As we race towards higher and higher sample rates (especially in the mastering community), it’s easy to forget this one simple fact: Sample rates higher than 96KHz serve little purpose other than to make equipment manufacturers more money.
Don’t get me wrong; moving from 16-bit to 24-bit was a major advance forward in sound production and reproduction. Increasing sample rates to 24-bit gave us noise floors that were below the self-noise that amplifiers typically produce, making the hiss that everyone accepted as a part of recording disappear, while simultaneously removing one of the critical concerns out of the audio equation; constantly worrying about levels. I can’t count the hours spent staring at VU meters on analog machines, trying to get that perfect maximum level before distortion, and with 24-bit, that’s barely a concern.
16-bit gave us a maximum signal-to-noise ratio of just -96db, and 24-bit gave us a maximum signal-to-noise ratio of -144db, freeing engineers to record as low as -48db before having reaching the noise floor of 16-bit recording! So, the main point I keep trying to get across in these articles, especially those who record in home studios and bring us their masters is this: Never let your master stereo bus peak above -3db. For me, the ideal level for home recordings are peaks at -6db, ensuring that there is plenty of headroom as well as plenty of distance from the noise floor, allowing the mastering engineer to deal with raising the levels to near-zero levels, using equipment specifically designed for these kinds of tasks.
This leaves us with deciding what the ideal sample rate for audio material is. But first; a very brief discussion on what sample rates do for sound: Theoretically, reconstructing an exact duplicate of a signal (in the digital realm) is possible only when the sampling frequency is more than twice the highest frequency of the signal being sampled or when the Nyquist frequency (half the sample rate) exceeds the highest frequency of the signal being sampled.
For example, if a signal has an upper band limit of 20KHz, a sampling frequency greater than 40KHz will avoid aliasing, theoretically allowing perfect reconstruction of the original signal. This is exactly why we use 44.1KHz as a standard for the reproduction of digital music on CD, and why 48KHz is used in professional applications and equipment.
Also, since humans theoretically can’t hear above 20KHz (most people can’t hear much past 15KHz), there seems to be no need for any sampling rates higher than 24-bit and 44.1KHz, is there?
And this is where the debate comes in: After conducting multiple listening tests using 16-bit/44.1KHz files, 24-bit/44.1KHz files, 24-bit/96KHz files, 24-bit/192KHz files, one thing became glaringly apparent: There is a “night and day” difference between a 16-bit master and a 24-bit master, but the difference between 24-bit and 44.1KHz or 96KHz or 192KHz becomes far less audible, if audible at all, when reproduced through audiophile-quality equipment. And for whatever reason, the clear loser in sound quality in the 24-bit testing, was the 192KHz sample rate.
So, what does that mean in the real world? Well, as one of the most highly-respected sets of ears in the industry; Bob Katz states; “A well-designed DAC should exhibit very little audible difference between sample rates.” In fact, after a series of stringent tests he did with colleagues, no one could detect any audible differences at all between different sample rates and converters on 24-bit files when using professional gear such as the Benchmark DAC1 that we rely on for all our conversions back to the analog world we exist in.
Another detail about higher sample rates is this: With a sample rate of 44.1KHz, to get a signal with a frequency range of 20Hz-20KHz, an extremely steep filter/slope is required to get to the Nyquist frequency. In other words; if the slope starts at 20K, it needs to be at -120db by the time it gets to 22.02KHz (half of 44.1KHz).
This only allows for a 2K-wide filter, making for a very steep slope. Yes, it’s a slope in the inaudible range, but less than perfect filters (such as consumer equipment) can cause “ripples” throughout the entire frequency range, manifesting as distortion, ringing, blurred sound, and other unpleasant audio artifacts.
With 96KHz, a signal needs to get to -120db by the time it reaches 48KHz instead of only 22.05KHz, meaning the filter/slope can be much gentler. This, although it’s not been able to clearly be determined in various listening tests, could be the reason why 96KHz can sometimes sound slightly better than 44.1KHz; a gentler slope has the potential to cause less “ripple” in the audible band. And this point is exactly the reason why we typically master at 24-bit/96KHz at the Playground.
Keep in mind, though; all of these discussions are relevant only to the specific gear used at mastering houses by mastering engineers, since the only way most consumers (At least at the time of this writing) hear their music is at the consumer standard of 16-bit and 44.1KHz at best. How it got there is only relevant to the people getting it there, and not to the actual consumer who has the final product in the form of a song.
And, with a home studio setup running off of your laptop, it might actually be a better idea sonically to record at 24-bit/44.1KHz instead of 24-bit/96KHz because it’s far easier on resources, most filters in home studio gear are optimized for the lower sample rate, and overall, you may actually get less distortion, smearing, ringing, and all the other negatives that can come with less-than-professional quality ADC’s and DAC’s.
Furthermore, tests have shown that 192KHz may actually be “too much” of a good thing when it comes to audio. To justify higher sample rates, many argue that the more you have to work with, the more room you have for error. But this is an argument based only in theory and not in actual fact. Not only do such high sample rates require an immense amount of processing power; what it’s actually processing has been found to have no audible impact on the final sound.
But the difference is so subtle to completely inaudible, what truly is the point?
And this is exactly in keeping with the main point I’m trying to convey in relation to mastering audio at The Playground: It’s easy to get caught up in the technology and the numbers, but the bottom line is whether or not you’re completely confident that you are walking out of the mastering house with something that not only sounds better than it did when you walked in, but with it’s sonic potential maximized in every way possible.
So, are you going to miss out on a clearer, more transparent-sounding mix if you give us 24-bit/44.1KHz masters instead of 24-bit/96KHz masters? Not at all! Half of 44.1KHz is 22.05KHz, which is already 2.05KHz above the range of human hearing. Taking into account that speakers only reproduce sounds up to 20KHz and only in a perfect environment (most speakers and headphones only reproduce up to 15KHz or 18KHz at best) and coupling that with music converted to MP3, and suddenly recording at 24-bit/192KHz seems a little silly.