Hey, I do that and I get somewhere with it! I've been doing it for 40 years though.
I typically use back of the envelope calculations and then I build a prototype. If the circuit doesn't do what I want I always know why, because I understand the limitations of my simplified algebra. Then I change a value or three and measure again. I always get where I want to be with a few trys.
Last edited:
Keith Howard, a very clever and reasonable man, did some listening tests, I believe single-blind, to find out if the Hiraga criteria held water. He used headphones and compared various distributions of harmonics he created as files to as completely a clean signal and as completely a distortionless signal chain as he could manage, and acknowledged that the differences were slight, but that he consistently preferred clean. I don't recall how close to or below the noise floor his clean signal was.
I'm tempted to categorize the Hiraga folks, however well-meaning and committed, as a subset of the effects box folks, with lots of handwaving about evolutionary developments and psychoacoustical speculations to make things seem deeper and more scientific. Conspicuously absent is support from real testing. And the latter is a whole lot of work and a whole lot less fun---but will be needed to convince those who are not already true believers.
John Linsley Hood used to trust the judgment of his cat. If it left the room there was something wrong. As appealing as that is, and as much as I love cats, I fear it is insufficient.
Olive suggested to me recently that no one as yet has done controlled tests on even the audibility of loudspeaker distortion! I'm sure some will cite studies that say otherwise, but with the requirements of qualification he would impose, I suspect he is right.
There is work to be done. In the meantime, enjoy the music and enjoy your hobby, and please be cognizant of expectation biases.
If they're low, which they are in any competent amp from the past several decades, they're inaudible. So the "harmonic progression" thing is inapplicable. That may be offensive to some, especially to those who are emotionally or financially invested in amplifier design, but that's what experiments show.
This has been troubling me. Whilst driving home from work. Ignoring any stability issues that might want you to need gain up there, by 60KHz your tweeters are down 30dB+ and your ears are down 140dB+ so unless there is some IM going on which mixes back into the audio band I must be missing something.
Yes, until you consider intermodulation and the possible products in the audio band.
By the same logic, we could do very well with amplifiers having an ULGF of 21KHz, who cares for any loop gain beyond this.
Did Keith Howard's test simulate distortion percentage increasing with level (music level varies constantly over a huge range) like real amplifiers do? Saying "x% distortion" describes an amp in test, not one in use for listening. You can't play an amp rated at 5% distortion and say you've 'heard' 5% distortion. The distortion could have been infinitesimal at the levels reached, average or peak. Or it could have been near 100% distortion at the average level if a distortion type like crossover notch distortion was involved.
I don't know of any amp that gives the same percentage distortion product levels for each input tone regardless of its level, such as would happen with the FFT multiple trick or simple seeming "add 1% distortion to everything" approach. Distortion in amps varies with level, as does distortion in ears. Distortion in amps is also relatively smooth vs. frequency. Not so speakers. I don't know about ears.
I don't know of any amp that gives the same percentage distortion product levels for each input tone regardless of its level, such as would happen with the FFT multiple trick or simple seeming "add 1% distortion to everything" approach. Distortion in amps varies with level, as does distortion in ears. Distortion in amps is also relatively smooth vs. frequency. Not so speakers. I don't know about ears.
Here is a more complete text of his argument:
"But most people's hearing mechanism ear/brain work the same way and if one is heavily exposed to live acoustic music then one knows what a natural sound is like...IF that person cares and IF that person has good skills in equating that sound with what is heard through a recording.
Most people I know who switched did so for convenience reasons knowing full well they were taking a step down in quality for the sake of a harmonious personal life. I know others besides myself that tried D and went back again to tubes. Hell, I tried three times with my own cash. I have heard at length the latest and greatest and I am not willing to go a fourth round based on what I hear...it is unnatural sounding for the very reasons Ralph (and several researchers in the field) have outlined. Unnatural harmonic BALANCE = unnatural sound. Cheever explains it quite well in his thesis. It is not just even or odd harmonics but like what Jean Hiraga said long ago a monotonic pattern (so even and odd where the next higher odd is always lower than the previous even in an exponential decay). This means a pattern with only even or only odd is wrong and anything with high order harmonics above say 9th (and then only at high volume levels) is also not following a natural pattern.
"This has been redrawn for clarity in Fig 2-3. Notice that the ear creates significant levels
of the second harmonic, nearly 10% of the fundamental for sound pressure levels (SPL's)
of 90dBA and above. Also the slope of the harmonic reduction versus input reduction
varies with the harmonic power, beginning at approximately 1:10 for the 3rd harmonic to
1:1 for the 9th harmonic."
"A reduced SPL range is shown. Even for the moderate S.P.L. of 80dBA, the 2nd
harmonic is at the equivalent of 65dBA or normal voice level, and the 3rd at 45dB. This is
still ~40dB above the mid-band threshold of hearing, yet one does not hear the
harmonics! Only a single pure tone is heard. The ear/brain appears to be able to
completely suppress the sound of a range of harmonics if they conform to this specific
pattern."
"If
the harmonics do not follow this pattern, the ear brain indeed detects these as new tones.
Therefore, for all but extreme frequencies and sound pressure levels, any electronics that
generate this harmonically consonant envelope will be transparent."
So, follow the pattern or get unnatural sound. It really is as simple (and as difficult) as that.
"We cannot disregard this function of the ear. For example, if the ear is presented
with an auxiliary sound distorted with a set of harmonics that are consonant with the
aural harmonics at 100 dBA but the actual sound pressure level of the fundamental is say
10 or 100 times (10 or 20dB) less, it will be perceived distorted"
Did you get this? The pattern changes with SPLs and if the distortion pattern matches a higher SPL but the fundamental is a lower SPL it will be perceptible as distortion.
"The relative deviation between an amplifiers distortion harmonics and the aural
harmonics, per harmonic, must better quantify the subjective sound quality of an
amplifier. In reviewing many different amplifiers I found that their harmonic signature
did not follow the aural harmonic envelope. Universally the distortion has high order
harmonics without the next lower order harmonics' complementary level. Contrary to the
history and evolution in audio design, high order harmonics, if they appear, MUST be joined by a family of lower order harmonics that follow the aural harmonic envelope."
Like all things in human perception it is about PATTERN recognition and not absolute levels. You need a 2nd followed by a 3rd that is 20db lower followed by a 4th, 5th etc. each with at least 10db lower than the previous. Very few amps get close to following this ideal.
Putzey's, like a lot of engineers, think that if they can supress everything enough it will be effectively invisible but there is always enough residual that it is not the case and since the residual doesn't follow the pattern it is audible even if it is very low in level. Putzey's and a lot of other engineers are trying to work against nature rather than working with nature to get the best sound.
Geddes reached similar conclusions but without all the ear analysis but with blind listening tests nonetheless.
So, in the end, it is about the application of correct SCIENCE to produce technology that most people would find to be natural sounding. Of course self-deception, marketing, convenience (weight, heat, size) and drivers other than pure sound quality will always exist and give a reason for unnatural sounding devices to exist (mostly driven by price because to truly follow the recommendations is invariably expensive).
Putzey's surely has impressive ENGINEERING but he is not following what has been determined by science and so his goals are misplaced. Audio is one of the few fields of technology where scientific understanding has been largely dismissed. It is frankly a bit bizarre.
It is only so subjective because so few are following good science when thinking about audio engineering decisions. You then get all kinds of flavors of wrong. I have said it before and will say it again...MOST amps are wrong...some a bit wrong some very wrong. Note I didn't specify a type. The ones I have found to be the least wrong are Class A tubes or hybrids. This also fits with the measurements from these types of devices. "
Hi Mr.Curl !
I think you are right ,
Atmasphere design come from that old electrical engineering school knowledge where maximum power transfer occur when ( Z generator = Z load ) .
Same rule as is in use in RF emission field today , except that now we have to match not only output Z of transmitter to antena system Z, but RF feeder power lines Z too for maximum RF power transfer .
For example,
one of my friend with his DIY OTL tube power amps power magnetostat Apogee Dueta LS (4 ohm) with exceptional great sonic success .
Those DIY amps are over 100W RMS each and are directly based on Atmasphere design , output power tubes are Russian 6C33C-B power triodes , each amp use 6 those power tubes in OPS , and those amps work without any inserted GNFB .
Curiosity is that hi try to power both Apogee magnetostats connected in parallel (2ohm) from just one OTL power amp ,and again there was no any sign of any sound degradation or amplifier overload at all .
Last edited:
Agreed in principle s long as we can come to some sort of agreement of what low is. Certainly if we are talking, say, 100dB below, then we're on the right track, but even then there are questions and caveats. Most of these I would chalk up to other mechanisms, namely intermodulation and transient behaviour, and this is because knowing what the instantaneous dynamic range of the ear is as well as audibility curves are, one would expect 100dB to be overkill yet something like that ballpark figure is often used to insure a transparent amp - just like we 'exaggerate' by providing a decade or so extra bandwidth on each end of the 'audio band'. While these practices are not really in question, there is a real question here: why some (note, SOME) amps with apparently horrendous THD (using the above rules of thumb) sound remarkably transparent to the ear.
While the above rules of thumb do indeed offer a path to a transparent amp, they do, as the saying goes, limit the number of ways one can skin a cat (don't ask me about the metaphor, it's an English language thing 😛).
Without going into too many examples I've had to deal with, I do have to say that some forms of distortion (like compression) can be captivating, for instance compression makes quiet details in music appear louder, therefore giving the illusion of more clarity and detail from a recording. There are a LOT of people who call themselves audiophiles yet do not recognise this at all. Some forms of harmonic distribution that generate small amounts of high order distortions, and also transient oscillations can sometimes also produce a similar effect. There is something about that cat leaving the room, I have experienced the effect myself numerous times, where at first everything seems so clear and defined only to produce a feeling of general discomfort and nervousness after longer listening. These things are quite difficult to measure, but if you narrow down what you suspect is happening, a load and signal scenario can be constructed where they can be observed.
I have also had a case where harmonic progression was 'ideal' according to Hiraga, but the amp sounded downright weird, like a very convincing synthesizer playing a MIDI track. It turned out there was a transient 'chase' condition between the follower output stage and it's error correction circuit because a bootstrap connection to the driver was inadvertently left connected to the output of the follower. This same amp at one point in it's development used a small amount of GNFB and looked really good on measurements but sounded completely dull, like someone put a blanket over the speakers - here the trap was too little available extra gain to produce a net positive effect from applying GNFB to lower the amplitude of the harmonics vs creating extra harmonics, something which Baxandall once wrote about.
I think the article was in Stereophile, so it should be pretty easy to find. If I could persuade him to comment in here I would, but he might be understandably reluctant. He did respond to an email I sent on a related topic.
Keith impresses me as very level-headed and not agenda-driven. And he would see through "simple-seeming" approaches. Had the results favored the "organic" Hiraga-like distortions, he would have been candid about the results. And he is not in the amplifier business as far as I know.
But his results suggest that, absent crossover distortion and especially absent clipping distortion, cleaner is preferred---by him at least.
EDIT: http://www.stereophile.com/reference/406howard/index.html#xd6G9WoQl57Lit8O.97
It was not single blind, my apologies. And from 2006---how time flies.
Last edited:
There are a remarkable number of poorly designed amplifiers out there, both of high GNFB and zero GNFB designs. And many of them sound pleasant enough. That's just a fact of preference and the human ear being remarkably tolerant to certain distortions.
What is sad is that it means that many take a defensive position of their preferred distortion patterns rather than actually just fessing up to liking it that way and accepting its not hifi.
What is sad is that it means that many take a defensive position of their preferred distortion patterns rather than actually just fessing up to liking it that way and accepting its not hifi.
as far as the ear goes
I think I mentioned elsewhere the use by some composers like Telemann, of the IM distortion of the ear-brain to produce quite-audible difference tones from two single-note instruments---in GPT's case, usually recorders. The amount of energy due to the nonlinearity of the air is tiny, but the quite-loud tones are readily heard by the players and nearby listeners. Benade shows some musical examples in his popular science book Horns, Strings, and Harmony.
When some years ago Cerwin-Vega hired a bunch of folks from the former Soviet Union, there were some papers produced by them about actual air nonlinearity mechanisms, with iirc some prescriptions for changing the seating arrangements of big bands to alleviate the effects. Somehow it didn't catch on...
I think I mentioned elsewhere the use by some composers like Telemann, of the IM distortion of the ear-brain to produce quite-audible difference tones from two single-note instruments---in GPT's case, usually recorders. The amount of energy due to the nonlinearity of the air is tiny, but the quite-loud tones are readily heard by the players and nearby listeners. Benade shows some musical examples in his popular science book Horns, Strings, and Harmony.
When some years ago Cerwin-Vega hired a bunch of folks from the former Soviet Union, there were some papers produced by them about actual air nonlinearity mechanisms, with iirc some prescriptions for changing the seating arrangements of big bands to alleviate the effects. Somehow it didn't catch on...
Hiraga's 20dB 2nd to 3rd harmonic step requirement seems to also ignore the effect of level on amplifiers. At what level is this required? I've measured a few non-gnfb SET and SS SE amps and with each one, the harmonic levels got closer to each other as levels increased, becoming almost equal across the board at a little above clipping.
RF amps don't usually have the luxury of lots of feedback because of physical delays (30 year retired RF engineer here). Almost universally, the HD2 increases 2dB for each 1dB of level. HD3 (or IMD3 more usually) goes 3dB per dB level. 4th goes 4:1, etc. This behavior is so predictable that distortion in RF or microwave amps is characterized by the concept of "intercept point", which is the (imaginary) power level at which the distortion product would equal the output level if the trend could keep up (of course it gives out near clipping).
RF amps don't usually have the luxury of lots of feedback because of physical delays (30 year retired RF engineer here). Almost universally, the HD2 increases 2dB for each 1dB of level. HD3 (or IMD3 more usually) goes 3dB per dB level. 4th goes 4:1, etc. This behavior is so predictable that distortion in RF or microwave amps is characterized by the concept of "intercept point", which is the (imaginary) power level at which the distortion product would equal the output level if the trend could keep up (of course it gives out near clipping).
That one sentence sums up the whole problem. Obviously this is an impossibility but even allowing for accidental choice of wording, how do you know that what you are measuring correlates with "not sounding different". There's been so much discussion already about how many decades ago a study concluded that THD is a very poor indicator of SQ. One smart-**** has insisted that "enlightened engineers" don't aim for lowest THD any more but has remained silent when asked what exactly they do use. This thread is supposed to be about the pros and cons of GNFB and there's nothing interesting about reading the statement "lowest THD and highest SNR etc are the only aspects of an amplifier that matter and your ears can't tell what sounds most like the original" expressed in twenty different ways.I want to defend engineers. I have a degree in electronics engineering and so far it has never made me think that human perception is something to be avoided and dismissed (maybe because I'm not professionally designing amplifiers?😛). We are not all closed-minded egotists. A good engineer sees the big picture and tries to find the best solution to the problem. In our case most of us want an amplifier that amplifies signals in such a way as to introduce the least amount of audible distortion. Until we have universal agreement on things that can be measured that have a meaningful correlation to what we perceive we will keep on going around in circles in this way. Those that insist that listening to amplifiers is pointless and prefer to use test instruments to measure THD+N vs freq etc only to determine how good they are should stick to threads devoted to achieving best possible conventional measurements instead of trying to derail potentially interesting discussions with egotistical condescending statements implying anyone prepared to listen or approach amp design with an open mind is deluded or wants "tone controls".
Is that intended as a statement of fact? In which case on what basis have you concluded they are poorly designed?
Have you scientifically proven that particular distortion patterns are irrelevant to accurate sound reproduction? If not who are you to say a particular distortion pattern is "not hifi"?
I think this sort of characterization by people of engineers is a strawman, reminiscent of Dr. Dre and Jimmy Iovine's insistence that they are guys with white lab coats and clipboards. I don't know anyone short of strictly-simulation jockeys who don't care about listening. Maybe I haven't been hanging out with enough geeks and nerds.
One of the problems, however, is that expectation bias dooms all who insist on listening to their "offspring" without any controls. It is a phenomenon particularly demonstrable with loudspeakers---you see them, you judge them independently (unless something really terrible is wrong) of what they sound like in a given room. You just simply cannot peek---but not to do so is beyond most anyone's capabilities.
But if we then layer this with spurious theories about why something sounds like we think it does, we're off to the races. If those theories include an approach to reproduction equipment framed as some sort of evaluation of a musical instrument, we're inviting further confoundings.
Making a flat-frequency-response and distortionless-below-clipping amplifier, a goal that many are approaching with considerable success, can sound like a pursuit of some cold lifeless thing. It is not. If you find deviations from this to be pleasing, go for it. But don't build them into the amplifier itself. It's the classic case of varying a plurality of things at once before figuring out the effect of a particular one.
Good. Show me audibility at -80dB, for that matter. Bill Waslo's Sousa test as well as Geddes's results suggest an even higher threshold for anything remotely resembling the distortions of real, physical amplifiers.
This is consistent with decades of ears-only tests, which keep showing the same boring things, over and over- unless you peek, any two decent modern amps engineered for linearity and flat frequency response will sound identical if they're level matched and not clipping. They can (and will) have different distortion distributions. You can design amps that have frequency response anomalies and/or high distortion and they will sound different. But that's irrelevant to amps designed to have the output replicate the input.
This is a problem that was solved decades ago.
