Sample Rates In Modern Record Making
By: Ruairi O’Flaherty of Nomograph Mastering and Unf*ck Your Studio
Edited by: Bradley Will
Editor’s note: This is an edited version of a post by Ruairi O’Flaherty on Oct 22, 2022 on the LWMR Discord Server intended as a starting point for beginners seeking guidance on the best sample rate for their needs. It has been posted here for posterity and ease of access. Enjoy.
A Tiny TL;DR
Which sample rate you use is not a deal breaker, or even close. You can make a great sounding record at any sample rate. There are edge cases where it can make a meaningful difference but in modern Pop, Rock and Hip Hop production it’s not that important.
Disclaimer - I’m going to play a little fast and loose with scientific terms here in an effort to make this useful for everyone.
What Matters? (Even a Little Bit)
Higher sample rates reduce aliasing
There is effectively no difference between 44.1k and 48k here but a meaningful difference at 96k and above. Aliasing is caused by non linear processes like distortion, and dynamics processing (compressor or limiter). Lots online about what it is and why it happens but you can sum it up by saying it’s an inharmonic, digital sounding hash that can feel bright but in a kind of ugly way (but remember sometimes ugly is good).
Higher sample rates reduce the impact of anti-imaging and anti-aliasing filters in AD to DA converters
Filtering is necessary for accurate capture and replay of analog signals and in the early days of digital it used to be a real challenge. Now it’s basically a solved problem. But if you are doing acoustic music with delicate transient information there is still a small but real advantage to higher rates.
Higher sample rates reduce overshoot and inter sample peaks on limiters
This really only matters if you are sending your loud mixes directly to release, otherwise your mastering girl or guy should handle that.
What Does Not Matter? (Even Though People Will Say It Does)
More samples (slices) per second equals a better representation of the analog waveform
Within half the band of the sample rate (say up to 22.05k for 44.1k) this is demonstrably not true. I know that is confounding, but digital and information theory is basically miraculous.
We can hear/feel/sense sounds above 20k
We truly really cannot and if you experiment with significant energy above 20k you’ll actually see problems creep in for electronics and speakers etc. Information above 20k is not our friend (space for harmonic information to live without aliasing back excepted).
Working at multiples of 44.1k is better because the sample rate conversion will work better
This used to be a little bit true a long time ago but hasn’t been for a long time. Use the SRC in iZotope RX on it’s stock preset to 44.1 and you are in pretty good shape.
Digital EQs work better at higher sample rates because of cramping
Cramping was an esoteric problem with older digital EQs that looked bad on analyzers but never actually sounded very bad. It’s a solved problem nowadays with most EQs.
Video and film is at 48k so we must start there
Yes, film and TV still work at 48k (although YouTube specs 44.1 now) but I have sent a million tracks for sync at 44.1 and it gets dropped in with SRC and away we go.
TL;DR
So what to do practically in the real world? If I had a super fast and stable computer I would record, produce, and mix at 96k. It’s subtly better in every way and you can introduce any of the “desirable” problems from lower sample rates at will. BUT BUT BUT ...if there was even the tiniest chance that the rig would hiccup, or my drives would not access fast enough, or that I would have a glitch in a real session then I would absolutely start at 48k.
TL:DR #2 - The Return
The final word. If I was asked “What sample rate should people make records at in 2022?” and was forced to choose then the answer would be 48k. Why?
Film and TV prefer it.
Labels prefer it.
Atmos is at 48k.
There are small but meaningful technical and sonic advantages.
It doesn’t place demands on CPU or drive space.
