and bases their understanding upon !!!
The widely variable voltage group delay is that at the (typical)loudspeaker terminal due to another series output choke (both driven by a perfect source), yet this includes a negative delay component caused by leading back-EMF generated by the loudspeaker.
This may be viewed as an expectable reactive response - BUT -
The first cycle of an amplifier's output starts out driving a loudspeaker system that has yet to deveop its back-EMF. Where anyone says it is unacceptable to have a suddenly starting sinewave then they simultaneously deny that an output choke can introduce loudspeaker voltage variation with frequency, and in music performance time, during the first 2 or 3 cycles of any hf waveform change such as transients, high frequencies, harmonic relationships !!!
This is significant because the varying voltage error momentarily is superimposed upon the entire waveform, and not just the component which generates it.
No series output choke means no frequency selective propagation delay variation in time between changing high frequencies, or energetic mids.
Why does everyone get at me, rather than respond about this affectation ?
Once this is understood it is but a small step to see how this distortion of the relationship of voltage in time applies equally to NFB amplifiers which, due to internal stabilisation networks within a global loop, render an entire amplifier inductive with inconstant response to loudspeaker generated back-EMF !!!
Yes the amplifier inductance value is small, but stage current inadequacies and class-AB crossovers exacerbate any effect.
Hi Rodolfo,
Is this computer generated (irrefutable) measurement not the sort of thing you seek in relation to output choke usage ?
Cheers ............ Graham.
Attachments
Where I write above - once this is understood,
when you model for example a straightforward 'Blameless' type circuit with fixed global feedback, but alter the internal stabilisation/inner loop arrangements, not only does this shift the zero voltage crossover (propagation delay) in time as the amplifier copes with leading current loudspeaker load, but the zero voltage crossover point on the waveform is also shifted.
Same input, same global loop, same load, same output, but different internal open loop characteristics, causes different zero volt crossover shifts in time at different loudspeaker dependent frequencies !!!
To image this you have to run several circuits simultaneously, because the time isolated viewpoint when running one alone shows nothing !!!
A shift in propagation delay due to loudspeaker loading equates to a voltage (coherence) error at a specifically chosen time = amplitude distortion.
Ultra low THD is meaningless here.
Thus increasing global NFB does not (CANNOT) make all amplifier outputs equally good.
Also, increasing the number of gain stages to increase global NFB could well make the propagation delay variation worse.
Cheers ........ Graham.
when you model for example a straightforward 'Blameless' type circuit with fixed global feedback, but alter the internal stabilisation/inner loop arrangements, not only does this shift the zero voltage crossover (propagation delay) in time as the amplifier copes with leading current loudspeaker load, but the zero voltage crossover point on the waveform is also shifted.
Same input, same global loop, same load, same output, but different internal open loop characteristics, causes different zero volt crossover shifts in time at different loudspeaker dependent frequencies !!!
To image this you have to run several circuits simultaneously, because the time isolated viewpoint when running one alone shows nothing !!!
A shift in propagation delay due to loudspeaker loading equates to a voltage (coherence) error at a specifically chosen time = amplitude distortion.
Ultra low THD is meaningless here.
Thus increasing global NFB does not (CANNOT) make all amplifier outputs equally good.
Also, increasing the number of gain stages to increase global NFB could well make the propagation delay variation worse.
Cheers ........ Graham.
Graham- as I posted back a ways in this thread, I can definitely see (measure) the effects of the output inductor within the audio band. This was with a resistive load, which IMO is worthless for getting to the bottom of why circuits may sound different, but quite useful for giving customers numbers. No doubt, I'd see more variation with a real (reactive) load, and your explanation makes sense. You're taking a system level approach, which IMO is often overlooked, though I can also see a difference in the style of description that likely conflicts with the way most engineers are taught to view things. IMO, one of the greatest values of these forums is trying gather new ways of looking at things, and maybe learning something 
My question is, as always, how big is the effect, and is it one I (or somebody with decent hearing) can hear? I have no clue at the moment. I do think it will show up in conventional measurements, though being an energy storage phenomena, it may be so subtle nobody recognizes it for what it really is, nor its importance. I haven't spent much time driving amplifier outputs as has been suggested here, but it seems like an easy way to see what back emf is doing.
Though it's nice to be able to put a nice label on things, like THD, my gold standard is differential measurements at very high gains, and with a music signal. Either the signal matches the input, or it doesn't.
To date, I've usually seen mismatches across the audio band that I've attributed to simple phase shift from roll-off components. The exception is some hash that seems directly connected to electrolytic capacitors. What bothered me was the amplitudes seemed larger than I'd expect for signals limited to the audio band. Now you've got me thinking that some of the error I see is not purely phase shift, but may result from the process you've described. A new avenue to pursue here.
Always believing in a multifaceted approach to problem solving, I'm also working with our government representatives here to outlaw transient signals in music under penalty of law- all music will have to be single frequency steady state tones, solving these darned amplifier problems once and for all
My question is, as always, how big is the effect, and is it one I (or somebody with decent hearing) can hear? I have no clue at the moment. I do think it will show up in conventional measurements, though being an energy storage phenomena, it may be so subtle nobody recognizes it for what it really is, nor its importance. I haven't spent much time driving amplifier outputs as has been suggested here, but it seems like an easy way to see what back emf is doing.
Though it's nice to be able to put a nice label on things, like THD, my gold standard is differential measurements at very high gains, and with a music signal. Either the signal matches the input, or it doesn't.
To date, I've usually seen mismatches across the audio band that I've attributed to simple phase shift from roll-off components. The exception is some hash that seems directly connected to electrolytic capacitors. What bothered me was the amplitudes seemed larger than I'd expect for signals limited to the audio band. Now you've got me thinking that some of the error I see is not purely phase shift, but may result from the process you've described. A new avenue to pursue here.
Always believing in a multifaceted approach to problem solving, I'm also working with our government representatives here to outlaw transient signals in music under penalty of law- all music will have to be single frequency steady state tones, solving these darned amplifier problems once and for all
This was my post in the NFB thread.
http://www.diyaudio.com/forums/showthread.php?postid=1157101#post1157101
Here the effect of the series output coil on a suddenly starting (transient/cymbal/triangle harmonic) wave is clearly demonstrated.
Cheers ......... Graham.
http://www.diyaudio.com/forums/showthread.php?postid=1157101#post1157101
Here the effect of the series output coil on a suddenly starting (transient/cymbal/triangle harmonic) wave is clearly demonstrated.
Cheers ......... Graham.
Graham Maynard said:
Graham,
Please note I am not taking position about the output choke, but my hunch is the low values mentioned make it irrelevant at audio frequencies.
That said, I promise to have a look at your simulation and result. Are we talking of a group delay ripple ranging from -1 to +4 uS?
Rodolfo
PS it should help if you clarify what signal are you plotting and with which reference.
Hi Graham
Varying the internal capacitance of a "Blameless" shifts the unity gain frequency up or down. But clearly a smaller Cdom would give a higher frequency response, and I agree with checking the performance of an amplifier to back emf's etc.
In fact, "blameless" circuits I have built have showed very small oscillations (~few mV) when the ft is set to ~10 MHz which gives a 300 kHz bandwidth for closed loop gains ~30, which has not helped, since adding more C elsewhere to stop this just makes things worse. I'm not saying this is a problem of the design, only my observations of the units I have built. Perhaps others have not seen this.
The point is that I agree with your observations that you could change the frequency response of a blameless quite considerably without changing the "audio" response of 20 Hz- 20 kHz, which Self did not mention as a consequence of changing the Cdom and/or Re on the input stage pair. My initial "error" in objecting to the use of Miller caps was to assume that slow output transistors needed the time constant/frequency response to be set rather lower than 10 MHz. - say, 2 or 3 MHz corresponding to the fT's of the output transistors, assuming that these dominate. In fact, if you do lower the response, you can eliminate the niggling oscillation in the units I built, which I mentioned, but this seriously causes possible input stage overloading with fast signals.
As we all recognise in this forum music is not sinewaves and I agree that this could well change the audible performance of an amp. Music is virtually a sequence of "first cycles", and people have pointed out that particular instruments/voices have asymmetrical characteristics. This suggests to me that any ac coupling will cause a low frequency offset every time... so we need to minimise LF ac components too. I prefer to use fully dc coupling (no decoupling capacitor in the feedback arm for example - this also eliminates capacitor noise/distortion), and in my current amp I have traded a little dc drift for dc coupling.
So the point is this ... the performance of an amplifier is likely to depend on the ability to respond to back emf's as you have said, and it is clear that the faster the amp., the more likely it is to deal with whatever transients are input whether from the speaker or input stage. Provided nothing untoward like cut-off affects it. And, in reply to those who state "frequency and time domains are equivalent" while this is basically true, it belies the subtle frequency response differences that, say, a Miller compensated amp. and a phase-lead compensated amp. (like Bailey's or Linsley-Hood's) have, and these are not the same, so their performance won't be either. The question ought to be at what point we are not able to hear the differences...
Unwanted emf's can be generated from the speaker - left hand channel picking up right hand output for example, as well as inductor/capacitor from the crossover components causing loading effects. It is right to continue checking the performance of amplifiers under different conditions - but I'm not sure how you accurately describe back emf's without a full electro-magnetic-acoustic simulation of the speaker. Most simulators don't do this yet, but perhaps speaker manufacturers have (real) acoustic load models...but there is plenty of scope for this to generate effects not simulated by most so far, I suspect.
otherwise, back emf's from the crossover should be pretty predictable and can be described in simulators relatively easily.
keep up the investigations!
cheers
John
Varying the internal capacitance of a "Blameless" shifts the unity gain frequency up or down. But clearly a smaller Cdom would give a higher frequency response, and I agree with checking the performance of an amplifier to back emf's etc.
In fact, "blameless" circuits I have built have showed very small oscillations (~few mV) when the ft is set to ~10 MHz which gives a 300 kHz bandwidth for closed loop gains ~30, which has not helped, since adding more C elsewhere to stop this just makes things worse. I'm not saying this is a problem of the design, only my observations of the units I have built. Perhaps others have not seen this.
The point is that I agree with your observations that you could change the frequency response of a blameless quite considerably without changing the "audio" response of 20 Hz- 20 kHz, which Self did not mention as a consequence of changing the Cdom and/or Re on the input stage pair. My initial "error" in objecting to the use of Miller caps was to assume that slow output transistors needed the time constant/frequency response to be set rather lower than 10 MHz. - say, 2 or 3 MHz corresponding to the fT's of the output transistors, assuming that these dominate. In fact, if you do lower the response, you can eliminate the niggling oscillation in the units I built, which I mentioned, but this seriously causes possible input stage overloading with fast signals.
As we all recognise in this forum music is not sinewaves and I agree that this could well change the audible performance of an amp. Music is virtually a sequence of "first cycles", and people have pointed out that particular instruments/voices have asymmetrical characteristics. This suggests to me that any ac coupling will cause a low frequency offset every time... so we need to minimise LF ac components too. I prefer to use fully dc coupling (no decoupling capacitor in the feedback arm for example - this also eliminates capacitor noise/distortion), and in my current amp I have traded a little dc drift for dc coupling.
So the point is this ... the performance of an amplifier is likely to depend on the ability to respond to back emf's as you have said, and it is clear that the faster the amp., the more likely it is to deal with whatever transients are input whether from the speaker or input stage. Provided nothing untoward like cut-off affects it. And, in reply to those who state "frequency and time domains are equivalent" while this is basically true, it belies the subtle frequency response differences that, say, a Miller compensated amp. and a phase-lead compensated amp. (like Bailey's or Linsley-Hood's) have, and these are not the same, so their performance won't be either. The question ought to be at what point we are not able to hear the differences...
Unwanted emf's can be generated from the speaker - left hand channel picking up right hand output for example, as well as inductor/capacitor from the crossover components causing loading effects. It is right to continue checking the performance of amplifiers under different conditions - but I'm not sure how you accurately describe back emf's without a full electro-magnetic-acoustic simulation of the speaker. Most simulators don't do this yet, but perhaps speaker manufacturers have (real) acoustic load models...but there is plenty of scope for this to generate effects not simulated by most so far, I suspect.
otherwise, back emf's from the crossover should be pretty predictable and can be described in simulators relatively easily.
keep up the investigations!
cheers
John
john curl said:
My current prototype amplifier (for personal HiFi use) under test doesn't seem to need an inductor at all, allthough it's slightly overcompensated into a purely resistive load. I could lighten the compesation for less HF THD, but since the THD-20 is below the 0.01% threshold of my crappy analyser, I probably wont bother.
I have used larger ouput inductors in other designs, incorporated into a LPF Zobel derivation for RFI supression, because these amps were used in areas of high EMI.
Does that make you happy?
- Status
- Not open for further replies.