THD is a trash measurement which means pretty much nothing. The order of harmonic distortion products is much more telling, and so is THD vs Power. Crossover distortion is said by many to be audible and very un-musical, yet shows low % THD and is only obvious by looking at a falling THD vs Power graph.
Many many many audiophiles have written on the subject that THD is useless, that class A sounds better, that feedback can make amps sound worse, that tubes sound better... etc. While the objectivist camp says this is pure bull, it actually has an objective basis in Phychoacoustics.
I have no idea why so many "informed objectivists" think that because something is electrically "transparent" that it will also be phychoacoustically "transparent". Electrical engineering gives us products that work; phychoacoustics gives us insight and goals on how to make products that sound better.
The human auditory system creates large amounts of harmonic distortion that is monotonic (increasing in harmonic "distortion" with increases in intensity/sound power). Also, any real-world sound increases in harmonic structure/ "distortion" as the object gets louder.
The human brain can filter out the high THD monotonic distortion products created by the ear. Now these distortion products are also falling in intensity vs order. This is why many audiophiles write that lower order distortion products are more begnin. The brain can easily mask them, especially if they are lower in intensity by order than what the ear creates. As order increases, the ear creates less harmonic distortion intensity- and is not as easily "filtered out" by the brain.
In summary, we want monotonic distortion that has harmonic distortion which decreases in power vs order.
Remember that lossy compression such as 320kbps MP3/AAC is indistinguishable from lossless in many cases for many people, yet can be up to 50% THD.
See post #15
GedLee Summa or Abbey12? 90* waveguide, and, distortion testing meaningless? - Techtalk Speaker Building, Audio, Video Discussion Forum
Geddes' white papers on THD being a useless measurement.
Perception
Nelson Pass on Feedback & Distortion
https://passlabs.com/press/audio-distortion-and-feedback
Personally I'm in the Class A SET zero feedback camp. Tubes have more gain with less distortion w/o Feedback than transistors. Their distortion products are monotonic, lower order, and decreasing in power vs order. Just choose one that measures flat and make sure you're getting one that increases in power with lower impedances if you're going to use "normal" speakers with reactive crossovers- such as the decware SE84 and EL34.
I also believe that (good) output transformers from SET's are highly beneficial. Beyond protecting the speakers from amplifier component failure, from my limited understanding they are essentially isolation transformers. Isolation transformers always have high CMRR, reducing noise. This is speculation though.. my knowledge in electrical engineering isn't exactly high. Checkout the performance on these. Take note of the CMRR: http://www.jensen-transformers.com/wp-content/uploads/2014/08/jt-10k61-1m.pdf
Many many many audiophiles have written on the subject that THD is useless, that class A sounds better, that feedback can make amps sound worse, that tubes sound better... etc. While the objectivist camp says this is pure bull, it actually has an objective basis in Phychoacoustics.
I have no idea why so many "informed objectivists" think that because something is electrically "transparent" that it will also be phychoacoustically "transparent". Electrical engineering gives us products that work; phychoacoustics gives us insight and goals on how to make products that sound better.
The human auditory system creates large amounts of harmonic distortion that is monotonic (increasing in harmonic "distortion" with increases in intensity/sound power). Also, any real-world sound increases in harmonic structure/ "distortion" as the object gets louder.
The human brain can filter out the high THD monotonic distortion products created by the ear. Now these distortion products are also falling in intensity vs order. This is why many audiophiles write that lower order distortion products are more begnin. The brain can easily mask them, especially if they are lower in intensity by order than what the ear creates. As order increases, the ear creates less harmonic distortion intensity- and is not as easily "filtered out" by the brain.
In summary, we want monotonic distortion that has harmonic distortion which decreases in power vs order.
Remember that lossy compression such as 320kbps MP3/AAC is indistinguishable from lossless in many cases for many people, yet can be up to 50% THD.
See post #15
GedLee Summa or Abbey12? 90* waveguide, and, distortion testing meaningless? - Techtalk Speaker Building, Audio, Video Discussion Forum
Geddes' white papers on THD being a useless measurement.
Perception
Nelson Pass on Feedback & Distortion
https://passlabs.com/press/audio-distortion-and-feedback
Personally I'm in the Class A SET zero feedback camp. Tubes have more gain with less distortion w/o Feedback than transistors. Their distortion products are monotonic, lower order, and decreasing in power vs order. Just choose one that measures flat and make sure you're getting one that increases in power with lower impedances if you're going to use "normal" speakers with reactive crossovers- such as the decware SE84 and EL34.
I also believe that (good) output transformers from SET's are highly beneficial. Beyond protecting the speakers from amplifier component failure, from my limited understanding they are essentially isolation transformers. Isolation transformers always have high CMRR, reducing noise. This is speculation though.. my knowledge in electrical engineering isn't exactly high. Checkout the performance on these. Take note of the CMRR: http://www.jensen-transformers.com/wp-content/uploads/2014/08/jt-10k61-1m.pdf
Last edited:
Exactly! It is not a big deal if inaudible distortions being easily measurable, for example high THD on a max power. But when measured low numbers represent nasty sounding distortions, for example wide tail of cross-over distortions when the sound decays, it is bad, and no addition of first ones to second ones can lead to "euphonic" result.
Looks like a lot of people explained it well while I was away. 
THD under a few % is not really related to sound quality. It lumps too many things together. At least plotting THD over frequency is a move in the right direction. But these days it's easy to plot the individual harmonics vs frequency, to get an idea of amp behavior. It's important to do that at different levels and loads, if you can.
The 300B as an output tube can be run with a 5K load for low THD, but a 2.5-3K will get a more favorable harmonic spectrum.
An amp with a regular fall-off of harmonics, with none unevenly accentuated or attenuated will be subjectively clean sounding. Human hearing is a little odd that way.
THD under a few % is not really related to sound quality. It lumps too many things together. At least plotting THD over frequency is a move in the right direction. But these days it's easy to plot the individual harmonics vs frequency, to get an idea of amp behavior. It's important to do that at different levels and loads, if you can.
The 300B as an output tube can be run with a 5K load for low THD, but a 2.5-3K will get a more favorable harmonic spectrum.
An amp with a regular fall-off of harmonics, with none unevenly accentuated or attenuated will be subjectively clean sounding. Human hearing is a little odd that way.
Yup this is exactly the problem with feedback and push/pull imho. You're trading a higher percentage of lower order distortion products that are probably inaudible, for a lower percentage of higher order distortion products that are definitely more easily audible.
Now I can see where %THD can be useful- if all distortion products are under your rooms noise floor, reguardless of order. This would mean distortion would be "inaudible" from a component. Chasing this in an amplifier is fruitless though imho, because speakers will probably never have distortion products this low.. not to mention a higher % of lower order distortion is already inaudible in many cases.
Of course we have to take into consideration that feedback also creates many other phenomenon that may have negative effects. Then we have crossover distortion that effects all sounds produced, and effects them most at the lowest power levels. Now when we realize that the first watt of energy from an amplifier has 10x+ the dynamic range of all other watts.. we realize how harmful crossover distortion truly is.
I hope my explainations are easily comprehended by noobs. This information is so essential in explaining why the heavily propagandize information of high-feedback, low % THD, IMD, high power modem amps are absolutely not superior.
High efficiency speakers & low wattage zero negative feedback class A SET amps forever
.
Last edited:
Thank you guys for the comments, but what I was asking is not the low order vs high order distortion.
My question is, why some SET amps have non flat "2nd harmonics" distortion vs freq response (intentionally or not). If high quality OPT is used, it would not be hard to make them flat, but even many of money-no-object type of SETs have high 2nd harmonics at lower frequency. So I wonder if higher 2nd harmonics at bass frequency has any "subjective" or "scientific" advantage, and what is the advantage, if it's so. Or if any disadvantage (trade-off) to make it flat?
My question is, why some SET amps have non flat "2nd harmonics" distortion vs freq response (intentionally or not). If high quality OPT is used, it would not be hard to make them flat, but even many of money-no-object type of SETs have high 2nd harmonics at lower frequency. So I wonder if higher 2nd harmonics at bass frequency has any "subjective" or "scientific" advantage, and what is the advantage, if it's so. Or if any disadvantage (trade-off) to make it flat?
Attachments
Last edited:
Low frequency 2nd harmonic sounds like guitar sound. There are special effects boxes (SS, computerized) that do this specifically for singer's voices. The sound is much richer and harmonious. Often called "sweeter" when referring to amplifiers that behave similarly. The problem with 2nd harmonic enhancement is that it usually results in increased IMD (quite annoying when audible) when more than one instrument or voice is present. Hence typical enhancing (distorting) SET amps usually don't work well with complex orchestral pieces.
One thing about higher harmonic "tails" should be noted. A squarer circuit (2.0 power) will produce a 2nd harmonic line alone. A 1.99 power effector will produce a smoothly tapering off tail of higher harmonics. Both will sound the same, being essentially the same distortion. Smoothly tapering harmonic tails are really just a mathematical artifact of FFT transforms, required to mathematically model various non-linear curvatures. (the FFT uses INTEGER powers to model a signal, and the signal may not conform to an exact integer power law, this results in tapering power series tails to model the signal, purely a mathematical artifice of the FFT integer power modeling process.) So one needs to be very careful in analyzing the FFT signatures of amplifiers.
The common argument about increasing Neg. Fdbk causing higher harmonics is a good example where the transfer function is steadily becoming LESS distorting, but harmonic tails arise in the process. This is NOT a problem at all. It IS a good example of widespread poor understanding of FFT results that is the real problem. The cause of endless arguments over FFT results and N Fdbk effects. (where neither side understands the subtleties)
Now -Isolated- higher harmonic lines ARE of concern, but tapering tails usually not. (depends on phase relationships too)
..
One thing about higher harmonic "tails" should be noted. A squarer circuit (2.0 power) will produce a 2nd harmonic line alone. A 1.99 power effector will produce a smoothly tapering off tail of higher harmonics. Both will sound the same, being essentially the same distortion. Smoothly tapering harmonic tails are really just a mathematical artifact of FFT transforms, required to mathematically model various non-linear curvatures. (the FFT uses INTEGER powers to model a signal, and the signal may not conform to an exact integer power law, this results in tapering power series tails to model the signal, purely a mathematical artifice of the FFT integer power modeling process.) So one needs to be very careful in analyzing the FFT signatures of amplifiers.
The common argument about increasing Neg. Fdbk causing higher harmonics is a good example where the transfer function is steadily becoming LESS distorting, but harmonic tails arise in the process. This is NOT a problem at all. It IS a good example of widespread poor understanding of FFT results that is the real problem. The cause of endless arguments over FFT results and N Fdbk effects. (where neither side understands the subtleties)
Now -Isolated- higher harmonic lines ARE of concern, but tapering tails usually not. (depends on phase relationships too)
..
Last edited:
We already explained you many times that there are no advantages in presence of any errors. Some of them may be more acceptable as trade-offs for other errors that sound nastier.
The Jensen transformer I linked is an example of an extremely good OPT. Even then the distortion rises at the lowest frequencies. It's just the nature of OPT's and I doubt throwing $20,000 on an OPT would yield different results. Again, this is most likely not audible anyway, so who cares?
If you're adamant to get around the OPT, you gotta look to OTL tube amps or Mr. Pass' SIT amplifier. Pretty sure that both will suck up around 300W+ continuously. Better have good A/C
. Nothing against Mr. Pass.Lower order? Okay I'll bite.
Let's use second harmonic which is an octave above the fundamental. So what you are telling me is if we have someone sing A=220 and then another person sings A=440 I won't hear the A=440 because my brain masks it? That would make it pretty difficult for me to play music with others.
Its just that the second harmonic is harmonius, we still hear it though I assure you.
You're ignoring power differences. At 80db, A 440 would have to be greater than or equal to -15dB the SPL of A 220 to be heard.
Again, post #15 GedLee Summa or Abbey12? 90* waveguide, and, distortion testing meaningless? - Techtalk Speaker Building, Audio, Video Discussion Forum
Thank you for your opinion, and I do not disagree with you. I just want to hear opinions from someone who find non flat response has certain advantage over flat one. If I go to an audio forum where people just talk about their sujective preference, I think I can find someone who prefer non flat amp over flat one without knowing what's going on, but I was wondering if someone very knowlegeable in this forum could explain more scientifically why this kind of non lineality is preffered by not a small number of audiophiles. We all know the reason why many people accept or prefer non linearlity such as second harmonics, but I had never heard the reason why low frequency 2nd harmonics is more acceptable or preferable than higher frequency 2nd harmonics.
Last edited:
You've already answered your question. Because it can be subjectively preferred.
FWIW, I have not read a good study on why that might be, but it is. I can also tell you that a strong dose of 3rd harmonic in the low end makes bass sound "fatter" and is generally liked. At what frequency that moves from fat to annoying I don't know. Somewhere around 200Hz would be my guess.
Basically, if you add low order harmonics to the bass, it tends to fatten up the sound and make it richer. That can be pleasant in the bass. For whatever reason, the ear finds audible distortion on low sounds pleasant, but not so much on higher sounds. Perhaps it's like any other sound. Play some pure sine tones and note their subjective effect. Which is more annoying, a 100 Hz tone or a 1000 Hz tone? Would added harmonics in those ranges be similar?
FWIW, I have not read a good study on why that might be, but it is. I can also tell you that a strong dose of 3rd harmonic in the low end makes bass sound "fatter" and is generally liked. At what frequency that moves from fat to annoying I don't know. Somewhere around 200Hz would be my guess.
Basically, if you add low order harmonics to the bass, it tends to fatten up the sound and make it richer. That can be pleasant in the bass. For whatever reason, the ear finds audible distortion on low sounds pleasant, but not so much on higher sounds. Perhaps it's like any other sound. Play some pure sine tones and note their subjective effect. Which is more annoying, a 100 Hz tone or a 1000 Hz tone? Would added harmonics in those ranges be similar?
If you have the Radiotron Designers Handbook (4th edition), look at pages 616 and 676.
There this subject is presented under the name "Synthetic Bass".
RDH4 can be downloaded here: http://www.tubebooks.org/Books/RDH4.pdf
On the reproduction side of things, its more ideal to have a characteristic that complements the Flecture-Munson curve.
Only is they're both singing perfect sine waves and the 440 is pianissimo.
Assuming the harmonic distortion is audible, I have a theory that this creates the effect of the missing fundamental. Basically you'll hear a tone an octave lower than the fundamental.
https://en.m.wikipedia.org/wiki/Missing_fundamental