Human hearing is generally defined as extending from 20Hz to 20kHz, and the Nyquist criteria states that a sampling rate of at least double the highest audio frequency must be used in order to accurately to represent the signal. An unwanted side-effect of analogue to digital conversion is 'aliasing', where musically unrelated signal noise enters the audio band, so an 'anti-aliasing' filter is used to prevent this, but because it is not possible to implement an infinitely steep filter, a slightly higher sample rate is used than is theoretically necessary. This is how the CD standard of 44.1kHz came about (a little more than double 20kHz), and DAT took this up to 48kHz to allow varispeed recording.
So if we can't hear above 20kHz, why record at very high sample rates such as 96kHz or even 192kHz? Some have suggested that although we don't actually hear 'supersonic' frequencies, we somehow perceive them by another means, but experiments that artificially add and remove these frequencies have shown this not to be the case. But 'golden eared' tests which have made A/B comparisons of the same equipment operating at standard and high sample rates have revealed that 96kHz & 192kHz recordings do sound subjectively better - so what's going on?
The answer lies primarily with the anti-aliasing filters - even the best designed filters produce a certain amount of 'ripple' that extends downwards from the cut-off point and gradually tails off, and this is perceived as a 'smearing' of high frequencies and the stereo image. But by moving the cutoff point way up (say to 96kHz), the ripple has all but disappeared by the time it gets down to audio frequencies, and hence, substantially improved sound quality. Importantly though, it follows that this improvement is maintained even after the sample rate has been converted down to normal rates, as no new A/D conversion is required, and therefore, no anti-aliasing filter.
Experienced Recording Studio Engineer GuidelinesIn the early days of high sample rate recording, there was a preference for using multiples of the 'CD' frequency (44.1kHz), i.e. 88.2kHz or 176.4Hz - this was because sample rate conversion was a very difficult thing to do at the time, and better quality could be achieved when converting down to 44.1kHz by using an integer multiple. However this is no longer the case, and our usual preferred rate for acoustic recording is 96kHz for a variety of reasons:
- 96kHz is the native frequency of many DVD audio releases
- Although 192kHz does sound a little better than 96kHz, the improvement is marginal
- Some plug-ins do not handle 192kHz audio
- Processing power is reduced to roughly a quarter of that available at 48kHz
Rupert Pfaff has worked in the music industry for over 20 years having operated a well known music shop in London before becoming a sound engineer and technical authority for mixing and mastering. Now involved with the day to day running and managing the online bookings for RecordingStudioLondon.