I think Self's dismissal of full scale 20 kHz long term testing in home music reproduction amps is mostly about economic Zobel component sizing
if the Zobel keeps the amp form oscillation then there's no need for a full amp power bandwidth rated resistor in the Zobel, <5W resistors are a lot cheaper than >50W
just because you could burn up a 5 W rated R in the output Zobel with a test signal doesn't mean its not a acceptable choice for consumer use
if the Zobel keeps the amp form oscillation then there's no need for a full amp power bandwidth rated resistor in the Zobel, <5W resistors are a lot cheaper than >50W
just because you could burn up a 5 W rated R in the output Zobel with a test signal doesn't mean its not a acceptable choice for consumer use
Yes and have found that with Zobels. I do beef it up during testing. A simple calculation.
He is very right to say 5 watt wire-wound and live with it being less than ideal. Those resistors will take far more. Wrap them through two holes in PCB and solder well. If not they drop off the PCB when saving amp and speakers. It is remarkable how well the Zobel works. 10 R 0.1 uF is about right regardless , I use 4R7 0.1uF . TDA 2050 type devices can be 1R 0.1 uF.
I made a op amp style 50 watt amp with built in active linear phase filter as if an op amp . A capacitor fell off and all was very unhappy. The Zobel stopped it getting serious. When corrected it worked very well and no faiiure. It was to drive a motor with a pure sine wave.
He is very right to say 5 watt wire-wound and live with it being less than ideal. Those resistors will take far more. Wrap them through two holes in PCB and solder well. If not they drop off the PCB when saving amp and speakers. It is remarkable how well the Zobel works. 10 R 0.1 uF is about right regardless , I use 4R7 0.1uF . TDA 2050 type devices can be 1R 0.1 uF.
I made a op amp style 50 watt amp with built in active linear phase filter as if an op amp . A capacitor fell off and all was very unhappy. The Zobel stopped it getting serious. When corrected it worked very well and no faiiure. It was to drive a motor with a pure sine wave.
But the first one will have better linearity (closed loop) everywhere below 20kHz, no?
Jan
a quite robust feedback result is that the excess loop gain reduces error by the loop gain at that frequency - I think more gain at the frequencies that are most audible is a fine consequence of high loop gain/low bandwidth if single pole compensation is your choice
the flat loop gain prescription only addresses one source of phase modulation - VAS and driver nonlinear C are others with different profiles, as are various Q hfe, ft roll off or modulation with output V and I
high feedback reduces all errors detected by the input diff stage - including phase errors - whatever their source
I also don't buy the "the distortion shouldn't vary with frequency" argument - if its low enough to be inaudible why should is rising from buried orders of magnitude below the Johnson Noise floor for signals << 1 kHz (musical instrument fundamentals anyone?) to still below audibility at 20 kHz matter?
in fact since some errors are increasing with frequency - to flatten them over audio would require a rising loop gain over audio frequencies - any out there try that?
arguments from music properties lead to more gain being desired at lower audio frequency particularly with a little knowledge of music power vs frequency distributions, then add in knowledge of temporal masking for just how inaudible small distortions around a transient are
the flat loop gain prescription only addresses one source of phase modulation - VAS and driver nonlinear C are others with different profiles, as are various Q hfe, ft roll off or modulation with output V and I
high feedback reduces all errors detected by the input diff stage - including phase errors - whatever their source
I also don't buy the "the distortion shouldn't vary with frequency" argument - if its low enough to be inaudible why should is rising from buried orders of magnitude below the Johnson Noise floor for signals << 1 kHz (musical instrument fundamentals anyone?) to still below audibility at 20 kHz matter?
in fact since some errors are increasing with frequency - to flatten them over audio would require a rising loop gain over audio frequencies - any out there try that?
arguments from music properties lead to more gain being desired at lower audio frequency particularly with a little knowledge of music power vs frequency distributions, then add in knowledge of temporal masking for just how inaudible small distortions around a transient are
Well it is interesting to see people chipping away at previous measurements, and even attacking ideas (like PIM) that are not fully developed even.
There seems to be a tug of war between two groups of people: Some who want to make better audio amps, and those who think that there is nothing wrong with most audio amps. What can I say? I am not completely happy with any amp that I have ever been exposed to, so I strive to learn even more, so that next year's designs will sound better still, just like I would presume some automotive designers might feel as well with their creations.
Others think that even the test 'standards' that we have developed are a waste of time and should be ignored. This is where I come in and try to get them to see that our previous research was not in vain, and still useful, just requiring a little more effort.
Why would anyone want to make less?
There seems to be a tug of war between two groups of people: Some who want to make better audio amps, and those who think that there is nothing wrong with most audio amps. What can I say? I am not completely happy with any amp that I have ever been exposed to, so I strive to learn even more, so that next year's designs will sound better still, just like I would presume some automotive designers might feel as well with their creations.
Others think that even the test 'standards' that we have developed are a waste of time and should be ignored. This is where I come in and try to get them to see that our previous research was not in vain, and still useful, just requiring a little more effort.
Why would anyone want to make less?
what measurements, please?
I'm all for things that can be measured - do you mean Quan's null results when he added merely 20 year old op amps to his 2nd paper?
we can easily repeat most any proposed test today with prosumer soundcards - 192k should cover anything labeled "music" coming out of commercial studios today
Scott, PMA others have done some of the obscure Multione/Noise Fill tests on these cards with results in the noise even with stock op amps, Al electrolytic coupling caps...
if desired there are ADC for medical imaging market delivering 120 dB SFDR, THD at 1 Ms/s
state a testable hypothesis, give required measurement resolution, justify re know recording studio practice, EE Signal Theory, Psychoacoutics...
I'm all for things that can be measured - do you mean Quan's null results when he added merely 20 year old op amps to his 2nd paper?
we can easily repeat most any proposed test today with prosumer soundcards - 192k should cover anything labeled "music" coming out of commercial studios today
Scott, PMA others have done some of the obscure Multione/Noise Fill tests on these cards with results in the noise even with stock op amps, Al electrolytic coupling caps...
if desired there are ADC for medical imaging market delivering 120 dB SFDR, THD at 1 Ms/s
state a testable hypothesis, give required measurement resolution, justify re know recording studio practice, EE Signal Theory, Psychoacoutics...
For example, TIM is an established standard measurement that is even included in modern test equipment like the Stanford Research SR1.
A great deal of time and trouble was put into creating this test. We did not want it to be too 'exotic' and we wanted the test signal to be very close to actual music on some occasions (reasonable worst case).
If you cannot pass TIM-30, then you have a compromised product. Maybe it is good enough for you, but technically it's compromised.
A great deal of time and trouble was put into creating this test. We did not want it to be too 'exotic' and we wanted the test signal to be very close to actual music on some occasions (reasonable worst case).
If you cannot pass TIM-30, then you have a compromised product. Maybe it is good enough for you, but technically it's compromised.
what's a good number? Bob's 1983 MOSFET Amplifier's 0.005%? (the text implies that was his noise floor)
Walt Jung http://waltjung.org/PDFs/SID_TIM_TAA77_P3.pdf also couldn't measure the DIM of a TL084
why do you think that there is anyone here proposing high feedback amps that would fail the test?
Walt Jung http://waltjung.org/PDFs/SID_TIM_TAA77_P3.pdf also couldn't measure the DIM of a TL084
why do you think that there is anyone here proposing high feedback amps that would fail the test?
Last edited:
Its the peak current and its relationship to the voltage at that instant that matters. I see a lot of arm waving and the Otala number of 50A quoted but little actual evidence to back the claim up. Perhaps all the high slew rate high current stuff is to ensure that the real operating area for music that is being listened to by humans is inside a very safe and linear part of the operating envelope. Maybe someone can show the waveforms to counter this?
Part of the dilemma of deciding how much power and speaker sensitivity is needed is because it varies depending upon the 'quality' of the end sound. If the reproduction is clean at near maximum levels for the system then one can easily look for, require more; one's hearing system easily adjusts to the higher SPLs, it still "sounds the same", and one would love for extra decibels to be possible. However, if higher levels just makes subjectively disturbing distortion artifacts more obvious, then a natural limit imposes itself - this loud, and no more, I feel uncomfortable with the higher volume ...
Cordell's MOSFET Amplifier paper gives some refs, claims they include measurements
more recent: Heavy Load: How Loudspeakers Torture Amplifiers | Stereophile.com
actually the "impedance dip" and "equivalent peak dissipation resistance" may be practical but simplistic - you can exceed even those numbers with a specific matched waveform that essentially pumps up each resonance and then reverses the phase of the drive signal to of all of them at the same instant
but this is textbook knowledge - even for "dumb" "conventional engineers" purporting to be audio designers who should at least be held to the knowledge of Cordell's and Self's audio power amplifier books today
both reference the phenomena - conclude that yes you do have to have more output stage current capacity than the nameplate loudspeaker R value would indicate - by factors of up to 6 - although Doug does again go for economy and suggest 2x is adequate if current limiting is clean, with quick recovery - but he isn't designing for the nose bleed, price is the object crowd
more recent: Heavy Load: How Loudspeakers Torture Amplifiers | Stereophile.com
actually the "impedance dip" and "equivalent peak dissipation resistance" may be practical but simplistic - you can exceed even those numbers with a specific matched waveform that essentially pumps up each resonance and then reverses the phase of the drive signal to of all of them at the same instant
but this is textbook knowledge - even for "dumb" "conventional engineers" purporting to be audio designers who should at least be held to the knowledge of Cordell's and Self's audio power amplifier books today
both reference the phenomena - conclude that yes you do have to have more output stage current capacity than the nameplate loudspeaker R value would indicate - by factors of up to 6 - although Doug does again go for economy and suggest 2x is adequate if current limiting is clean, with quick recovery - but he isn't designing for the nose bleed, price is the object crowd
Last edited:
Scott, this is the star moment of diplomatic speech.
I'll be damned if I know which of the two you'd take.
Obviously, I'd take the 20 kHz OLB one.
Scott, this is the star moment of diplomatic speech.
I'll be damned if I know which of the two you'd take.
Obviously, I'd take the 20 kHz OLB one.
I would take the other one because it has the same THD at 20kHz but progressively lower at all lower frequencies. Free lunch etc.
Jan
- Status
- Not open for further replies.