The famous thinker, Edward De Bono, believes that most of life’s errors are errors of perception – not errors of logic. Read on and decide for yourself if this is case in the 96kHz debate.
Is it really worth the expense and effort required to properly configure a 48kHz/96kHz system? A few die-hard engineers are sticking to their 44.1kHz guns but the apparent advantages of higher sample-rates when used throughout the audio production process are worth further investigation.
A widely accepted belief lingers in the minds of audio engineers that even if you are using equipment throughout your audio chain that is able to process signals of up to 44.1kHz it is a waste of time because human hearing is limited to 20kHz anyway, and the Nyquist theorem demonstrates that 44.1kHz sample-rates certainly do deliver all that’s necessary – if that belief is true. If that’s true and that’s the only relevant consideration, then that’s the end of the discussion because the logic holds – what could possibly be the point of trying to capture and reproduce signals above 20kHz if we can’t hear (perceive) them? If you stop there the logic will lead you to the conclusion that all this fuss about high sample-rates is just a new cycle of revenue-generating marketing deceptions from big audio companies. This might just be an example of the kind of error of perception Mr De Bono was talking about. Here’s what the tech-heads are saying…
Human auditory perception spans well above 20kHz
A paper published as a result of research performed at the Human Information Science Laboratories in Japan by Tsutomu Oohashi* demonstrated that humans accurately report the presence or absence of signals well above 20kHz. In his experiments, Oohashi and his colleagues found that subjects showed that they could perceive changes in the quality of material played back with and without the information above 20kHz. Despite the subjects not being told when the signal above 20kHz was present or absent the subjects consistently reported that the music sounded better when the signal above 20kHz was present (*see references at the end of this article).
Another fascinating study by James Boyk** of the California Institute of Technology showed that a cymbal strike for example, has frequency components approaching 100kHz with 40% of it’s sound energy above 20kHz. Many other instruments also exhibit strong harmonic energy well over 20kHz. Clearly there is information well above 20kHz to record and it seems we can in fact hear them when listening to the actual instruments themselves but buyers of recorded music will not benefit in this particular way from owning recordings made at higher sampling rates until their systems are capable of faithfully reproducing those frequencies above 20khz, in theory.
The fact that so much sound exists above 20kHz might partly account for common engineering frustrations with the apparent inability of present recording techniques and technology to produce a recording that sounds exactly like the ‘real thing’. Since we are, it seems, able to perceive those frequencies over 20kHz it provides a possible explanation as to why we are disappointed when we hear cymbals played back over our studio monitors which might (if you’ve spent your money well) reproduce frequencies up to 22kHz.
So, when viewed from a cost-to-benefit perspective, since most consumer equipment could not play back those frequencies above 20kHz even if they were present, there seems to be little justification for the investment required to implement a 96kHz system. However, the most important benefit of using a higher sampling rate has yet to be mentioned.
The improved audio characteristics of 96kHz sampling rates are audible below 20kHz***
Digital equipment is designed with filters that are (generally speaking) non-linear at the upper bands of the frequencies it is designed to reproduce. This means that distortion (albeit subtle) is present in, or at least very close to, those high frequencies processed through your average 44.1kHz studio equipment. When consumers use their home or car systems to play back music recorded on this studio equipment they will reproduce on their systems the cumulative distortions put there by the studio equipment on every recorded channel in the mix (and possibly add some distortions of their own).
Dan Lavry’s position on this seems to be that there are ways of getting around these distortions introduced by studio equipment running at 44.1kHz but those solutions boil down to expensive implementations mathematical algorithms and excwellent circuit design – the cost of which is passed on to the consumer. On the other hand the implementation of higher sampling rates affords manufacturers a more economical way of moving the distortions higher up or even beyond the limit of the audio spectrum that consumer sound systems are currently able to reproduce. So, whereas the average consumer’s sound system is still only reproducing frequencies up to around 18kHz, if the music was carefully recorded, mixed and mastered at higher sampling rates the part of the spectrum that consumers hear will be largely devoid of the cumulative distortions that are introduced to the music through multiple instances of equipment-related non-linearities.
What this means is that audio professionals are able to provide improved quality of audio signal to the consumer through the use of high sample rates throughout the recording process. There is a caveat worth mentioning though, which is that economically implemented 96kHz sampling might sound different to pristine implementation of 44.1kHz sampling, but that doesn’t mean it’s better from a purely technical perspective. So when comparing the performance of the systems a distinction needs to be made between a pure-science (measurement) comparison of the systems in terms of the fulfillment of their technical purpose, and the audible character each possesses as a result of how they were implemented (circuitry etc). You might like 96kHz more but it’s not necessarily because it does it’s job better.
The painful fact remains that the world’s consumer car and home-stereo systems will take a while to catch up enough for consumers to fully benefit from pro-audio technological progress, but that seems like a poor reason not to embrace it. Or is it?
References and Essential reading
* For more information on the research Mr Oohashi has done into the “supersonic effect” please visit: http://jn.physiology.org/cgi/content/full/83/6/3548#SEC4
** For more information on the research James Boyk has done into the limits of human auditory perception visit: http://www.cco.caltech.edu/~boyk/spectra/spectra.htm
*** For a more thorough treatment of the topic of sample rates, read the thoughts of Dan Lavry of Lavry engineering (makers of some of the worlds best AD/DA converters) who has much to say on this controversial and heavily technical topic. His thoughts, from which much of the contents of this article are derived, can be found at the ‘support’ pages on www.lavryengineering.com in a tutorial called “sampling theory”. It is well worth the read.