But there are a lot more opportunity for intermodulations (rather than harmonic distortions) to happen, and for them to fall in the audible ranges, with audio reproduction gear. For instance, third order harmonic distortion of anything above about 6kHz isn't in the audible range, but 3rd order intermod of 7k and 8k will fall at 6k and 9k, and both distortions are from the same nonlinearities.
Not anywhere near as much as politicians, religious fanatics, philosophers . . .
In a previous post, someone said something more or less to the effect that engineering is a particularly bad area for "beliefs," presumably as opposed to going on science and physical facts. It seemed to me there are some other areas that have produced as bad or worse results by going on "beliefs." Makes engineering that way look not so bad.
I like to design to better than what one piece of gear's listening distortion detection threshold is. As I say every couple years..... we listen to a whole string of stages from mic preamp thru console and recording and playback system to speaker. If each amplifying stage was 1-2% it wouldnt sound so great in the end. So I/you need to design any piece in the chain to be much lower so the end result will remain under the detection threshold (<.05% for me).
THx-RNMarsh
Last edited:
Agree IMD is qualitatively worse and is produced by when multiple frequencies pass through a non-linearity. However, the IM product amplitudes (sum and differences), as I recall, are maximum when the original input frequencies are of equal amplitude. Anyway, my point is that different amounts of HD and IMD my be produced in a given situation. If what you end up with is mostly low order HD, then you may not be very good at noticing it in a listening test. So, when we are talking about the limits of distortion people can hear, some questions arise, such as what it the exact makeup of the particular distortion, what is the age and gender of the listener, and has the listener had any training to recognize distortion? Of course, there are likely other questions as well, but I think you see it could get complicated to sort out all the variables.
It is an interesting view but:
Musical instruments have inharmonic overtones to begin with.
Inharmonicity
sure, no ones going to go to the effort to make an amp so bad you need all of psychoacoustic lossy codec theory to justify
well, except for the Schiit Ragnarok bias scheme
Schiit Audio Ragnarok integrated amplifier Measurements | Stereophile.com
well, except for the Schiit Ragnarok bias scheme
Schiit Audio Ragnarok integrated amplifier Measurements | Stereophile.com
Sure, the piano is an interesting example, and the inharmonicity undoubtedly accounts for much of its distinctive sound.
One of the most baffling of designs. What a terrible idea!sure, no ones going to go to the effort to make an amp so bad you need all of psychoacoustic lossy codec theory to justify
well, except for the Schiit Ragnarok bias scheme
Schiit Audio Ragnarok integrated amplifier Measurements | Stereophile.com
But it has by and large been done. I don't have a comprehensive bibliography, but some in here may.
Harmonic distortion of a single-line instrument is hard to hear if it is of low order and not gigantic in magnitude, as it many times is equivalent (although with a different curve of growth with level) to a mere change in equalization. An RF engineer wrote an infuriating piece for Electronics World which provoked an email I should have slept on, in which he took audio people to task for supposedly ignoring IM. I pointed out that there were plenty of IM measurements in audio, despite not finding their way to manufacturer's specifications very often. I said a good deal more and my and Self's truncated missive got Ian Hickman to lament the decline in civility in the LTEs.
I have noticed that people who don't mind, or even claim to prefer harmonic distortion, or worse insist that everyone likes it (that idiot in aXp who alleged this and presented a single-stage JFET stage buffered in and out with unity-gain buffers comes to mind) tend to listen to rather specific material, often solo vocals with sparse instrumental accompaniment. When I auditioned one person's system, I searched for any program material with which I was familiar out of a wall of CDs. I finally found ONE orchestral recording, some Pops orchestra favorites one. I didn't know it but I knew what most of the pieces ought to sound like.
The general audibility of harmonic distortion has been known for more than 75 years. In the 'Radiotron Designers Handbook' 1941, etc the ORDER of the harmonics was given a subjective limit that holds up today. So it is NOT the ability to measure harmonic distortion, but that harmonic distortion is but one parameter, and once it is well controlled, other factors intrude.
This was first pointed out by Richard Heyser, more than 45 years ago. He wrote a number of obscure articles trying to set the stage for understanding this to both audiophiles and audio designers alike.
He came up with some surprising conclusions even back then, like global negative feedback is a big subjective detriment to audio sound quality. He found this out in the late 50's and early 60's by experiment. He always said: "If any 2 people can hear an audio problem, it is real."
In acoustics, he said that even a 1 foot time delay between individual loudspeaker components could possibly be audible. Yet, he like me, used a single K-horn as our reference speaker though the 60's and early 70's, which has a lot more time delay between speakers than 1 foot. This is because it had other qualities, like low harmonic distortion, and low FM distortion as well.
Unfortunately, Richard Heyser died relatively early in his career, and if he had lived longer, audio would be in a better place, and other measurement approaches would have been developed that better approximated human hearing discernment.
This was first pointed out by Richard Heyser, more than 45 years ago. He wrote a number of obscure articles trying to set the stage for understanding this to both audiophiles and audio designers alike.
He came up with some surprising conclusions even back then, like global negative feedback is a big subjective detriment to audio sound quality. He found this out in the late 50's and early 60's by experiment. He always said: "If any 2 people can hear an audio problem, it is real."
In acoustics, he said that even a 1 foot time delay between individual loudspeaker components could possibly be audible. Yet, he like me, used a single K-horn as our reference speaker though the 60's and early 70's, which has a lot more time delay between speakers than 1 foot. This is because it had other qualities, like low harmonic distortion, and low FM distortion as well.
Unfortunately, Richard Heyser died relatively early in his career, and if he had lived longer, audio would be in a better place, and other measurement approaches would have been developed that better approximated human hearing discernment.
Compare them to what? I have in the past been able to compare with live sound, having heard a concert live and then heard the next night's concert at home on the BBC - and heard the same hall (RAH) in each case.kamis said:
No. Do you?
As I have said more than once: if you think we are measuring the wrong things, then try listening to music on a system with 2.5kHz bandwidth and 25% THD (and maybe -30dB noise). You will soon hear that we are measuring the right things. There may be more things we ought to measure too, but the things we are measuring are relevant to good sound reproduction quality - although perhaps not to everyone's music production taste.john curl said:
Very tiny amounts are inaudible. Small amounts may be preferred by some people. Problems only occur when they claim that their preferred distortion has better fidelity than smaller distortions.Markw4 said:
Last edited:
It is amazing that I get the most criticism from people who don't design audio products professionally. Real audio designers know what I am talking about.
For the record: Back in the 1960's, our test equipment consisted of either wave analyzers or SMPTE IM analyzers. Usually for repair and optimization, SMPTE IM was about all you needed for electronics, either tube or transistor. However, as time went on, we found that SMPTE IM was NOT measuring a critical parameter in amplifier design, and that was slew rate induced distortion or TIM. Matti Otala, Walt Jung, and many others, including me, started to find ways to measure this problem in the mid 1970's. We found that either special test signals like the sine-square test waveform or extended sine wave measurement up to 50KHz or so, would give us consistent results. SMPTE IM was still valuable, but only for very low frequency performance, let's say 10Hz-1000Hz.
About this time one could get a THD meter that was both inherently low distortion and over a reasonable frequency range. The two main manufacturers were HP and Sound Technology that brought out the THD testers that are normally used even today. HP also built wave analyzers that were even more precise, but unfortunately more difficult to get a quick result from, but they were still very valuable for analog tape, or for Sine-Square TIM testing. However, they were very expensive and usually reserved for test labs. These instruments could measure down to about -100dB. This is usually more than enough to adequately measure discrete, relatively low feedback electronics, but not very effective with high feedback, and usually fairly noisy IC circuits. We could hardly measure anything, except slew rate generated distortion, with this older test equipment. However, as far as harmonic or IM distortion was concerned, it was more than enough.
Today, using FFT techniques, we can measure much lower amounts of distortion, as well as removing inherent noise from the distortion spectrum, and we get fine results, but they don't seem to mean much, subjectively. We still have to listen to our designs IC or otherwise, to find out if we have been really successful. This is where people cannot understand WHY some circuitry that measures in the -80dB or so, still sounds great, yet other circuitry that might measure -120dB or better doesn't sound so good. It is NOT the distortion residual that creates the subjective difference, but OTHER FACTORS. Finding these 'other factors' is one of the greatest challenges for audio designers, today. We still have a long way to go. Hopefully, someone of Richard Heyser's academic and intuitive stature will come forth and put us back on track.
For the record: Back in the 1960's, our test equipment consisted of either wave analyzers or SMPTE IM analyzers. Usually for repair and optimization, SMPTE IM was about all you needed for electronics, either tube or transistor. However, as time went on, we found that SMPTE IM was NOT measuring a critical parameter in amplifier design, and that was slew rate induced distortion or TIM. Matti Otala, Walt Jung, and many others, including me, started to find ways to measure this problem in the mid 1970's. We found that either special test signals like the sine-square test waveform or extended sine wave measurement up to 50KHz or so, would give us consistent results. SMPTE IM was still valuable, but only for very low frequency performance, let's say 10Hz-1000Hz.
About this time one could get a THD meter that was both inherently low distortion and over a reasonable frequency range. The two main manufacturers were HP and Sound Technology that brought out the THD testers that are normally used even today. HP also built wave analyzers that were even more precise, but unfortunately more difficult to get a quick result from, but they were still very valuable for analog tape, or for Sine-Square TIM testing. However, they were very expensive and usually reserved for test labs. These instruments could measure down to about -100dB. This is usually more than enough to adequately measure discrete, relatively low feedback electronics, but not very effective with high feedback, and usually fairly noisy IC circuits. We could hardly measure anything, except slew rate generated distortion, with this older test equipment. However, as far as harmonic or IM distortion was concerned, it was more than enough.
Today, using FFT techniques, we can measure much lower amounts of distortion, as well as removing inherent noise from the distortion spectrum, and we get fine results, but they don't seem to mean much, subjectively. We still have to listen to our designs IC or otherwise, to find out if we have been really successful. This is where people cannot understand WHY some circuitry that measures in the -80dB or so, still sounds great, yet other circuitry that might measure -120dB or better doesn't sound so good. It is NOT the distortion residual that creates the subjective difference, but OTHER FACTORS. Finding these 'other factors' is one of the greatest challenges for audio designers, today. We still have a long way to go. Hopefully, someone of Richard Heyser's academic and intuitive stature will come forth and put us back on track.
Last edited:
I'd have to dig it out myself, as I recall the conclusions are more about how multitone IMD tests compare, nonlinear system modeling including Volterra
not "0.xx% can't be heard"
and their focus was on pro loudspeakers, stadium levels where air nonlinearity induced distortion can be measured
Last edited:
And yet one of the (currently*) most prolific and most successful audio designers around at the moment eschews all this bull**** and just concentrates on getting the performance right by using huge amounts of negative feedback. He breaks every one of your rules and wins the awards, and most importantly the OEM contracts.
* One cannot currently compare his works in quantity to the lifetime output of some of the greybeards of audio, but he may well catch them up.
Of course if you have a bunch of amplifiers with 2N3055's and psuedo-complementary outputs you might make any conclusion. You state this over and over, I would like a cite where he states that with the meaning you keep giving it.
- Status
- Not open for further replies.