Literature
"Compared to acoustics, religion is a very exact science." (Anon.)
Articles by the author
Background Knowledge
- Bob Katz, Mastering Audio, Focal Press 2002, ISBN 0-240-80545-3, www.digido.com
- Dave Moulton, Golden Ears, training CDs for sound engineers, KIQ Productions 1995
- Jim Taylor, DVD Demystified, 3rd Edition, McGraw-Hill 2006
- Dan Lavry, Sample Theory White Paper et al., www.lavryengineering.com
- Free Online Hearing Test
Annotations
^[0] Myths and Legends...
Audio technology is a little fussy and can work for or against the music. It doesn't make it easier that the understanding of analog and digital technology is very much obscured by a legion of myths and legends. Without studying this technology in good detail and applying it in actual recording situations, it is impossible to make the right decisions, both technically and musically. — "It's the ear, not the gear!"
^[1] The Sound of Digital
The sound of digital audio has often been described as "cold" or "sterile." One reason for this may be the absence of—potentially pleasant sounding—distortions. We were used to the presence of these distortions, such as added harmonics or filter effects, from all analog devices like tube amplifiers, record players, and so on. Not every digital signal chain will sound clear and transparent by itself, there are many possibilities to produce an unsatisfying result, both technically and musically. Moreover, digital distortions sound a lot less pleasant than their analog counterparts, many of which have been deliberately created or simply accepted as a given part of the sonic result. This is what the "vintage" hype is all about, and it is up to the discerning mind to find out how the typical analog distortions actually work and come into play.
It has taken digital audio technology a long way to overcome its teething troubles, and only very recently did we reach a state where it's safe to say that the results sound truly excellent (I guess, however, that this has been said many times before, so I'm reluctant to speak of perfection). The truth is that good digital converters are not cheap, and most of all they need to be fed with an undistorted, pristine analog signal to begin with. When we get all these things right, the feeling of being present at the same place where the music was created, as well as the impression that all recording technology simply vanishes from the perception, are clear indicators of a good recording involving good components.
^[2] Why 24 Bit/96 kHz? Why so much—or so little?
24 bit recording offers a much higher vertical resolution compared to 16 bit, i. e. the audio waveforms are digitised up to 2^8 = 256 times finer compared to the CD standard. This is the main difference between modern audio formats and older digital technologies.
Higher sample rates, which digitise frequencies far above 20 kHz in the front end, are more difficult to evaluate. Humans hear very little or nothing above 20 kHz. Most humans over the age of 30 have a ceiling of about 15-17 kHz, or a little higher if they are lucky and haven't been to too many rock concerts. Nevertheless, we are only talking about the last half octave—below that, there are nine other audible octaves in which a lot of music is happening. Thus, we should concentrate on those.
So the reproduction of ultrasonics cannot be the reason for sampling audio at 96 kHz or 192 kHz, although many products exist that claim to do just that. It hard to tell apart technological superiority from sales talk with a microphone such as the Sennheiser MKH 800, which records frequencies of up to 50 kHz, even if the subsequent elements of the audio chain are in no condition to process or reproduce such frequency content, most of all the human sense of hearing. Who is going to sit in with you in your mastering session to detect distortions above 20 kHz—your cat or your dolphin? ;-)
Moreover, higher sample rates do not increase precision. Everything our ears can hear is conveniently covered by a sample rate of 44.1 kHz.
Roughly speaking, there are two reasons left that make a data rate of 24 Bit / 96 kHz desirable;
- A sampling rate of 64, 88.2, or 96 kHz has more tolerance for possible filtering side effects and rounding errors that can occur during digital processing.
- The latency, or response time, is minutely better at higher sample rates when working on computer systems that are fast enough to cope with the considerable data overhead or 2x or even 4x the lower rate. Needless to say, this only has an effect if the same high rate is used on both ends of the system.
Balancing the pros and cons, we end up with a sweet spot between 48 and 96 kHz, where all possible unwanted artifacts and distortions are conveniently moved to a frequency range above the human hearing limit. Going higher than 96 kHz, it becomes increasingly difficult to trade off the technical advantages (shorter latency) against the disadvantages (decreased precision, foldback artifacts, superfluous data). Moreover, higher sample rates increase the data size and processor load proportionally, but the audible content remains exactly the same—on a good day. Whatever some listeners may describe as "better sound" mostly dissipates into thin air when looking closely or doing a blind-folded A/B/X test; often, devices operating at 192 kHz sample rate are becoming less precise and introduce distortion that some perceive as being interesting or pleasant, even though the signal is effectively diminished in quality.
Conclusion: The most noteable difference lies between the traditional CD standard with its 16 bit/44,1 kHz format, and the higher rates using 24 bit at 44,1 kHz and up. The audible differences when using higher sample rates are very minute and also depend on other factors such as filter design or clock jitter.
At the end of the day, the bandwidth of the human ear is pretty constant (When was your last hearing test (another hearing test here)?). All that matters is the quality of what the D/A or A/D converters recreate. Here, the differences are significant and very distinguishable, but almost completely independent of higher sample rates.
In my work, sample rates between 48 kHz and 96 kHz have proven the most useful, together with top-end traditional audio gear, ensuring a near-linear signal chain. It makes no sense trying to cut corners here. Apart from that, I try to achieve a minimum of disturbances in front of the microphones. :)
|
|