Not quite true, Andrew, and not everyone.......
There is the maxim that designing amps without NFB could need as much as that too, depending on how many thousands of hours the designer has spent into the topology!
I'm not disagreeing with you vehemently, but there are good and bad designers on both camps.
Like so many issues in engineering, I tend to lean to low GNFB amps, the middle ground, somewhere around 25-32dB. Why the figure? Seems like a nice figure to design to, and seems to sound pretty good too. I admit that the best sound I have heard is a grotesquely inefficient SE mosfet, running at 3A with a mosfet CCS and 50V Vcc. Glorious 28Wrms with a 6SL7 front end. But it is not particularly practical with 150W of heat to dispense with in the Melbourne summer. Absolutely no GFB and very little local FB.
BTW, your '.....thousands of hours spent in learning' would be 10,000.
That's what it really takes for a half smart person to fully learn how to design good amps.
Hugh
Last edited:
No. Not in this thread.
If you wish to see my measurements, refer XRK971's figures done on a few of my open source amps - the ALPHA, and the ALPHA Nirvana.
If you wish to see my measurements, refer XRK971's figures done on a few of my open source amps - the ALPHA, and the ALPHA Nirvana.
I have little doubt you are right and this investment probably pays off well in many engineering tasks.
But does it translate into better sound? Practice shows that some of the most "engineered" amps are also among the worst sounding. An unpleasant discrepancy between theory and practice.
Once again: a surprising number of the ultra high end amps currently in manufacture use none, or a very limited amount of nfb. Why? Could it be because adjectives such as cold, clinical, unnatural and tiresome do not exactly attract paying customers in their droves?
A combination of marketing and the fact that people in the market for 100k amplifiers care about looks and bragging rights more than sound. Benchmark can't make AHB2s fast enough. That should say something about discering consumers?
Last edited:
Look at this thread here, the late Mauro Penasa did some quite interesting tests, starting with this post:
https://www.diyaudio.com/forums/solid-state/57995-distortion-5.html#post654100
Then he posted a report (2 parts) about his findings:
https://www.diyaudio.com/forums/solid-state/57995-distortion-10.html#post667616
https://www.diyaudio.com/forums/solid-state/57995-distortion-11.html#post667619
Mauro didn't master the english language too well, so sometimes it might be challenging to fully understand what he wanted to say, but nevertheless, he found some interesting things.
Unfortunately, it got forgotten by many. (What a big loss his passing was, my thoughts go to his family)
Apart that, we see a lot of measurements into resistive loads, and only few into reactive loads simulating a real loudspeaker. Especially multitone tests.
Now we should also reread Nelson's article, maybe...
I can only say that I have a hard time to believe that even good ears can reliably identify deliberately added 2nd or 3rd harmonic to the music, if lower than -60dB (or 0.1%) .
And such a distortion level is happily achievable with amplifiers having no global feedback.
We also may calculate the frequency error of an output stage running open loop into a real, typical loudspeaker. Say 4 BJT output pairs, each biased at 80mA, 0.33 Ohm Emitter resistor, driven from a low AC impedance driver.
I think it will be difficult to discern anything, if not listening to white or pink noise, but to music.
https://www.diyaudio.com/forums/solid-state/57995-distortion-5.html#post654100
Then he posted a report (2 parts) about his findings:
https://www.diyaudio.com/forums/solid-state/57995-distortion-10.html#post667616
https://www.diyaudio.com/forums/solid-state/57995-distortion-11.html#post667619
Mauro didn't master the english language too well, so sometimes it might be challenging to fully understand what he wanted to say, but nevertheless, he found some interesting things.
Unfortunately, it got forgotten by many. (What a big loss his passing was, my thoughts go to his family)
Apart that, we see a lot of measurements into resistive loads, and only few into reactive loads simulating a real loudspeaker. Especially multitone tests.
Now we should also reread Nelson's article, maybe...
I can only say that I have a hard time to believe that even good ears can reliably identify deliberately added 2nd or 3rd harmonic to the music, if lower than -60dB (or 0.1%) .
And such a distortion level is happily achievable with amplifiers having no global feedback.
We also may calculate the frequency error of an output stage running open loop into a real, typical loudspeaker. Say 4 BJT output pairs, each biased at 80mA, 0.33 Ohm Emitter resistor, driven from a low AC impedance driver.
I think it will be difficult to discern anything, if not listening to white or pink noise, but to music.
If you were referring to current PFB that has gained short-lived fame in some tube amplifiers in the 1950ies, surely not. It has been applied in low doses to decrease the amp's output impedance to almost zero (even negative impedance behaviours were possible, but not practical, of course) in order to meliorate control over the speaker, hence decreasing the speaker's distortions.
Best regards!
I believe this is central to this argument. Especially if the distortion does not rise with frequency.
And a quote from the late, great Charles Hansen:
"Here is the best proof that negative feedback does not work:
IF NEGATIVE FEEDBACK ACTUALLY WORKED, THEN ALL AMPLIFIERS WOULD BE PERFECT, AND ALL AMPLIFIERS WOULD SOUND EXACTLY THE SAME."
😎
You are clearly ignorant of the First Watt F7 which carries on the fame of current PFB 😀
You need more biasing and lower Re for this to work well. NAD S300 is a good example how it can be done.
Oh boy! Obviously he is and you are ingorant of the fact that (voltage) GNFB won't decrease THD and output impedance to zero and increase bandwith to infinity. Instead it decreases both the first parameters by the GNFB factor and increase bandwith to some amount, depending on the design. It has numerously been shown within this thread that an open loop design that is as linear and as low THD as possible is mandatory.
Maybe I am. The result most probably is shown in #1743, first diagram.
Best regards!
Last edited:
Here are some measurements of an amp running open loop, from the late Charles Hansen:
Ayre Acoustics MX-R Twenty monoblock power amplifier Measurements | Stereophile.com
Output Z 0.26 - 0.28 Ohms, results in a frequency error of +/- 0.25 dB with their real LS load.
Distortion not low from an academic point of view, but I'd say low enough not to be detected in a controlled listening test...
Ayre Acoustics MX-R Twenty monoblock power amplifier Measurements | Stereophile.com
Output Z 0.26 - 0.28 Ohms, results in a frequency error of +/- 0.25 dB with their real LS load.
Distortion not low from an academic point of view, but I'd say low enough not to be detected in a controlled listening test...
I think you just destroyed any remaining believe-ability you still had ...
Anyone who still thinks that feedback produces perfect amps has not understood the basics. And such opinions can safely be ignored. Unless you are selling stuff, of course - than you'll have a field day ;-)
Jan
Very good. I fully agree. However, as soon as we get there, suddenly the need to achieve even lower distortion and Zout figures -by closing the loop- is gone.
And the "huge benefit for audio" becomes somewhat questionable.
I'd really be interested if there is anyone able to detect H2 and H3 below -60dB (flat in frequency as analog_sa has pointed out correctly) in a controlled test.
I certainly am not.
Last edited:
!
Hey, i am just the messenger here 😀
Oh, i see you have removed the quotation marks...shame on you.