Thanks for the link, seems to be an interesting article. I am often saying that we examine deterministic periodic test signals, which is contrary to the nature of musical signals.
Hopefully I speak Russian
This paper stresses on possible un-deterministic properties of the transfer function, and its influence on output signals. Sine wave has minimum information, and any even narrow transmission channel can handle it. While music is an information dense signal, and limited channel transmission possibilities, caused by stochastic component of transmission function, can be a reason that information at the output is not complete.
Are you sure this is not just a way of saying that if you know the non-linerarity and it is time invariant you can undo it mathematically? If the distortion is bad enough the problem rapidly gets intractable and ultra sensitive to noise. This is similar to Weiner filtering in optics or inverting almost singular matricies.
Hello, Scott
It is difficult for me to comment this professionally, I am sure you are write as for the important role of numerical artefacts. But Chumakov speaks on an analytical level, and states about the basic impossibility to recognize information in its complete form, if the transfer function is not comletely deterministic.
I attract attention to this paper, because I was impressed by complete coincidence of the conclusions made in the paper, with my listening experience and intuition. I for already few years do not care about linearity, but take all possible measures for lowering noise, use external power supplies, multystage RC filtering in PS, diode switching noise suppression, high quality parts, no NFB or very short-loop and fast current NFB, all-jFET schematics, SE class A, etc. All these measures are aimed at diminishing stochastic components of the transfer function.
I was very encouraged that my intuition has got analytical support. For a long time I listened, that distortions have a sub-class, called losses. Now I can imagine, about what losses people spoke.
Last edited:
Here is another paper on stochastic nonlinearity
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.77.2980&rep=rep1&type=pdf
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.77.2980&rep=rep1&type=pdf
http://www.filigrane.ru/Chumakov_Stochastic_Nonlinearity_EAT.pdf
it works on a Mac with Google translate:
The analysis of the electroacoustic channel as a channel for information transfer.
A model of the influence of distortion components in auditory perception channel audio
of information.
Evaluate the effect of deterministic and stochastic nonlinear distortions
on the subjective perception of audio information.
The technique of measurement and the percentage for stochastic nonlinear
distortion power amplifiers, audio-frequency measurements based on feedback
intermodulation AMI (RIMD).
Presented and discussed results of the measurements and subjective assessment of quality
amplifier designed with a view to ensuring the requirements for the reverse
nonlinearity, and without such records.
Formulated quantitative eligibility criteria and the subjective perception of the measured parameters of the stochastic nonlinear electroacoustic tract.
it works on a Mac with Google translate:
The analysis of the electroacoustic channel as a channel for information transfer.
A model of the influence of distortion components in auditory perception channel audio
of information.
Evaluate the effect of deterministic and stochastic nonlinear distortions
on the subjective perception of audio information.
The technique of measurement and the percentage for stochastic nonlinear
distortion power amplifiers, audio-frequency measurements based on feedback
intermodulation AMI (RIMD).
Presented and discussed results of the measurements and subjective assessment of quality
amplifier designed with a view to ensuring the requirements for the reverse
nonlinearity, and without such records.
Formulated quantitative eligibility criteria and the subjective perception of the measured parameters of the stochastic nonlinear electroacoustic tract.
This translation may be misleading. RIMD means Reverse Intermodulation Distortion, and it is measured by injection of 2nd signal to the output. This is not about "feedback intermodulation".
"If the signal in the channel adds noise, the signal point adopted shifted from its true (expected) values. Since the direction and magnitude bias can not be predicted, the noise creates a small region of uncertainty around the "true" signal point. If, due to the effects of noise signal received "Hit" is not in your area "web" of the receiver - this is a mistake and part of the transmitted information is lost. Equation (3) just determines the number of potentially distinct areas of uncertainty for each dimension of the "web" receiver. However, the unpredictable shift signal point from its true situation can be caused not only by noise but also by other processes in the channel vtransmission (electro-tract). These random effects can be further reduced to an equivalent noise. "
"Thus, the presence in the channel transmission Information stochastic constituents, regardless of the nature of their of (additive noise, random modulation transfer function) reduce the throughput ability the channel. It means that if the rate of arrival information from the source will be more this bandwidth, the transmitted information can not be accepted in full and part of it will be distorted."
"There will probably be appropriate to recall a much lower visibility low-order harmonics (2nd and 3rd) in comparison with high-order harmonics. One of the reasons may be that a proper hearing has nonlinearity short order, thus adapting to the nonlinearities of electroacoustic tract similar order is much more efficient. It should be emphasized that it is not a masking product nonlinearities of electroacoustic tract products of their own non-linearity of hearing, and the almost complete compensation products such total nonlinearity. Uncompensated Products deterministic nonlinear path made by color, due to the curvature the signal space Stochastic distortion principle can not be compensated for any hearing, or any techniques. Their presence leads to a decrease in information bandwidth of electro-tract. This means that that if you attempt to transfer this channel of information at a rate greater than bandwidth, such a transfer is unreliable [4] and some of its is lost."
Pavel feel free to correct any bad Google translation. I think these quotes are clear enough.
"Thus, the presence in the channel transmission Information stochastic constituents, regardless of the nature of their of (additive noise, random modulation transfer function) reduce the throughput ability the channel. It means that if the rate of arrival information from the source will be more this bandwidth, the transmitted information can not be accepted in full and part of it will be distorted."
"There will probably be appropriate to recall a much lower visibility low-order harmonics (2nd and 3rd) in comparison with high-order harmonics. One of the reasons may be that a proper hearing has nonlinearity short order, thus adapting to the nonlinearities of electroacoustic tract similar order is much more efficient. It should be emphasized that it is not a masking product nonlinearities of electroacoustic tract products of their own non-linearity of hearing, and the almost complete compensation products such total nonlinearity. Uncompensated Products deterministic nonlinear path made by color, due to the curvature the signal space Stochastic distortion principle can not be compensated for any hearing, or any techniques. Their presence leads to a decrease in information bandwidth of electro-tract. This means that that if you attempt to transfer this channel of information at a rate greater than bandwidth, such a transfer is unreliable [4] and some of its is lost."
Pavel feel free to correct any bad Google translation. I think these quotes are clear enough.
Hi,
Sounds to me like Hawkesfords fuzzy distortion.
Antique news.
Showed up with noiseloading testing in the 80's.
Ciao T
Sounds to me like Hawkesfords fuzzy distortion.
Antique news.
Showed up with noiseloading testing in the 80's.
Ciao T
PS,
Incidentally, if you use FFT and average multiple samples "fuzzy distortion" magically disappears, sadly the human hearing seems to be "averaging challenged".
Ciao T
Incidentally, if you use FFT and average multiple samples "fuzzy distortion" magically disappears, sadly the human hearing seems to be "averaging challenged".
Ciao T
clear enough to say it is needless elaboration on detailed description of the noise while Shannon-Hartley Channel Capacity Theorem always applies?
are there measurement of this "stochastic nonlinearity" in high performance audio amps? - I'd look at the noise floor in FFT with multitone test signals - when has noise modulation been an issue in decent amps? - no averaging needed
the output "Interface Intermodulation" was analysized by Cherry more than 2 decades ago and nothing "interesting" was found - feedback reduces it just fine at audio
are there measurement of this "stochastic nonlinearity" in high performance audio amps? - I'd look at the noise floor in FFT with multitone test signals - when has noise modulation been an issue in decent amps? - no averaging needed
the output "Interface Intermodulation" was analysized by Cherry more than 2 decades ago and nothing "interesting" was found - feedback reduces it just fine at audio
Last edited:
Actually I prefer the one about the motorist who gets a flat tire. Pulls over to change it and takes off the hub cap. Then removes the lug nuts and places them in the hub cap. Pulls off the old tire. Puts on the new tire and turns around to grab the lug nuts. He just misses, taps the hub cap and the lug nuts flip up, roll off and all go right down a drain grate he didn't notice.
He starts looking around and realizes he is stopped right outside the fence of a mental institution. There is a crowd at the fence that has been watching him! He is at a loss for what to do. One of the inmates starts to speak to him.
"Just pull a lug nut off of each of the other three wheels and you can get going to buy more!" The motorist is speechless. The inmate looks at him again and says "Around here we may be crazy, but we are not stupid!"
He starts looking around and realizes he is stopped right outside the fence of a mental institution. There is a crowd at the fence that has been watching him! He is at a loss for what to do. One of the inmates starts to speak to him.
"Just pull a lug nut off of each of the other three wheels and you can get going to buy more!" The motorist is speechless. The inmate looks at him again and says "Around here we may be crazy, but we are not stupid!"
Oh Danny boy (Cheever), the pipes, the pipes are ca-hal-ling.
From glen to glen..and down the mountain side.
This paper seems heavy on the mathematical analysis, but does not seem to delve into psychoacoustics very much. Perhaps it is the translation from Russian into English or the choice of passages to quote that is responsible for the lacking.
In any case, the conclusions seem opposite of human perception. Our ear and brain system is quite capable of ignoring uncorrelated noise in a system. That is why tape is often superior in perceived sound compared to digital recording, despite that vastly inferior dynamic range of tape and relatively high noise floor.
On the other front, nonlinear distortion introduces new frequencies into the signal which our hearing system can often interpret as increased loudness. Not that increased loudness sounds bad, but it certainly is not faithful to the original. One of the reasons that vinyl can sound 'better' than digital recording is this amplitude-dependent harmonic distortion, which acts to make louder passages sound even louder than they should - or effectively acts as dynamic range expansion to the human listener. We all know that there is a human hearing system preference for 'louder' sounds, so any distortion that introduces new frequencies in higher critical bands compared to the original signal will sound louder and thus 'better', and if this effect is amplitude dependent then it will even serve to sound more 'dynamic.'
Anyway, the paper (as quoted) reads more like a study on data compression and information preservation, totally without regard for how the human hearing system interprets noise versus harmonic distortion.
This is not to say that reducing noise is not a valid endeavor. An attentive listener is more likely to perceive noise, even though it can easily be ignored (such as when listening to music while driving a car). I certainly welcome the signal-to-noise ratio improvements made in the transition from analog tape to digital tape. I also prefer a quiet car to a noisy one if I am going to listen to music. However, the effect on listening of nonlinear distortion is a bit less readily perceived consciously, at least not without the help of scientific testing of human hearing perception. There are quite many situations in which nonlinear distortion serves to 'enhance' the sound in ways that are perceived as 'better' - but I would argue that these nonlinear systems are less faithful to the original.
If your goal is to enhance music, then a focus on reducing noise while completely ignoring nonlinear effects is probably valid. However, if your goal is to faithfully reproduce music, then I suggest that you need to educate yourself on the ways the nonlinear distortion can fool your human hearing system into preferring a slightly distorted reproduction. As Nelson Pass often alludes, those 10,000 hours of listening are important for training the ear and brain together. I would add that it helps to learn about psychoacoustics while gaining those listening hours, so as to better distinguish aesthetically pleasing distortion from less distorted and more faithful reproductions.
- Status
- Not open for further replies.