amplifierguru said:
I have examined your 'scenario', and alas, in the immortal words of Baxandal (RIP), found it to be sub-optimal in all respects..
H Andy C,
To answer your question succinctly - The unity gain frequency of the integrator is dependant on the size of the amplifier and the defined gain based on the requirement of full swing from max input signal, at 100KHz. practical circuit considerations. If it had 60V output swing and was driven from a chip amp with max output of 15V then gain would be 4 at 100KHz and unity gain for the integrator would be 400KHz.
Within the nested loop this is internally stabilised, as being seperately discussed with Mikeks, to give a UGS stable loop with minimal excess phase shift over the 45deg at closure.
Mikeks - you're telling me that applying Ccomp to ground from a Vas OR emitter degen is ALL BAD - how so?
To answer your question succinctly - The unity gain frequency of the integrator is dependant on the size of the amplifier and the defined gain based on the requirement of full swing from max input signal, at 100KHz. practical circuit considerations. If it had 60V output swing and was driven from a chip amp with max output of 15V then gain would be 4 at 100KHz and unity gain for the integrator would be 400KHz.
Within the nested loop this is internally stabilised, as being seperately discussed with Mikeks, to give a UGS stable loop with minimal excess phase shift over the 45deg at closure.
Mikeks - you're telling me that applying Ccomp to ground from a Vas OR emitter degen is ALL BAD - how so?
amplifierguru said:
Listen to the sample files; bow head:
http://www.gedlee.com/distortion_perception.htm
amplifierguru said:
Let's try this again 🙂. I was not asking what the unity gain frequency of the integrator was. That's trivially computed from its time constant, which you already (indirectly) stated based on its gain at 100 kHz. Rather, I was asking what the unity loop gain frequency was. That is, it's a feedback amplifier. What's the frequency for which the magnitude of its loop gain (or return ratio, whatever you choose to call it) is one?
Example: Make an integrator out of an op-amp whose gain-bandwidth product is 10 MHz. For an integrator, the feedback factor is 1 at high frequencies, so the unity loop gain frequency will be approximately the gain-bandwidth product, or 10 MHz.
Good one Jeff Mai, and you believe this?
I'm afraid I'm in no position to evaluate GedLee's wave files except through my computer's little speakers but, if that's all I need - I'll replace my stereo with a portable and wind it up for 110dB as see how palatable for how long that is.
And when Metallica goes on tour, think of the money they'll save on roadies and equipment luugging when no ampont of distortion is detrimental - they can carry a portable each.
Thanks for opening my eyes.
I'm afraid I'm in no position to evaluate GedLee's wave files except through my computer's little speakers but, if that's all I need - I'll replace my stereo with a portable and wind it up for 110dB as see how palatable for how long that is.
And when Metallica goes on tour, think of the money they'll save on roadies and equipment luugging when no ampont of distortion is detrimental - they can carry a portable each.
Thanks for opening my eyes.
amplifierguru said:
No need for any of that. The samples are so obvious that even small computer speakers will show the difference.
Thought experiment: how much distortion (in overall percentage terms) does an MP3 have compared to it's uncompressed original? Because more than 75% of the original data is thrown away, it must be a huge figure (well into double digits) even on the highest bit-rate MP3s. Yet picking the difference between it and the original is not an easy thing to do. Given this fact, how can expressing distortion as a single number have any correlation to SQ?
Hi Andy C,
We're referring to the nested integrator, yes? The block diagram in my post 60 was conceptual. Depending on the practical implementation the gain bandwidth (GBW) within the A2 and output stage integrator loop depends on whether MOSFETs or BJTs are used, in combination with the Vas, Ccomp,and input stage and it's degen. It just needs to be UGS stable with minimum excess phase shift . Typically it could be 2-10 MHz.
We're referring to the nested integrator, yes? The block diagram in my post 60 was conceptual. Depending on the practical implementation the gain bandwidth (GBW) within the A2 and output stage integrator loop depends on whether MOSFETs or BJTs are used, in combination with the Vas, Ccomp,and input stage and it's degen. It just needs to be UGS stable with minimum excess phase shift . Typically it could be 2-10 MHz.
amplifierguru said:
Oh, and nobody said "no amount of distortion is detrimental." What was said is that THD does not correlate to SQ. There is a huge difference in those two statements.
The point is that some harmonic distortion products are vastly more important in determining SQ than others, making THD a meaningless number.
I'm with Jeff on this... i have held for decades that THD has little correlation to sounf quality... Earl Geddes has been scientifically proving that this is the case.
dave
dave
Jeff Mai.
And those warm sounding valve euphonics don't exist either?
Best to render THD to ZERO then while we ascertain just what is 'insignificant'.
There that wasn't too painful and we got there in only 2 posts AND I didn't have to listen to any wave files on my tinny computer speakers.
And those warm sounding valve euphonics don't exist either?
Best to render THD to ZERO then while we ascertain just what is 'insignificant'.
There that wasn't too painful and we got there in only 2 posts AND I didn't have to listen to any wave files on my tinny computer speakers.
amplifierguru said:
??? Huh?
amplifierguru said:
The world is waiting. Let us see this ZERO THD amplifier.
jeff mai said:
Well, that was a fun exercise! Just got done listening to them. The sound card of my computer is an E-Mu 1212m (very low distortion). I plugged in my Sennheiser 580 headphones directly to the analog output of the sound card. It was the first time I tried this. The card drove them nicely.
What surprised me was how bad the one with 0.1 percent THD sounded - completely unlistenable. However, they don't fully disclose the nature of the distortion. Clearly, the distortion is a function of level. So where is the plot of THD vs. level for the nonlinearity? I suspect it's a very abrupt, nasty nonlinearity near zero crossing such that the distortion could be much greater than 0.1 percent at very low signal levels.
I doubt that even the most hardcore objectivist would claim that one THD figure at one frequency and one signal level is all that's needed to determine subjective sound quality. So in a sense the whole thing is somewhat of a strawman. Though I would definitely change my tune if they showed a plot of distortion vs. input level showing a monotonic reduction of distortion as the signal level decreases.
Jeff Mai,
Sought after valve euphonics are simply selected low order harmonics - indicating the very audibility of moderate distortion and it's use to 'improve' sound quality!
ZERO THD is simply THD below the measuring ability of normal test equipment. At least one 'over the top' amplifier has already claimed this. It's simple with nested FB.
Hi Andy C,
Thanks for the feedback. You're right of course - 0.1% of zero crossing notch distortion ( reminiscant of early PA amps) sounds partic BAAAD! So does the half wave spray of harmonics from poor PSRR/supply impedance (see early FFT this thread) Whereas nice euphonic 2nd OR 3rd is a different earful completely even at 5%.
Sought after valve euphonics are simply selected low order harmonics - indicating the very audibility of moderate distortion and it's use to 'improve' sound quality!
ZERO THD is simply THD below the measuring ability of normal test equipment. At least one 'over the top' amplifier has already claimed this. It's simple with nested FB.
Hi Andy C,
Thanks for the feedback. You're right of course - 0.1% of zero crossing notch distortion ( reminiscant of early PA amps) sounds partic BAAAD! So does the half wave spray of harmonics from poor PSRR/supply impedance (see early FFT this thread) Whereas nice euphonic 2nd OR 3rd is a different earful completely even at 5%.
amplifierguru said:
Alternate explanation: the selected low order harmonics *mask* more noticeable high order harmonics while having little negative impact on the SQ themsleves. The Gedlee study mentions that H2 and H3 up to several percent was not detected by test subjects on music signals.
Ah masking! Sounds like a cover up to me.
The 2nd harmonic in the half-wave commutation currents and subsequent associated voltages superimposed on the supply of Class AB amps and injected into the signal path through PSR weakness don't seem to do well at masking the spray of higher harmonics.
And, from experience, doesn't do much for letting through 'detail' and 'air' and 'space' and 'depth'.
The 2nd harmonic in the half-wave commutation currents and subsequent associated voltages superimposed on the supply of Class AB amps and injected into the signal path through PSR weakness don't seem to do well at masking the spray of higher harmonics.
And, from experience, doesn't do much for letting through 'detail' and 'air' and 'space' and 'depth'.
Attachments
Guru: 😀 Very amusing reply. (I do construct both Class D, Class A and Class A/B amplifiers though. The Class A/B ones without any feedback loops at all, not even in the output stage have a THD of 0.07% Class A's slightly lower but no significant (or imaginary) 😀 difference).
However you didn't answer my question about the feedback compensated unliear stage. Well i will give you a hint then. The unlinear stage is actually a modulator or also known in RF tech as a 'mixer'. The unlinear distortion from the mixer is compensated by the feedback, so you will not see the THD on the output. But if more than one frequency (or another waveform than a single sine wave) are sent through mirror frequencies will arise and clutter the sound performance.
Allow me to also joke on your 'music is really just a bunch of sinewaves' claim: Those guys who make MP3 encoders would realy applause you. 😀
THD vs. soundquality non existant correllation: I am aware this is a highly controversial claim on my behalf, because you theoretically founded guys don't like it if you can't measure it. BUT never the less, round up all the tests of amplifiers you can find in any hifi magazine in the world. Read the ratings of the sound quality, then compare with THD measurements. Let me know if you find any correllation at all. 😉 You won't!
As a further support for my claim, i can list a whole bunch of stuff in hifi that you can hear but not measure But then most readers of this forum, i guess, can do the same. Imaginary or not. 😀
Lastly i did not try to attack your pride in any way, merely tried to point out there might be other solutions than the 'usual' ones.
🙂
However you didn't answer my question about the feedback compensated unliear stage. Well i will give you a hint then. The unlinear stage is actually a modulator or also known in RF tech as a 'mixer'. The unlinear distortion from the mixer is compensated by the feedback, so you will not see the THD on the output. But if more than one frequency (or another waveform than a single sine wave) are sent through mirror frequencies will arise and clutter the sound performance.
Allow me to also joke on your 'music is really just a bunch of sinewaves' claim: Those guys who make MP3 encoders would realy applause you. 😀
THD vs. soundquality non existant correllation: I am aware this is a highly controversial claim on my behalf, because you theoretically founded guys don't like it if you can't measure it. BUT never the less, round up all the tests of amplifiers you can find in any hifi magazine in the world. Read the ratings of the sound quality, then compare with THD measurements. Let me know if you find any correllation at all. 😉 You won't!
As a further support for my claim, i can list a whole bunch of stuff in hifi that you can hear but not measure But then most readers of this forum, i guess, can do the same. Imaginary or not. 😀
Lastly i did not try to attack your pride in any way, merely tried to point out there might be other solutions than the 'usual' ones.
🙂
Hi Lars,
What is it with you Scandinavians and feedback loop phobia? Feedback is your friend - particularly where output stages and reactive loads are concerned. More is better.
And music REALLY is just a bunch of sinewaves!
As for " You theoretically founded guys don't like it if you can't MEASURE it" - what is that statement but a contradiction. But, hey, it's called science. Who was it that said something along the lines of "If you can't measure it, you know nothing about it" But if you really mean that I'm out of touch with reality (which has been known to happen particularly after a few charts of home brew), here's an excerpt from my 90's Australian Hi Fi Eidetic pre/poweramp review -
" The instantaneous power available was quite awesome -regardless of the impedance or efficiency of the speakers connected to it. The GB2b seemed to drive anything I connected to it with complete aplomb. Certainly, I cannot really imagine that any normal domestic user will require more power than the GB2b is able to provide in it's 4-8 ohm configuration.
Despite it's ability to deliver power , the pre/power combination proved equally capable, on complex passages, of revealing the subtleties behind full orchestrations. A reduction in the noise floor was evident (over it's predecessor) at least in rooms with sufficiently low background noise to allow one to appreciate it. For example, down near the threshold of audibility, notes seem to crystallise out of silence, stay as long as necessary, then disappear back into a perfect grainless silence.
The stereo image was magnificant. Not only were instruments precisely and correctly positioned in terms of width, it was also possibleto precisely locate their position in three dimensional space, thus adding image height and stage depth to the equation.
Bass performance was tight without being hard and the midrange possessed an uncanny naturalness that seemed to open up the midrange. the high frequencies were clean and clear, irrespective of the volume level, and....
The phono stage was completely transparent and I could not overload it, no matter how hard I tried.
Since Eidetic first became available it has earned a reputation for giant-killer performance and reliability. And, what is more, it has achieved this reputation at prices which put most imported amplifiers to shame."
THD<0.005% (20-20K), 90dB crosstalk (20K) 150W/ch with 2 dB dynamic headroom. from one 300VA toroid.
Science mate, it works!
What is it with you Scandinavians and feedback loop phobia? Feedback is your friend - particularly where output stages and reactive loads are concerned. More is better.
And music REALLY is just a bunch of sinewaves!
As for " You theoretically founded guys don't like it if you can't MEASURE it" - what is that statement but a contradiction. But, hey, it's called science. Who was it that said something along the lines of "If you can't measure it, you know nothing about it" But if you really mean that I'm out of touch with reality (which has been known to happen particularly after a few charts of home brew), here's an excerpt from my 90's Australian Hi Fi Eidetic pre/poweramp review -
" The instantaneous power available was quite awesome -regardless of the impedance or efficiency of the speakers connected to it. The GB2b seemed to drive anything I connected to it with complete aplomb. Certainly, I cannot really imagine that any normal domestic user will require more power than the GB2b is able to provide in it's 4-8 ohm configuration.
Despite it's ability to deliver power , the pre/power combination proved equally capable, on complex passages, of revealing the subtleties behind full orchestrations. A reduction in the noise floor was evident (over it's predecessor) at least in rooms with sufficiently low background noise to allow one to appreciate it. For example, down near the threshold of audibility, notes seem to crystallise out of silence, stay as long as necessary, then disappear back into a perfect grainless silence.
The stereo image was magnificant. Not only were instruments precisely and correctly positioned in terms of width, it was also possibleto precisely locate their position in three dimensional space, thus adding image height and stage depth to the equation.
Bass performance was tight without being hard and the midrange possessed an uncanny naturalness that seemed to open up the midrange. the high frequencies were clean and clear, irrespective of the volume level, and....
The phono stage was completely transparent and I could not overload it, no matter how hard I tried.
Since Eidetic first became available it has earned a reputation for giant-killer performance and reliability. And, what is more, it has achieved this reputation at prices which put most imported amplifiers to shame."
THD<0.005% (20-20K), 90dB crosstalk (20K) 150W/ch with 2 dB dynamic headroom. from one 300VA toroid.
Science mate, it works!
Guru: OK If you want to talk science .. 😀
Actually science agrees that you can not measure something without changing it.
It's called the theory of complementarity. 😀
But then let me ask you: Do you claim that if you can hear a difference, then it follows that you can also measure a difference?
And vice versa?
Actually science agrees that you can not measure something without changing it.
It's called the theory of complementarity. 😀
But then let me ask you: Do you claim that if you can hear a difference, then it follows that you can also measure a difference?
And vice versa?
If you can hear a difference, but can't measure one then:
(a) Your ears/brain are deceiving you.
or (b). You are not measuring the right thing.
(a) Your ears/brain are deceiving you.
or (b). You are not measuring the right thing.
Thanks Ouroboros, My feelings exactly.
That's why if I hear a difference I always confirm it with others whose aural acuity I value and, if it's really important, double blind testing.
When the result is, for example, that FET input stages sound better, I'll use FET input stages and mull over what it is I have to measure to put numbers on it.
I won't not use them because I haven't yet quantified why.
That's why if I hear a difference I always confirm it with others whose aural acuity I value and, if it's really important, double blind testing.
When the result is, for example, that FET input stages sound better, I'll use FET input stages and mull over what it is I have to measure to put numbers on it.
I won't not use them because I haven't yet quantified why.
- Status
- Not open for further replies.