grt@twitch.UUCP ( G.R.Tomasevich) (02/21/86)
> 1 (hack, cough, wheeze) speakers. It wasn't that bad at low volume. > However, most car stereo amplifiers have gross phase distortion (pull out > your Ford factory radio and run a proof on it) as well as a frequency The human ear is insensitive to phase and will not be affected by phase distortion, provided the amplitude vs frequency characteristic is not altered. The RIAA equalization for records produced horrendous phase shifts at some frequencies, but we don't care. We do speech processing here, so we are aware of what one can do to speech, at least. Some people in this department have done 'hi fi' coding, too. The cheap amplifiers have other problems, probably including intermodulation distortion. -- George Tomasevich, ihnp4!twitch!grt AT&T Bell Laboratories, Holmdel, NJ
mat@mtx5a.UUCP (m.terribile) (02/24/86)
> > 1 (hack, cough, wheeze) speakers. It wasn't that bad at low volume. > > However, most car stereo amplifiers have gross phase distortion (pull out > > your Ford factory radio and run a proof on it) as well as a frequency > > The human ear is insensitive to phase and will not be affected by phase > distortion, provided the amplitude vs frequency characteristic is not altered. This is conventional wisdom. And it's true -- phase distortions are irrelevant to the intelligibility of speech, and to music so long as other distortions predominate. But in much of today's really good equipment, other distortions do not always predominate. Tests a while ago (someone know the actual reference) indicated that phase distortions CAN cause subtle but audible changes in timbre when they occur between about 100 and 2000 Hz, with the most sensitive region being about 600 to 1200, if I recall. You also mention the RIAA equialization doing violence to relative phase. Well, perhaps this is what the vinyl-over-CD freaks are so used to? (or why Carver was able to measure funny differences between vinyl and CD?) -- from Mole End Mark Terribile (scrape .. dig ) mtx5b!mat (Please mail to mtx5b!mat, NOT mtx5a! mat, or to mtx5a!mtx5b!mat) ,.. .,, ,,, ..,***_*.
rdp@teddy.UUCP (02/24/86)
In a recent article, George Tomasevitch <ihnp4!twitch!grt> asserts: > > The human ear is insensitive to phase and will not be affected by phase > distortion, provided the amplitude vs frequency characteristic is not altered. > The RIAA equalization for records produced horrendous phase shifts > at some frequencies, but we don't care. We do speech processing here, > so we are aware of what one can do to speech, at least. Some people in > this department have done 'hi fi' coding, too. The cheap amplifiers have > other problems, probably including intermodulation distortion. > -- Sorry, Mr. Heyser [1][2][3], sorry Mr. Fincham [4], sorry Mr. Lian [5], sorry Herr Linkwitz, sorry Messrs. Lipshitz, Pollock, and Vanderkooy [6] and sorry to all those once capable researchers who had determined that many kinds of phase shift are audible, Mr. Tomasevitch has, with the flourish of keyboard, made the ear "insensitive to phase". Poor phase, ears are no longer sensitive to it! <technical flame on> The example given, "RIAA equalization for records introduces horrendous phase shifts at some frequencies, but we don't care" is ill chosen and an obvious red herring. First of all, the RIAA equalization does not produce "horrendous" phase shifts, as the absolute magnitude of the phase shift never exceeds 90 degrees (no more than one pole or zero in effect at any one time). Secondly, Note that the phase errors introduced should be the exact opposite of those impressed on the signal by the RIAA pre-equalization and playback characteristics of the reproduction system. The result is a net phase shift of 0 degrees across the audio band. Hence, the reason we don't care about phase shift in RIAA networks is because there shouldn't be any! Much speech processing is done using mere spectral energy vs. time and, in these measurements, phase is indeed not an issue (although I shan't pretend to tell Mr. Tomasevitch about his business, as he MAY know more about it than I). He may well be correct, in a very limited context, but 50 years of psycho- acoustical research indicates that phase distortion is quite audible. Certainly such a blanket pontification such as "the ear is insensitive to phase" falls flat on its face in light of the well-demonstrated fact that the ear needs very accurate phase information at midrange frequencies to localize apparent sound sources [7] (this is how stereo works!). I might be willing to swallow a statement such as, "Under the conditions of high noise levels, ambiguous reverberent fields and lacking the ability to sit in an optimum position, as found in an automobile, and given the lack of attention of the vast majority of radio stations and recording companies to issues such as phase, and given that many car stereo systems have speakers that have terrible phase characteristics, especially given typical placement, I feel that other anomolies in the reproduction system may have more significance than phase." <technical flame on low, waiting for responses> Dick Pierce But a few references: [1] Heyser, R. C., "Loudspeaker Phase Characteristics and Time Delay Distortions," J. Audio Eng. Soc., Part I - vol. 17, no. 1, pp. 30, 1969 Jan., Part II - vol 17, no. 3, p. 130, 1969 Mar. [2] Heyser, R. C., "The Delay Plane, Objective Analysis of Subjective Properties," J. Audio Eng. Soc. Part I - vol. 21, no.11, pp 609- 701, 1973 Nov., Part II - vol. 21, no. 12, pp. 786-791, 1973 Dec. [3] Heyser, R. C., "Determining the Acoustic Position for Proper Phase Response of Transducers," J. Audio Eng. Soc., vol. 32, no. 1, pp. 23-24, 1984 Jan. [4] Fincham, L. R., "The Subjective Importance of Uniform Group Delay at Low Frequencies," J. Audio Eng. Soc., vol. 33, no. 6. pp. 436-439, 1985 June. [5] Lian, R., "Is Linear Phase Worthwhile?," presented at the AES Conv., 1981 May, preprint no. 1732. [6] Lipshitz, S. P., Pollock, M., and Vanderkooy, J., "On the Audibility of Mid-Tange Phase Distortion in Audio Systems,", J. Audio Eng. Soc., vol. 30, no. 9, p. 580, 1982 Sep. [7] Moir, J. "Speaker Directivity and Sound Quality," Wireless World, vol. 87, no. 1541, pp. 32-38, 1981 Feb.
spp@ucbvax.BERKELEY.EDU (Stephen P Pope) (02/27/86)
I don't remember the reference, but one psychoacoustic study showed that, for speaker crossovers, the phase distorion was found to be inaudible (under the condtions of this study) so long as the frequency response is smooth in the region where the phase change vs. frequncy was high. If the frequency response was uneven, the phase shift was very noticible. Hence the conclusion, human hearing is insensitive to phase shift if the frequncy response is flat. However, this obviously can't be generalized to the multiple radians of phase shift found in the passband of a sharp-cutoff LPF, as the CD makers have found out. There's also the irrelevant fact that phase carries no information in speech. steve
ron@brl-smoke.ARPA (Ron Natalie <ron>) (02/28/86)
> > 1 (hack, cough, wheeze) speakers. It wasn't that bad at low volume. > > However, most car stereo amplifiers have gross phase distortion (pull out > > your Ford factory radio and run a proof on it) as well as a frequency > > The human ear is insensitive to phase and will not be affected by phase > distortion, provided the amplitude vs frequency characteristic is not altered. > The RIAA equalization for records produced horrendous phase shifts > at some frequencies, but we don't care. We do speech processing here, > so we are aware of what one can do to speech, at least. Some people in > this department have done 'hi fi' coding, too. The cheap amplifiers have > other problems, probably including intermodulation distortion. Well you telephone people's ear (singular) may be insensitive to phase, but that makes up part of how we percieve sound direction. In addition, if you have two sound sources (like stereo, maybe?) their phase relative to each other IS essential as making random changes is phase will cause different perceived sounds when it's summed down into even your single ear. -Ron
mohan@topaz.RUTGERS.EDU (Sunil Mohan) (02/28/86)
In article <12077@ucbvax.BERKELEY.EDU> ... > > ... one psychoacoustic study > showed that, for speaker crossovers .... > .... hearing is insensitive to phase > shift if the frequncy response is flat. However, this obviously > can't be generalized ... > > There's also the irrelevant fact that phase carries no information > in speech. Yes, but do all these studies that conclude phase carries no or little info also consider DIRECTION of source as relevant information ? Such a study could well be looking for only likeness to live sound. If the reproduced-sound source in the experiments was monophonic, for instance, then directionality is obviously not being taken into account. It would be interesting to see what the 'conditions of study' are for such experiments. -- Sunil UUCP: ...{harvard, seismo, ut-sally, sri-iu, ihnp4!packard}!topaz!mohan ARPA: Mohan@RUTGERS
djb@riccb.UUCP (Dave J. Burris ) (03/01/86)
> > Tests a while ago (someone know the actual reference) indicated that > phase distortions CAN cause subtle but audible changes in timbre when > they occur between about 100 and 2000 Hz, with the most sensitive > region being about 600 to 1200, if I recall. > The keyword here is SUBTLE. Most source agree that a CONSTANT phase shift is not distinguishable within X degrees (not sure of the number. I can attest to the fact that phase cause timbre changes. Can you say phase shifter? Can you say flanger? Can you say digital delay? I knew you could! -- Dave Burris ..!ihnp4!ihopa!riccb!djb Rockwell Switching Systems, Downers Grove, Il.
levy@ttrdc.UUCP (Daniel R. Levy) (03/03/86)
<Oh oh here it comes. Watch out boy, it'll chew you up! \ Oh oh here it comes. The LINE EATER! [Line eater]> In article <12077@ucbvax.BERKELEY.EDU>, spp@ucbvax.BERKELEY.EDU (Stephen P Pope) writes: >I don't remember the reference, but one psychoacoustic study >showed that, for speaker crossovers, the phase distorion was >found to be inaudible (under the condtions of this study) >so long as the frequency response is smooth in the region where >the phase change vs. frequncy was high. If the frequency >response was uneven, the phase shift was very noticible. >Hence the conclusion, human hearing is insensitive to phase >shift if the frequncy response is flat. However, this obviously >can't be generalized to the multiple radians of phase shift >found in the passband of a sharp-cutoff LPF, as the CD makers >have found out. >There's also the irrelevant fact that phase carries no information >in speech. >steve Way back when I was in undergrad EE, I learned that the effect of making a phase shift in one component of some waveforms could, if the shift was correct, cause the waveform to have extremely sharp (theoretically tend- ing to infinity) peaks. This would pose an obvious problem for the faith- ful reproduction of those waveforms, even setting aside the issue of whether it "sounds the same." (Which I doubt it would, since it would push the ear itself to mechanical extremes during the peaks.) Note I am not saying that the TOTAL ENERGY of the peaks tended to infinity (obviously impossible when starting out with a signal of finite power) but just that the INSTANT- ANEOUS (current, voltage, power) level did. -- ------------------------------- Disclaimer: The views contained herein are | dan levy | yvel nad | my own and are not at all those of my em- | an engihacker @ | ployer or the administrator of any computer | at&t computer systems division | upon which I may hack. | skokie, illinois | -------------------------------- Path: ..!{akgua,homxb,ihnp4,ltuxa,mvuxa, vax135}!ttrdc!levy