pd (12/16/82)
I have some comments on the The Audiophile vs. Digital Recording debate. The audiophile's position, that I have heard often, is "It doesn't sound right, so there's something wrong with it." To which the Digital enthusiast retorts : "All audio signals can be expressed as a sum of sine waves with frequencies up to 20 Khz; and Digital techniques can accurately reproduce sine waves up to 20 Khz. So Digital recordings can accurately reproduce all Audio signals." At which point the audiophile retreats, baffled and beaten. However. Fourier Transforms attempt to MODEL a REALITY. There are merely a way of mathematically representing a REAL phenomenon. Fourier never claimed (like the good Scientist he was) that his frequency domain representation of an audio signal (for example) was a complete explanation of reality. A real scientist attitude would be not to reject the audiophiles perception of poor quality of Digital recordings because it didn't fit within Fourier- and z-transform models, but to devise repeatable, comprehensive experiments to gather data on the Audiophiles' dislike of digital recordings, and then attempt to explain it! Who knows. We might actually learn something! Prem Devanbu
ark (12/17/82)
I have yet to see documentation of a carefully constructed double-blind experiment in which a panel of "experienced listeners" consistently found a well-made digital recording to sound inferior to the best analog recordings.
karn (12/17/82)
A properly designed digital audio system records waveforms in the time domain by sampling them at a rate greater than twice the highest frequency component in the input signal. It doesn't matter what the sampling rate is, so long as it is at least the Nyquist rate (2x f) and the anti-aliasing input filter doesn't distort or roll off the highest frequency desired input signal. A 200 khz sampling rate would be a complete waste of bits. The human ear CANNOT distinguish between a 20 khz sine wave and a 20 khz non-sinusoidal wave, unless your hearing extends to at least 40 khz. This is an established fact. For this reason, a digital audio system's inablility to reproduce 20 khz square waves means nothing. As far as phase balance errors between channels are concerned, I find this hard to accept as a real problem. If both channels are not sampled simultaneously, this is fine so long as the D/A converters reconstructing the output signals update their outputs with the same timing. If not, this would be an inexcusable design error which could have been easily avoided. I'm sure that analog tape heads introduce more phase balance errors simply by being out of azimuth alignment than any production digital recording system. To show the effect of azimuth alignment on channel-to-chanell phase delay, try this experiment: Play a high frequency tone (a head alignment tape always has this) with the output channels combined (mono). Adjust the head azimuth. You will hear "beats" as the adjustment is made; this is the result of the two tracks going alternately in and out of phase as the angle of the playback head is adjusted. The proper adjustment is to peak the tone in the "center" beat; I usually find that I can get to the proper point by adjusting azimuth first on music known to have been recorded by a well-aligned deck, peaking the high frequencies by ear, then making the fine adjustment with the high frequency alignment tape. Only a small change in azimuth would be sufficient to cause a 10 degree phase imbalance between channels at 10-15 khz. Back to digital audio, my point is that almost any shortcoming of digital audio, imagined or otherwise, is far more likely to actually occur in conventional analog recorders. Try aligning even a good quality analog deck at high speed with good tape, and you'll be amazed at how imprecise they really are compared with digital's known performance. Again, I challenge anybody to produce a properly controlled, double-blind study that shows 50 khz, 16 bit digital audio to be inferior to the best possible analog recording techniques. Phil Karn
FtG (12/17/82)
This digital discussion reminds me of the great hi-fi debate of so many years ago. It seems the "experts" knew that there was no point in reproducing high frequencies since human beans couldn't hear them. Well... somebody performed an experiment using live musicians (as opposed to dead ones?) separated by a system of baffles from the listeners. By adjusting the baffles, high frequencies could be damped out. The listeners noticed a significant loss of fidelity. The experts hadn't taken into account the huge difference between detecting a single tone and "sensing" a full range of harmonics. Thus hi-fi was born. (Coming next week, Uncle Ferd will tell you the story of stereo!!!) It seems that the digital people should know about these experiments and taken into account the complex nature of hearing. In any case, I can easily imagine the experts to be wrong again. FtG
thomas (12/17/82)
At the risk of being flamed off the net (watch out for the Geisthounds!), I
will add my two cents to this debate.
1. How many of you can hear 20kHz anyway? Can you hear the whine of a TV
set (15.75kHz)? If so, can you hear the ultrasonic burglar alarms in
some retail stores (K-Mart has them around here) (about 18kHz)? If you
answer no to either of these questions, then you have nothing to worry
about when it comes to 20kHz signals (whether sine or square).
2. You certainly can't record a 20kHz square wave on most analog recording
media, either. Most serious listeners have cassette decks these days.
How many cassette tapes have a frequency response above 20kHz? How
can you say "digital is inferior because it won't reproduce a 20kHz
square wave" when your current medium won't either? If somebody can
show that the human ear can distinguish between a sine wave and a
square wave at the upper limit of hearing, then this is a motivation
to improve things. Note that feeding a sine wave and a square wave
into a speaker and listening for the difference doesn't count as a
proper comparison (as somebody pointed out), because of non-linearities
in the reproduction system, a square wave can actually cause some
lower frequencies to be generated by the speaker. A valid comparison
would involve generating sine and square pressure waves in the air.
3. Even listening comparisons must be done carefully. I used to work for
a small digital audio firm, and we got one of a competitor's recorders
in once (a large recording firm had a large backlog because editing
using this competitor's system was very difficult, and they asked for
help in getting a couple recordings edited). The converters in this
recorder were so badly adjusted that they were introducing about 20dB
worth of noise into the recording and reproduction! Although the
theoretical S/N ratio was 96dB, the actual measured S/N was 70dB.
Needless to say, this made music played on this system not sound as
good as it could have. It would be a shame if sloppy manufacturing
practices ruined the potential of digital recording.
4. For those of you who claim that listening is the real test, I have done
a lot of listening to REAL digital recordings (not the record pressings,
but the actual bits themselves). They have been, in general, the best
reproductions of music I have ever heard. Instruments such as trumpets
and drums, which suffer badly in most traditional recording media,
sound just like trumpets and drums. The imaging is fantastic (arguing
against those who claim that digital recording will necessarily
introduce huge phase shifts). When I closed my eyes, I could imagine
that I was sitting in the ideal seat in an otherwise empty concert hall,
listening to a flawless performance (since mistakes can be edited out).
It was the best listening experience I have ever had.
I for one, can't wait for home digital (but when I get mine, I'm going to take
it to my friendly local digital music company and get those D/A converters
adjusted, that's for sure).
Digitally yours,
Spencergwyn@BRL (12/17/82)
From: Doug Gwyn <gwyn@BRL> If in fact only frequencies from 0 through 20KHz are relevant for audio perception, and if in fact correct Fourier transformation is being done to audio signals (sampled at 40KHz or above), then the reconstructed signal WILL match the original. This is not a model, it is one of the most solidly-established results of mathematics. I note that when true "high fidelity" first became available, there were those who thought IT sounded "wrong". It appears to depend on what one is accustomed to. In the case of recordings, analog disks are typically compressed and digital disks typically are not; this is certainly audibly detectable. There are audiophiles who claim to be able to hear the difference between equipment with 0.001% THD and 0.01% THD; this while putting their output into loudspeakers that cannot do nearly that well! Some of think they are hallucinating.
VaughanW.REFLECS@HI-Multics (12/18/82)
A few random comments: 1. Throughout the Sixties, a lot of audiophiles complained that transistor amplifiers didn't sound as good as tube amplifiers. They were uniformly ridiculed by people who claimed that steady-state frequency response charts told the whole story, and on the basis of these charts, transistors were better than tubes. I don't really need to remind anyone that in the late Seventies the audiophiles were proved to have been correct, when TIM (transient intermodulation) distortion was discovered. Although the 1965-vintage transistor amplifiers had far less harmonic distortion than a 1959-vintage tube amplifier, their TIM distortion was orders of magnitude higher. The lesson is that good steady-state figures are necessary but not sufficient. 2. The human ear is not sensitive to sinusoidal waveforms much above 18kHz. If we believe in Fourier transforms (which is much like believing in 1+1=2: it's ordinary mathematics), we may conclude that the ear is also insensitive to repetitive signals of any waveform and whose repetition rate is significantly above 18 kHz. But Fourier transforms don't deal with transients. A waveform must be repetitive for Fourier analysis to apply. Besides, the ear is known to be exquisitely sensitive to very small phase diferences in arriving waveforms: One can accurately, with closed eyes, locate an arriving sound \in a vertical plane/ by using the differential delay lines built into the external ears (ref. Scientific American articles some years ago) - where the differences in wavefront arrival time may be measured in microseconds! So much for the Fletcher-Munson curve. 3. The mathematical analysis of waveforms is a relatively well-understood field. Not so the perception of sound, which involves elements of psychology and the organization of neural networks and processing within the brain. We often make the mistake of identifying two things that match only partially; thus, we identify color with frequency of light (a false identification according to the Land theory of color vision), and we identify pitch with frequency of sound (a fallacy familiar to any piano tuner). Why, then, should we identify perceived quality of sound with mathematical distortion of the sound's waveform?
mat (12/18/82)
I for one CAN hear TV raster, though I don't know if it is the 17+ kHz signal or some lower frequency that is being generated. In any case it is one of the most offensive sounds that I know. Yes, my terminal bothers me a little too! I can also hear SOME of the ultrasonic store alarms, and can 'feel' some others. This notwithstanding, I find most digital recordings just fine. Some of them seem a bit on the bright side after listening to 10 or 15 year old recordings ... but is that the fault of the digital process or of the older, less accurate recordings ? Every medium introduces SOME effects, either as a result of the medium itself or as a result of the paraphenalia that accompanies the medium. There are some people who feel that Avery Fischer Hall is too 'bright' and prefer Carnagie hall. There are some who find the visible delay in sound propagation in Carnagie Hall to be disconcerting and prefer the more 'accurate' Avery Fischer. Are recordings made in anechoic chambers with perfect mikes? What about reflections from the recording artists body, the mike stands, etc? And how many of us listen at the ideal locations for proper phasing relative to our speakers ... in an anechoic chamber. Yes, there ARE bad things that can happen to the HiFi signal ... but not all things that happen must be bad, and some of these can thoroughly mask many of the 'distortions' that get introduced. --NEXT PLEASE!
VaughanW@HI-MULTICS (12/25/82)
From: VaughanW at HI-MULTICS (Bill Vaughan) Though I can't hear the difference between .01 and .001 percent THD (or I don't think I can), I would be careful about saying people are hallucinating. There are several scenarios I can think of which would explain why one might be able to hear low levels of distortion through a filter having more distortion. 1) Suppose the coloration of the amplifier's distortion differs significantly from that of the speaker? e.g. the amplifier has more odd harmonic distortion at certain frequencies, the speaker has more even harmonic distortion at other frequencies, etc. Odd harmonic distortion is easier to hear than even HD (or so I've read somewhere). This now begins to sound like a classical problem in signal processing, i.e retrieving a weak signal (the distortion of the amplifier) from strong interference (the distortion of the speakers). But that's unlikely, because the amp. and speakers are likely to be used together a lot & you wouldn't be able to tell what coloration came from what box -- unless you were doing an A-B test in an audio showroom (the best way to buy a system anyway). 2) Maybe the reference was to intermodulation distortion rather than THD. I wouldn't be surprised to find that people can hear the difference between .01% and .001% IMD. IMD is particularly annoying - flutter on a turntable or tape machine is an example. 3) Don't forget that you can see colors through colored sunglasses. The brain is a far far better signal processor than any test equipment yet invented. The problem lies in telling just what it is processing and how. 4) We tend to measure the characteristics of audio equipment under steady-state conditions, though what we listen to through the equipment is far from steady-state. (Kind of like looking for a lost coin under a streetlamp on the wrong side of the street.) A box that has very good steady-state specifications may have very poor transient response. Maybe the .001% and .01% figures are both wrong under the actual conditions. ---(1)---
dya@unc-c.UUCP (11/27/83)
References: akpcb.9148 Actually, to preserve angular resolution between channels ( which is critical to one's perception of direction) then perhaps much more than simply " some marginal increment above 2 * the highest reproduced frequency." Those who claim that digital " sounds more sterile " than analog desparately miss their compression and peak limiting. What I would like to know is why everyone is willing to put up with far worse than mere analog recording when they turn on the FM tuner? A student at UNCC who is also works for WFAE-FM (campus snob station) reports that even their G/R meter on their Optimod (stereo generator) reads -20 dB; i.e. a signal which is 20 dB over 0 vU is still 0 vU. Would you tolerate a 20 dB reduction in dynamic range? And Optimod is one of the good stereo generators; i.e. its pumping and breathing is much less offensive than say a Thomson-CSF (CBS Laboratories) Volumax and Audimax processors. An NPR affiliate punching the G/R meter ?? Well, WROQ-FM, the local MTV music outlet, once broadcast some of the digital discs. The resulting signal was so bad, because their AGC / peak limiting couldn't cope with the dynamic range or "rms level slew rate." Some of our local FM's transmit from 1-4 dB of dynamic range ( by continuous RMS integration measurements) over a 40 dB input range. Same with TV sound (even on high audio fidelity videocassettes.) You guys should flame at your local FM for transmitting such hogwash. Even the most affluent record purchaser gets tired of their albums. By the way, most consumer amplifiers are not up to the task of reproducing digitally recorded music. In fact, most consumer amplifiers are not up to the task of reproducing plain old ordinary records. Who wants to buy a music source and software with 90 dB + actual dynamic range when for them, cranking up Journey to decent party volume causes the panel lights to blink wildly (indicating poor power supply regulation.) Whilst everyone is building quieter and more distortion-free amplification, why do you have to spend $$$$$$$ to get tolerable performance? Those who have the blinking lights amplifier are going to be mighty disappointed when they crank up digital records to decent party volume. I can't wait for digitally recorded music, either. If excessive dynamic range is a problem, I'll truck out my DBX compressor. As for being rid of analog evils ( record tics, pops, clicks, slip-sticking, surface noise, warps, the hassle of using the Discwasher and Zerostat, ; tape noise and dropouts,) a little phase distortion is worth it. After all, what everyone in this debate is forgetting is that the peak limiting that goes into cutting conventional records and duplicating tapes does far more harm to phase response than failure to sample at 3 or 4 times the Nyquist frequency. Even the nonlinearities in your phono cartridge's transient response cause far more phase rotation than any digital system. As for the subject of analog tape heads, yecch! Have you ever seen a 500 hz square wave reproduced adaquately on anything we can afford? Even an ITC 750 (which is the only tape deck I can get my hands on that also reproduces music reasonably well) can't do this too well. Not that this keeps me awake, but not having to consider "cartridge- preamp impedance interaction" and so on will be nice, too. As will the reduction of RFI-induced noise in my cartridge's coils, having to shell out $ 100 for a new cartridge periodically, and being highly irritated at the fact that even half-speeds click and pop.... Right now, the limiting factor in most homes is their furnace, not analog s/n (unless you use stereophones.) Just to be rid of the other evils is why digital audio is worth it. If I could just get my local FM stations to be as good as my worst records ???? ..Dave
rpw3@fortune.UUCP (12/16/83)
#R:unc-c:-117000:fortune:8600003:000:1995
fortune!rpw3 Dec 15 23:15:00 1983
"WATCH MR. WIZARD" COMES TO USENET (See "Experiment:", below)
-------------
Not to worry about phase distortion due to sampling rate. There is
an interesting thing about the human ear:
- At LOW frequencies (a few hundred Hertz and below), we use
phase information to locate things, while
- At HIGH frequencies (a few kiloHertz and up), we use
amplitude (loudness) to locate things.
In between, we use a mixture, as one might expect. The cross-over
between the two occurs at frequencies whose wavelengths are roughly
equal to the width of your head (surprise!), since making use of
phase information implies that the signal diffracts (not scatters)
around your head. (Sort of over-the-hairline sonar)
Experiment: You can prove this to yourself by putting a tone-generator
on to your amp + one speaker in a "soft" or "quiet" room. Set for a
single sine wave, and turn up the volume a bit. For very low ( <200 )
and very high ( >6000 ) freqs, you will think the sound is coming from
the speaker, but for certain middle frequencies, as you turn your head
(maybe eyes closed, to help), the sound will appear to "jump" around
the room. Try it. Fun.
[As an exercise for the reader, measure your head, divide by the
speed of sound, take the reciprocal, and listen to Andy Devine
saying "No, no, Froggie!"]
Don't use a "hard" room, or your head will wander in and out of
standing waves of the high frequencies, making you think phase is
important there too. It is, in a "hard" room, but only because the
standing waves actually modulate the amplitude at your ear.
So don't worry about your digital disks. The main problem with
sampling is aliasing harmonics back into lower frequencies "growls",
and (if they are AT ALL competent), the raw audio will have gone
through an anti-aliasing filter before being digitized.
Rob Warnock
UUCP: {sri-unix,amd70,hpda,harpo,ihnp4,allegra}!fortune!rpw3
DDD: (415)595-8444
USPS: Fortune Systems Corp, 101 Twin Dolphins Drive, Redwood City, CA 94065