kimr@sri-unix (07/16/82)
For you digital audio fans, I have a thought question. There best current information on "audible differences" (admittedly reported by golden-ears, but incorporated in products by manufacturers such as NAD and Cotter) is that the fastest signal rise time the ear is sensitive to is in the 9 to 14 microsecond range. A digital signal recorded at 50kHz has a maximum possible rise time of 20 microseconds. Admittedly, analog recordings have their problems, but rise time limitations are not among them. Why isn't our technology capable of supporting, say 100kHz sampling frequencies? My favorite solution for the digital disk (assuming the 100kHz a-to-d's are the limiting factor) is to record the program using analog technology and then "half speed master" the digital disk at 50kHz and play it back at 100kHz. I, for one, will never purchase a digital recording, disk, or disk player until the sampling rates are improved. Kim Rochat
jj (07/19/82)
I must admit that I find it difficult to reply to the article by Kim Rochat, although not because the arguements are difficult to refute. They are, in fact, not that difficult to refute, to another signal processing person. Disgegarding the opinion statements about Golden Ears, I will begin my comments by addressing the statements about rise time in analog equipment. It is indeed limited, and quite severly, (in a record, for example) after several passes through the playback device. (This statement holds for magnetic recording, also, but a discussion of the causes is beyond this note). In addition, records, which do generate signals with a very high rise time, most often do so (even in the case of the best records and cartridges) as a result of tracing distortion or mistracking. Even if the signal being reproduced does have an undistorted risetime on the order of 10us, the transducer that reproduces it and the travel through the ear will spread it out, due both to filtering and dispersion. There have been several, although few and far between, papers that have discussed the variability of DISTORTION with respect to signal risetime. Most transducers, do, in fact, have measurable non-linearities that arise during highly slewed parts of the signal. These non-linearities have been shown to be perceptable.<i.e. how does the test that you refer to avoid measuring THAT as a perceptable effect?> Regarding the recording of digital at half speed: There are indeed converters that are capable of converting at 100kHx. The converter is not the problem. STORAGE of the information is the problem. Your scheme of slowing down the digitization of the signal will do the following: 1) Make no difference whatsoever in the Digital sense<assuming the equipment is working correctly> 2) Introduce a distortion of about 2% <minimum> due to the nonlinearities in the magnetic recording process. 3) Limit the dynamic range to 60-65dB OR introduce low frequency distortion as a result of compression/expansion. 4) Introduce the most annoying (to me) problem with magnetic recording tape: Head Scrape/Flutter. This bizzare process is the result of sliding a tape, with a finite coeffecient of friction, against the tape head. It has nothing to do with the flutter due the drive system, etc. This is the process that causes the "closed in' sound that is often attributed to magnetic recording. 5) Double the storage required for the signal. To summarize. Your method of making digital recordings would eleminate the beneficial aspects of digital recording, without making any change in the problems involved.
thomas (07/19/82)
raeh My thanks to Jim Johnson for the comments on the 'rise time/sampling frequency' issue. I can flame for quite a while about people who think they can hear 50kHz (or even 100kHz), but I'll refrain. (I often wonder if all the AM radio signals pulsing through their ears drive them crazy.) I want to emphatically second the point that any analog recording stage in the process totally destroys the digital advantage (except its archival quality, more on this later). The Fleetwood Mac album "Tusk" (pre-released in platinum, as explained in net.records) was recorded analog (so they could play their 32-track mixdown games), then transferred to a digital form for editting and (I think) mastering. ON THE RECORD you could hear the tape hiss from the single analog stage. GROSS!!! Why then, you say, did they do the digital stage at all? The way I heard the story is that the analog MASTER tapes for "Rumours" (these are the originals, no way to recreate them) suffered during the process of producing the 10 million copies they sold that the high end was severely depressed in the later pressings! They wanted the 'never changes no matter how many times you play it or copy it' quality of the digital recording medium for this reason. To me, this is the overweening reason to go digital, even if there were no other differences. A copy of a digital recording is EXACTLY the same as the original, no added noise or distortion. Tom Stockham (a pioneer in digital recording technology) often puts it this way: "Digital recording is to analog recording as books are to oral tradition." Too much, I'll stop now. =Spencer Thomas
pdh (08/04/82)
I, too have heard of a number of good arguments against digital recording techniques using the 50Khz sampling rates. The best one I have been able to come up with so far is that, at that sampling rate, potential channel-to- channel phasing misalignment at much over 15KHz COULD be tremendous, possibly as high as 20 or 30 degrees. This phase misalignment would vary with frequency, due to the way the A/D's work, and so the final result is a piece of music with very vague and poorly defined imaging. I have noticed this in every digital album I have heard to date (about 30). The big deal with this is that the problem won't go inaudibly away until the sampling rate is up around 300KHz or more, which is WAY out of the range of technology. One possible solution to the problem would be to digitize the signal from the microphones onto several different digital master tapes, and then use a computer to correlate the signals, and assure proper phase alignment. This should work by statistically making the maximum phase alignment proportional to the number of master tapes being correlated into a single master. Peter Henry hplabs!pdh
jj (08/04/82)
The different articles posted to the net about digital audio demonstrate a very interesting phenominon, i.e. that of complaining about (I lack the vocabulary to use the proper word here) various alleged problems that exist in digital audio, without either a correct statement of the so-called problem or a reasonable comparison with the performance of the current analog systems. The complaints about the phase response of digital systems are just plain WRONG! I can't attack the argument on technical grounds because it doesn't have any basis in either theory or practice. There have been hardware problems that have, in the past, led to such problems, but I do not consider that a problem with even a set of the most well known hardware constitutes a flaw in the concept. The problem alluded to in the article that this responds to alludes to phase problems. The correspondant does not mention whether this is a fixed, random, or time varying phase problem. Neither is the magnitude of the "phase problem" compared with the phase response of a tape deck such as that used to master a disc. To clear things up a bit.. If the "phase problem" is random, it will cause the system to have a total snr of about 10dB, regardless of the number of bits, etc. Since this is clearly not the case, I dispose of this case. If the phase is time varying (at a fast rate--) it will cause modulation effects that have similar results (no snr). If the phase varies very slowly, it has the same effect as a tape deck. If the phase is non-tracking, but fixed, it doesn't matter, since it's like a perfectly aligned tape deck with NO head skew.(Fat chance) I am becoming tired of defending digital audio systems against people who have complaints which are either wrong or taken out of context. Is there someone else out there with an interest in digital audio? If so, and your are interested in continuing the debate, mail me. If no, I will abandon this group to the uninformed digit haters who seem to think that any new technology should be perfect,and who disregard the performance of current systems in their haste to tear down a new idea. I guess that new ideas will always suffer from conservatives who fear that they will become obsolete. (Very likely, if they refuse to become informed.)