Sam recently made a post explaining how higher sample rates (88.2kHz, 96kHz, 192kHz, etc.) are not better and arguably more problematic than lower sample rates. So if bigger isn’t better, where’s the sweet spot? The answer is 48k. Here’s why.
Your computer can’t handle it
The reality is that working at higher sample rates uses way more processing power and hard drive space than 48 or 44.1, and that has real implications for your workflow.
To illustrate this point: I once received files for a mix session from a tracking engineer who always worked at 96kHz. I figured I’d keep my mix sessions at that sample rate and see how it went - hey, maybe I’d hear whatever magic it was that he was hearing. Who knows?
Well, I don’t know that I heard any jump in quality, but I can tell you for certain what did happen: by the time I was running enough plugins to finish the mix, the session would crash. I made a copy at 48k and finished the mix with no issues. The client, of course, was furious…
Just kidding. Nobody noticed, if the client even cared in the first place. (He didn’t.)
The even-integer-multiple myth
Ok, fine - so maybe 96k is overkill. But shouldn’t we work at 88.2 if the final release is in 44.1? There’s a common belief out there that audio will downsample more cleanly if there’s a whole-number ratio between source and target (2:1 in this case).
This claim makes sense on the surface, but it’s mostly bunk. Modern sample rate conversion isn’t just your DAW throwing away every other sample when you go from 88.2 to 44.1. A good SRC basically reconstructs the waveform, filters out frequencies the new sample rate can’t represent, and creates a new set of samples at the target rate. That process can be done well or badly, but it doesn’t magically become clean just because the two sample rates are even multiples.
44.1 and the Nyquist theorem
What about just working at 44.1 then, and skipping the conversion altogether? There’s nothing inherently broken or bad about 44.1. But during recording, playback, and SRC, digital audio systems rely on filtering around the Nyquist limit - the highest frequency that sample rate can represent. Those filters have a little more runway between the upper edge of human hearing and the Nyquist limit when working at 48k - about 4kHz, vs. 2kHz at 44.1.
Is that a difference you can hear? Honestly, probably not. But if you care about “best practices,” this is an easy slam dunk, without the drawbacks of higher rates.
Industry standards matter
Ultimately, the real argument for 48k audio is the most boring of them all: it’s the standard audio format for video. Staying in 44.1 is fine for music, if you’re confident it will never end up getting used for any kind of sync. But why would you want that?
If your music gets used for video - movies, TV, streaming, games, ads, YouTube, social media - then somebody is going to have to do sample rate conversion at some point unless it’s already at 48k. So here’s my question: wouldn’t you rather that conversion be handled intentionally, by you, who knows what your music is supposed to sound like?
Look, maybe you hear some magic that happens at 88.2 and above that I can’t. But the bottom line is that 48k is the sample rate that comes with the most real-world benefits and the fewest drawbacks, and no fairy dust required to believe any of it.
