Sample rate, as you may have guessed, is the number of samples per second in a piece of digital audio. We measure this in Hz (or kHz) and it is used to determine the frequency range that you are able to record audio. This is also known as Nyquist’s Sampling Theorem. As an example, a sample rate of 10kHz (10,000Hz) would be able to reproduce audible frequencies up to 5kHz with anti-aliasing.
Nyquist Theory in More Detail
The reason for us needing double the sample rate of the highest frequency we want to hear is rooted in the concept of the audio wave. For every positive section of a wave, we have the corresponding negative section. By having twice as many sampling points per cycle we need to reproduce, we can create a far more accurate representation when converting from analogue to digital. Taking this concept further, by using an even higher sampling rate, we can record more points in each cycle. This allows for a greater level of accuracy when converting from analogue to digital.
Red Book Standard, Nyquist Converters, and Folding Frequencies
For a long time, the industry standard for red book format (commercial CD release) has been a sample rate of 44.1kHz. Based on Nyquist theory, this allows us to reproduce frequencies up to 22,050Hz. Perfectly adequate for music, right? Whilst this is true, the argument for higher sampling rates has become more prevalent in recent years due to advances in technology. Equally, many streaming services are now allowing 48kHz as a sample rate, including iTunes.
Many argue that a sample rate of 96kHz offers a more complete bandwidth. It is less likely to result in aliasing issues when recording very harsh, high-end sounds like drum cymbals, brass, and white noise. This certainly holds true. However, another stipulation of Nyquist converters is that any frequencies that surpass the sample rate threshold will be folded back into themselves due to sum and difference. This means harmonic content from bright instruments recorded at a 44.1kHz sample rate could potentially fold back, creating high-frequency distortion.
As an example, let’s say that you record a very bright string section. Potential harmonic overtones could sit around 25kHz. 25kHz recorded at a red book sampling rate of 44.1kHz would result in aliased fold-back, creating tones at 19.1kHz. This is something our ears are not only perfectly capable of perceiving but something that will mostly sound like distortion. Another thing to bear in mind is that the closer to 0dBFS you are, the more apparent this will become. This is where setting appropriate recording levels is imperative.
CPU Power & HDD Space
One of the most important things to consider is what your computer can handle. If we double our sample rate, we will double our file sizes. Not only this, but your CPU will have to work twice as hard to read from a larger file. This means you won’t be able to handle as many sources or plugins as usual if you switch up to a higher sample rate. Equally, if you start using higher bit-depths as well, your files are going to get even bigger. If your rig can handle it then go for it but do be wary of system degradation.
192kHz Sample Rate!?
Yes, this does exist. A sample rate with the ability to reproduce frequencies up to 96,000Hz. Why? That’s the million-dollar question. Even within super high-fidelity recording and limitless computer power, I see no realistic situation where this is necessary. If you’re reading this and you think there is, please do drop me a comment and I’d love to discuss it.
48kHz Sample Rate is the One for Me
My suggestion to any of you reading this, stick with a 48kHz sample rate. The likelihood is that you are working with music. Therefore, you’re unlikely to need to deal with any super high-frequency, out of this world type audio. Maybe if you’re doing some creative sound design, you might find some use for higher sample rates. However, it’s still unlikely. Giving yourself that little bit of extra top end over a 44.1kHz sample rate will help to deal with the harmonic fold back that we mentioned earlier. It will also ensure minimal conversion issues and stop possible clipping over your LUFS/dBTP thresholds. It will also result in sensible CPU usage, manageable file sizes and a hardly noticeable increase in RAM load over 44.1kHz.
However, one thing to bear in mind here is what format you’re ultimately going to end up with. If you’re looking to put your releases out on CD, starting with and sticking to a sample rate of 44.1kHz makes far more sense. This will minimise sample decimation and provide better quality over a downsampled file. Equally, if you intend to render your files to 48kHz but want to work with higher sample rates, using a sample rate that is a multiple of that integer (96kHz, 192kHz) is a far simpler equation when downsampling. This will result in higher quality.
Sample rates come in many shapes and sizes. The thing you really need to know is that, based on the science of human hearing, you needn’t bother with anything over 48kHz. Nyquist theory tells us that this is sufficient, and will be handled well by our ADCs (Analogue to Digital Converters). You won’t have to deal with ridiculous file sizes. Your PC won’t die under the stress and you can continue working as you always have been.
For those of you who are interested in the more complex science behind the reasons why super high sample rates aren’t useful, please check out this fantastic article by Monty at Xiph.