http://www.diyaudio.com/forums/soli...audio-power-output-stages-44.html#post2036835
Very nice post, Brian.
Very nice post, Brian.
Member
Joined 2009
Paid Member
This thread developed nicely overnight, very interesting !
My brain continues to tell me that large amounts of NFB is generally a desirable feature, reducing overall distortion and that middle of the road NFB levels is not desirable because it has the potential to accenuate the relative magnitude of high order harmonics without being able to push them into the noise floor. I know this is a generalization but I acknowledge the dangers with it being so.
This is interesting because this is consistent with most of the 'complaints' I read about NFB - that it harms enjoyment of the higher frequencies. So I have to wonder why ? - and if I understand what I read the answer is the accuracy of the NFB is insufficient because of a) phase error (sorry if this is the wrong terminology still) and b) inadequate NFB due to bandwidth limitations of the amplifier.
And as pointed out by AKSA, c) accuracy of the error amplifier.
I assume (?) the first two of these limitations are related to the output stage (boy, that puts this thread back on topic almost !!!) and so all I need to do is ensure enough bandwidth-gain product to high frequencies I will have good HF performance from my amplifier ?
And this means fast output devices (1 point for Lumba)
Why does this sound too easy ?
My brain continues to tell me that large amounts of NFB is generally a desirable feature, reducing overall distortion and that middle of the road NFB levels is not desirable because it has the potential to accenuate the relative magnitude of high order harmonics without being able to push them into the noise floor. I know this is a generalization but I acknowledge the dangers with it being so.
This is interesting because this is consistent with most of the 'complaints' I read about NFB - that it harms enjoyment of the higher frequencies. So I have to wonder why ? - and if I understand what I read the answer is the accuracy of the NFB is insufficient because of a) phase error (sorry if this is the wrong terminology still) and b) inadequate NFB due to bandwidth limitations of the amplifier.
And as pointed out by AKSA, c) accuracy of the error amplifier.
I assume (?) the first two of these limitations are related to the output stage (boy, that puts this thread back on topic almost !!!) and so all I need to do is ensure enough bandwidth-gain product to high frequencies I will have good HF performance from my amplifier ?
And this means fast output devices (1 point for Lumba)
Why does this sound too easy ?
Interstingly, the latest Stereophile (the amplifer special) has some reviews that can be summed up as follows:-
Marantz 40 watt amp - looking at JA's test measurements, this is clearly a high feedback design. Got a great review
Aesthetix - a tube/SS zero GNFB - got a good review
A Bryston - a high feedback design (again, deduced from the plots) - not so good review
and so it goes on . . . .
Marantz 40 watt amp - looking at JA's test measurements, this is clearly a high feedback design. Got a great review
Aesthetix - a tube/SS zero GNFB - got a good review
A Bryston - a high feedback design (again, deduced from the plots) - not so good review
and so it goes on . . . .
Because in many "high-feedback" designs, the loop-gain rolls off from a fairly low frequency. The treble sounds worse because there is less feedback (and hence less less distortion reduction) in the treble.
Interesting reply. I have to add that i could not properly evaluate the time shift (which i prefer to call smearing) of the NFB due to the transducers used in the evaluation, nor the "speed". I will with some new silicon toys and proper ultralow distortion tweeters in the next 6 months.
I think there may be other factors that slipped to us.
First and foremost is clean power to the OPS but not only to it.
The lowest distortion amplifier was also much more powerful (not that the other wattage was a bottleneck, though) and the PSU was not entirely optimized (this is a big factor, on which Piercarlo is on the same line of thought). It did feel faster though.
hmmm
As Bob said, bad use and misunderstood of an ingenious engineering tool, giving room for all these nonsense theories over there...
regards,
Hi Lumba,
No disagreement there, but the reality is that there is always some amount of interaction. A good example is when a designer uses only a Darlinton output stage instead of a Triple. The load presented to the VAS by a Darlington output stage is only about 20k when driving a 4 ohms load. This may dominate the output impedance of the VAS.
Going to a Triple (Locanthi T circuit) is what I always advocate, but a bit more care is sometimes needed with the Triple to keep it from instability.
Cheers,
Bob
btw, can everybody mention his prefered bipolar power devices?..
and give some clues about the reason of these choices...
personnaly, i used the 2SA1302/2SC3281 pair extensively
and recently a few 2SA1943/2SC5200 as their replacement parts..
they were a great improvement at the time,i just remember
the audible improvement it gave compared to the then reigning
MJ1500X that was stuck in almost every us made amp...
currently ,i use few bjts , as i settled long ago
for the hitachi/renesas power mosfets, although i still
use sanken s MP1620/MN2488 darlingtons in simple designs
and as replacement parts even for amps that have a discrete darlington..
and give some clues about the reason of these choices...
personnaly, i used the 2SA1302/2SC3281 pair extensively
and recently a few 2SA1943/2SC5200 as their replacement parts..
they were a great improvement at the time,i just remember
the audible improvement it gave compared to the then reigning
MJ1500X that was stuck in almost every us made amp...
currently ,i use few bjts , as i settled long ago
for the hitachi/renesas power mosfets, although i still
use sanken s MP1620/MN2488 darlingtons in simple designs
and as replacement parts even for amps that have a discrete darlington..
Lumba,
This is not really true. There is no such thing as "conservation of errors" in negative feedback. Be cautious with your analogies between physics and electrical engineering.
However, what you may be thinking of is how noise-shaping DACs and ADCs work. They do in fact re-shape the noise spectrum and put the noise power up to a part of the frequency spectrum well above hearing.
Cheers,
Bob
No, he speak against NFB for itself, whitout any distinction.
If you mean that time problems have been underestimated for 49 years, then we agree
The problems, acting as claimed by Lumba, are actually inexistent out of its mind.
See my other post, and also tiefbassuebertr's who explained better than me. Even Self measured THD higher than 20khz
If you like to think yourself able to ear ultrasonic frequency, feel free of believe it.
Human ear sensibility is just as plotted many decades ago by Fletcher- Munson. Other sensibilities are undimostrated (except from audio-gurus obviously!
). Fletcher-Munson diagram shows clearly that human sensitivity peak just in a narrow band between 1-3 kHz. lower and upper frequencies perception is strongly attenuated. If hypothetically you are able to ear frequencies of about 20 kHz you need to boost them of about 15-20 dB for hearing it with the same intensity of a 3 kHz signal.Ah, for clarity: Fletcher-Munson curves and those updated in ISO 226 (2003) gave audibility of frequencies above 15 kHz just as "estimated", what means they were unable to find enough peoples for confirming data and that human beings that really ear those frequencies are more exceptions than rules.
Second armonic above 7 kHz is just the only DIRECT armonic you can hear (pheraps you can hear a well boosted third if you have really - but REALLY, in physician terms, not audiophile terms!
- good ears... not more). Other armonics, if not processed by intermodulating non linearities that downmix them in audible band, are for human beings simply inexisting.The TRUE reason for which SOME instrumentally bad amplifier (not ALL as usually many guru claim for) soun "better" of some other really better amplifiers i've yet explain in the previous reply and, adversely the beliefs of many actracted by alleged "audio misteries", there is nothing of misterious: is simply psichoacustics.
You are optimistic. The effective noise floor of many common GOOD appliances are about 0.05 %, what roughly lay in the 70 dB S/N region. Not so rare as it should be, noise floor lay in 60 dB SN region, what is equivalent to a 0.1 % (NOTE: these level are to be intendend at normal earing level - about 0.1 to 1 Watt - not at maximum output level).
If you think about the similarity of these level and the "misteriously good, bad amplifiers" level of distortion, you may acquire a first clue of that really happen. That is appearing "unnatural" and "unpleasant" is simply the ABSENCE OF PERCEIVABLE NON LINEARITIES. As proven in many deprived sensorial experiences, lacks of stimulus or their plain absence may appear really "more loudly" (and discomforting) of a mild noisy environment.
And this is true also in music listening. The reasons of this discomfort stem from a simple fact: DON'T EXIST ANY KIND of recorded sound event that really match ACTUAL sound event as happen in world. Any recorded event - no count how good is the equipment used in doing it - may appear less or more pleasant than other... but it NEVER appear "natural" because it's NOT natural. And a TOO clean audio reproduction simply reveal it. This is also the reason because many peoples take in disdain the headphones (especially of closed type): because headphones, keeping out environmental noise, slam in the ear many recording artifacts that usually are interstitially filled by domestic noises and similar.
The reason of that is partially contained in the text you've quoted. The key point that usually few in audio take in adequate count, is what really matter in audio listening: ratios among signals, not their absolute values. A thing usually considered only when facing with compressed music that in reality count EVERLY in ANY audio listening situation. Our earing-brain system should be considered more similar to a "demodulator" instead to a linear adcquiring systems. A "demodulator" which we may figure acting as the old thelephonic modems that, from a stream of "noise" decoded all data we see on monitors when connected to web: images, letters, sounds, and so on.
If we don't dismiss our wisdom of seeing listening as a linear process where perceptions are "linearly" connected to (supposed) unknown technical limits in amplifier and the rest of audio chain, we simply will continue to understand nothing of the whole matter. And we'll struck against audio phenomena as this: why a CD reversed on tape cassette sound more pleasurable of original IN SPITE OF THE MEASURABLE quality WORSENING apported by the process? Why adding a slight amount of random noise on the floor of digital recording, these become more "analogized"? Why reversing a Long Playng on CD we obtain a more "edible" sounding result than what is obtained from direct reversing of original audio masters? Until we'll continue skipping to answer at these questions (as to many others similar of course!) we'll don't understand anything and we'll continue to recriminate about "time", "global feedback" and so on that really count as a fly on horse's head. Today others things really matter.
The point about PSU is extremely good and you know how much we agree on this. why? Because it is the main cause of HF distortions
... and this drive to another "big question": why old tube circuits (but also many classic solid state preamplifiers) were so highly refined in their power supply section whence "modern" solid state (often Op-Amp based) amplifiers use, more often than it appear justified, minimal part count power supplies? Old circuits were heavily exposed to "injuries" caused by poor SVRR ratio and this of course lead the old designers to be more accurate in care of it. But we are really sure about "immunity" of modern solid state circuits just because they can afford to use huge amounts of NFB for artificially increasing SVR starting from a usually poorer one? Or pheraps even here we need some rethinking, similar to that performed on other topics about NFB (mis)use?
Hi
Piercarlo
Again you don't know the matter about you are speaking. NFB can only redistribuite signal components spreading it on spectrum, not energizing it, otherwise it became POSITIVE not negative feedback.. By itself NFB can't ADJOINT NOTHING. And raising signal noise relating floor or something that "restrict dynamic range" is postulating that NFB add NEW non linearities and also INCREASE the existing ones. And this is plainly false. First of all from mathematic viewpoint.
Piercarlo
I disagree. He was being pessimistic.
Most amps can achieve a noise floor better than 100dB below maximum output and some approach -120dB.
If we refer these to 1W output (for a 100W amplifier) then we are talking noise ~-80dB to -100dB below 1W, i.e. 0.01% to 0.001% of 1W
Hi Brian, nice to you back again
Thank you. And a Happy New Year to both of you.
I really like what they've done to the look and feel of the site.
Brian
The 0,01% i was referring to the typical hobbyst appliances used to measure distortion - Piercarlo interpreted it right. I should have used a semicolon.
About amps i would really like to have the chance to measure the exact same amp with some "trick" to change its distortion (only quantity, not distribution!) from lets say the above 0,01% to 0,02% and 0,005% (yes at 1W/1m). That woud be enough i guess to answer my original question.
Andrew, can you remind me the formula of -xxdB to % for distortion levels? thanks.
About amps i would really like to have the chance to measure the exact same amp with some "trick" to change its distortion (only quantity, not distribution!) from lets say the above 0,01% to 0,02% and 0,005% (yes at 1W/1m). That woud be enough i guess to answer my original question.
Andrew, can you remind me the formula of -xxdB to % for distortion levels? thanks.
it s 0.000001 % to 0.00000001 % of 1W , since the s/n ratio is
a voltage ratio, not a power ratio..
The exactly name of the according article about the "T-Circuit" is follow:
"An Ultra-Low Distortion Direct-Current Amplifier" from the author
BART N. LOCANTHI (James B. Lansing Sound, Inc.
http://www.harman.com/EN-US/OurComp...ip/Documents/Scientific Publications/1091.pdf
This article now I have found (I search for a long time, but I don't know the appropriate keywords until this day - thank you, Bob
An other truck FULL of metropolitan legends... What's this? Special discount offers after Xmas holidays?
Piercarlo
PS - At best, for now, we can enjoy of living in post stalinian era: your opinions are, in electronic, the exact equivalent of that were the Lysenko's thoughts about genetics.
10% = 0.1 = -20dB,
1% = 0.01 = -40dB,
0.1% = 0.001 = -60dB,
0.01% = 0.0001 = -80dB,
0.001% = 0.00001 = -100dB,
0.0001% = 0.000001 = -120dB.
dB value = 20 * Log [voltage ratio]
If the voltage ratio >1 the dB value is +ve.
If the voltage ratio is <1 the dB value is -ve.
Most calculators get this right.
Once more (sigh)...
I dont hear the ultrasonic frequences by themselves, but the armonics of lower frequencies and all this content is elaborated by the brain. Feel free to not believe it (i know most people dont).
Until some more studies will be done on this subject, i think it is a moot point.
Human ear sensibility is just as plotted many decades ago by Fletcher- Munson. Other sensibilities are undimostrated (except from audio-gurus obviously!
Correct. But something is changing, albeit slowly. For instance, I read sometime ago an AES paper where finally they recognized that the ambient noise do NOT cover completely the dynamic range.
The ear is able to hear sounds below the noise floor and the brain is able to distinct them. Alleluja!
This is not so unrelated to the noise floor/thd topic under debate here.
You just stated a known fact.
Please note that the tests were done with simple signals, which are good at finding objective measurements such as sound intensity but not for distortion.
If/when they'do some analog study comparing DISTORTION types --possibly on more complex signals-- you'll see some surprises. I bet anything you want on the following: the ear is more sensitive to higher frequency distortion than it is to low freq ones -- assuming a point where the sensitivity according to those curves IS THE SAME.
Second armonic above 7 kHz is just the only DIRECT armonic you can hear (pheraps you can hear a well boosted third if you have really - but REALLY, in physician terms, not audiophile terms!
The TRUE reason for which SOME instrumentally bad amplifier (not ALL as usually many guru claim for) soun "better" of some other really better amplifiers i've yet explain in the previous reply and, adversely the beliefs of many actracted by alleged "audio misteries", there is nothing of misterious: is simply psichoacustics.
It is not psychoacustics, unless your definition is "any elaboration that the brain does to the sounds perceived by the ear".
I already know i'll have to rent some serious equipment for a week or such.
I disagree, to a point.
With a blameless (really) reproduction chain and some digital manipulation it is possible to recreate at a certain approximation a sound that our ears/brain can classify as "natural"... or at least that's my challenge.
This is a very good explanation, and i believe it's true.
And this also explain why -let's say- ear sensitivity to higher order armonics is not linear but exponential.
Also agreed.
Take for instance some top of the line Sansui SS amps of the 80s, before they had to make budget-forced compromises.