Sample Rate and Bit Depth: What Actually Matters

Two Settings, Ten Years of Internet Arguments

Every DAW (digital audio workstation — the software you record into) makes you pick a sample rate and a bit depth before you've played a note. Those two dropdown menus have generated more forum warfare than almost anything else in home recording, which is strange, because the practical answer fits in one sentence: record at 44.1 or 48 kHz, 24-bit, and get back to making music.

That's the conclusion. The rest of this article is the reasoning, because understanding why those settings are enough will save you from two expensive mistakes: buying gear to chase numbers you can't hear, and blaming your settings for problems that are actually about technique.

What These Settings Actually Control

Digital audio works by measuring an electrical signal over and over, very fast, and storing each measurement as a number. Sample rate and bit depth describe the two dimensions of that process.

Sample rate: how often you measure

Sample rate is the number of measurements ("samples") taken per second, in kilohertz (kHz — thousands of times per second). At 44.1 kHz, your interface measures the incoming signal 44,100 times every second.

The math that governs this is the Nyquist theorem, and it's worth knowing because it kills most of the marketing around high sample rates: to capture a frequency accurately, you need a sample rate a little more than twice that frequency. The highest frequency a system can represent — half the sample rate — is called the Nyquist frequency.

Human hearing tops out around 20 kHz in young, healthy ears, and most adults hear noticeably less than that. A 44.1 kHz sample rate captures everything up to about 22 kHz. In other words, the lowest standard sample rate already covers the entire range of human hearing with room to spare.

A common misconception is that higher sample rates capture a "smoother" waveform, as if digital audio were a connect-the-dots drawing that gets less jagged with more dots. That's not how reconstruction works. Within the frequency limit set by Nyquist, the original waveform is rebuilt exactly — not approximately. More samples per second don't make 10 kHz sound better; they only extend the ceiling to frequencies you can't hear anyway.

Bit depth: how precisely you measure

Bit depth is the resolution of each individual measurement. More bits means each sample can land on a finer grid of possible values, and the practical consequence is dynamic range — the distance between the loudest signal a system can hold and its noise floor (the low-level hiss under everything).

The rule of thumb: each bit adds roughly 6 dB of dynamic range.

• 16-bit gives you about 96 dB — the CD standard, and genuinely plenty for finished music.
• 24-bit gives you about 144 dB in theory. Real converters can't actually deliver that — the analog electronics around them generate thermal noise before you get there — but even mid-priced interfaces manage well over 100 dB of usable range.

Here's the key reframe: bit depth is not about "quality" in the sense of clarity or detail. It's about how far down the noise floor sits. A 24-bit recording of a bad performance through a badly placed mic sounds exactly as bad as a 16-bit one, just with quieter hiss underneath.

What Sample Rate Should You Record At?

44.1 vs 48 kHz: the only comparison that matters day to day

These two rates exist for historical reasons, not sonic ones. 44.1 kHz was the compact disc standard; 48 kHz grew up in video and broadcast and remains the standard everywhere audio meets picture.

Sonically, the difference between them is nothing you will ever hear — both cover the full audible range. The decision is about workflow:

• Working with video in any form — YouTube content, film scoring, podcasts destined for video platforms, game audio? Use 48 kHz. It's the delivery standard, and staying in it end-to-end avoids conversion.
• Music only? Either is fine. 44.1 kHz matches CD and is what most streaming services historically received; 48 kHz is equally accepted everywhere now.
• Collaborating? Match whatever the project already uses. A session recorded at mixed rates is an avoidable headache.

If you want a single default and never want to think about it again: 48 kHz. It's compatible with everything and video-safe. But if your DAW project is already at 44.1, there is no reason to convert — modern sample rate conversion is audibly transparent when done properly, but the best conversion is the one you never have to do.

88.2, 96, and 192 kHz: where the argument actually lives

This is one of the few areas in audio where the honest answer is "the evidence is genuinely mixed" — so here is both sides, fairly.

The case for higher rates. Some engineers prefer to work at 88.2 or 96 kHz, and the least hand-wavy argument involves aliasing: when a plugin generates new frequencies through distortion or saturation (any "nonlinear" processing), harmonics that land above the Nyquist frequency fold back down into the audible range as inharmonic junk. A higher project rate pushes that ceiling up, so there's less folding. There are also arguments about gentler anti-aliasing filter behavior in converters at higher rates.

The case against. Controlled, blind listening tests have consistently struggled to show that listeners can distinguish properly made 44.1/48 kHz audio from higher-rate versions of the same material. Meanwhile the costs are concrete: 96 kHz doubles your file sizes, doubles the samples your CPU must process for every plugin, and cuts your available track and plugin count before the system chokes. And the aliasing argument has weakened over time, because most modern saturation and clipping plugins oversample internally — they run their own math at a multiplied rate and filter before coming back down, solving the problem where it occurs instead of taxing the whole session.

One caution about testing this yourself: louder always sounds better, so any casual A/B comparison where levels aren't precisely matched will "prove" whatever you played louder. And keep comparison levels moderate — sustained listening above roughly 85 dB SPL risks permanent hearing damage, and your ears are the one piece of equipment you can't replace.

Our position: if your machine is powerful, your material is acoustic and sparse, and working at 96 kHz makes you feel better, it does no harm. But nobody has ever heard a home recording and said "shame it was tracked at 48k." The gap between amateur and professional recordings is microphone placement, room sound, performance, and gain structure — never the sample rate menu.

16-Bit vs 24-Bit: This One Has a Clear Answer

Unlike the sample rate debate, the bit depth question has a settled, near-universal answer: record at 24-bit. Here's the reasoning, because it's genuinely useful to understand.

When you record, you have to leave headroom — space between your loudest peaks and the digital ceiling (0 dBFS, the maximum level a digital system can represent), because an unexpected loud moment that hits the ceiling clips, and digital clipping is ugly and unfixable. Sensible practice is to aim peaks somewhere around -18 to -10 dBFS, which is the core of good gain staging.

At 16-bit, that safety margin eats meaningfully into your 96 dB of range: every dB of headroom you leave is a dB of range you're not using, and quiet passages sit that much closer to the noise floor. At 24-bit, the noise floor is so far down that you can track with generous, relaxed headroom and lose nothing audible. That's the whole trade: 24-bit costs you 50% more disk space — trivial today — and buys you the freedom to record conservatively without penalty.

16-bit still has a job: it's a perfectly good delivery format. CDs are 16-bit, and 16-bit files are transparent for finished, mastered music. Record and mix at 24; deliver at whatever the destination asks for.

32-Bit Float: Useful Tool, Misunderstood Spec

You'll see "32-bit float" in two very different places, and conflating them causes confusion.

Inside your DAW, mixing math has been done in 32- or 64-bit floating point for decades. Floating point can represent values far above 0 dBFS internally, which is why a channel can go "into the red" inside the mix engine without actually clipping, as long as the final output is brought back under the ceiling. You don't choose this; it's how the software works.

As a recording format, 32-bit float matters mainly in a specific gear category: field recorders with dual analog-to-digital converters, which capture the same signal at two gain levels at once and stitch the results into a file that effectively cannot clip at the file level. For run-and-gun location sound with unpredictable levels, that's genuinely valuable. But note the caveat: the analog input stage can still overload. 32-bit float removes one clipping point, not all of them.

Recording 32-bit float files from a normal audio interface adds nothing, because the converter inside is a 24-bit (fixed-point) device — the extra bits store no additional information about the signal. It doesn't hurt anything except disk space, but it isn't an upgrade.

Dither: The One Rule at Export

When you reduce bit depth — typically exporting a 24-bit mix to a 16-bit file — the discarded low-level information becomes quantization distortion: a gritty artifact correlated with the signal, most audible on quiet material like reverb tails and fades. Dither fixes this: a whisper of specially shaped noise added during the reduction that converts that distortion into benign, steady hiss far below anything you'll notice.

The rules are short:

1. Dither once, as the very last processing step, only when actually reducing bit depth.
2. Exporting at 24-bit or 32-bit float? No dither needed.
3. Never dither an intermediate file you'll process further — that belongs at the mastering stage, alongside loudness decisions.

Any dither type your DAW offers is fine. Agonizing over dither flavors is audiophile theater; forgetting dither entirely on a 16-bit export is a real, if subtle, error.

Settings Cheat Sheet

Two clarifications that trip people up:

• Sample rate is not buffer size. The setting that controls latency (the delay between playing and hearing yourself) is the audio buffer, measured in samples, and it's adjusted separately. Raising the sample rate does shorten latency slightly — same buffer, faster clock — but at real CPU cost; lowering the buffer is the honest fix.
• Changing a project's sample rate mid-stream is messy. Audio recorded at one rate and played at another comes out at the wrong speed and pitch, so DAWs must convert files when rates change. It works, but it's a step that exists only to be gotten wrong. Set the rate when you create the project and leave it alone.

Where the Sound Actually Comes From

It's worth ending on the anti-climax, because it's the useful part. Sample rate and bit depth are solved problems. The formats standardized decades ago already exceed human hearing and the dynamic range of any room you'll record in. Once you've set 48 kHz / 24-bit, every audible improvement left on the table lives elsewhere: the instrument, the performance, the mic position, the room, and the levels you record at.

Takeaway: record at 44.1 or 48 kHz and 24-bit, dither once when exporting to 16-bit, and spend the mental energy you just saved on mic placement.