Digital Audio Quality
quality depends on your settings going into and out of your DAW.
#file-qualityQUALITY of an audio file =
1 SAMPLE RATE (kilohertz/kHz) — how many snapshots per second
2 BIT DEPTH (bits) — how much info in each snapshot
3 FILE TYPE (wav · flac · aac · mp3) — what it’s saved into
mix_v3_48k_24bit.wav
① sample rate
② bit depth
③ file type
1 SAMPLE RATE (kHz)
44.1 kHz — CD quality
48 kHz — standard recording and file quality: video, podcast, streaming
96+ kHz — studio max / ultra quality
48 kHz captures all sound audible to humans (20 Hz–20 kHz).
96+ holds extra info — for stretching & pitching, max-fidelity room capture.
2 BIT DEPTH (bits)
16-bit — only if a spec demands it (CD)
24-bit — most record AND bounce here
32-bit float — records peaks past 0 as a safety net (the converter can still clip); larger files
more bits = finer volume steps and a lower noise floor.
3 FILE TYPE & BITRATE (kbps — kilobits per second)
WAV / AIFF — UNCOMPRESSED: they keep every bit
FLAC / ALAC — LOSSLESS COMPRESSION: shrink the file, lose nothing
MP3 / AAC — LOSSY COMPRESSION: discard some detail; their number is BITRATE (kbps).
256/320 is the minimum quality for sharing, archiving & DJing — 128 will be noticeably less detailed (higher bitrate = bigger file, still just a few MB per song).
kHz describes the audio · kbps describes the compression
.m4a is just a container that holds AAC or ALAC data — iTunes downloads are AAC 256 with a .m4a extension.
WHAT DO I BOUNCE? (bounce = export) depends on what you’re making —
final track for streaming upload
→
48 kHz / 24-bit WAV
CD
→
44.1 kHz / 16-bit — dither on export
sharing / DJing
→
320 kbps MP3 / 256 kbps m4a — or better
these demos need headphones — phone and laptop speakers can’t reproduce the low end, and the iPhone mute switch silences them entirely.
SET YOUR LEVEL — once
this is steady pink noise. turn your system up until it’s comfortable but not loud, then leave it — every demo plays relative to this level, and nothing renormalizes later.
pink noise stepped down the ruler — −6 is half the signal, not half the loudness. | It takes about −10 to sound “half as loud” — that gap is why meters and ears disagree.
THE CEILING AT 0
0 dBFS is the ceiling.
Anything past it gets CLIPPED — the tops of the sound wave get cut off. This causes distortion.
Distortion can be an artistic choice, but it’s not usually applied in this way.
the drums stay the same loudness — past 0 only the harshness grows. that grit is the tops being shaved off.
HEADROOM & THE CHAIN
The space you leave below 0 = HEADROOM.
(ex. peaks that reach −6 dB are leaving 6 dB of headroom)
You can always turn up a low-volume take — but clipping can be permanent.
No “red-lining” (clipping) anywhere in the chain — don’t push your inputs (preamp, converter channel, DAW) into the red — ie. leave headroom on your tracks.
THE 32-BIT FLOAT SAFETY NET
The one safety net: 32-bit float recording lets you record and then recover peaks past 0 on your DAW track —
but if the preamp or converter clipped on the way in, nothing brings it back.
Most engineers say good practice is to avoid clipping at any bit depth.
HOW DECIBELS (dB) WORK
Decibels don’t count up and down evenly — they multiply.
Every −6 dB cuts the signal in half: −6 = half, −12 = a quarter, −18 = an eighth, −24 = a sixteenth (see the ruler).
A 16-bit file’s floor is −96 dB, 24-bit’s −144 — practically, anything ~60 dB under the mix is inaudible.
1 COMPRESSION
A compressor reduces the distance between the loudest and quietest points of an audio signal by running it through a gain-reduction circuit.
It nudges the loudest moments down so quieter ones can come up.
When the signal crosses a threshold, it gets turned down — in hardware or software.
Your DAW ships with one.
Compressors all have their own quirks and sound characteristics — engineers use different ones for different things.
audio compression ≠ file compression — MP3 compresses the file on export; a compressor in your signal chain processes the audio itself.
2 LIMITING
A LIMITER is a type of compressor — the extreme version. It creates a wall peaks cannot pass.
Most often used at the very end of the chain for overall loudness, on your DAW’s main output (your master track / mix bus / 2-bus).
Set the ceiling between −1 and −0.1 dBTP.
Lossy encoding (MP3/AAC) can nudge peaks up on export, so −1 prevents clipping. −0.1 runs hotter, with a little clip risk.
A WAV keeps every bit — no encoder to push peaks up, so a ceiling at 0 stays 0 in the file. Playback converters can still overshoot between samples, so leave a little.
Make sure your mix doesn’t clip, then raise the limiter gain and/or mix volume until you hit a good loudness.
3 SIDECHAIN COMPRESSION
the compressor triggers from an outside signal.
For example, every time the kick-drum mic passes a set level (say −6 dB), a compressor on other elements fires —
an intentional pulse downward at that moment, letting the kick show more in the mix.
COMPRESSORS1176 · LA-2A · CL1B · SSL bus comp
LIMITERSPro-L 2 · Waves L2 · Ozone Maximizer
DYNAMIC vs BRICKED
genre-based + taste-based — hotly debated
typical deliveries: −12 to −8 LUFS
Platforms normalize around −14. A good gut check: your loudest moments shouldn’t be much quieter than the last track on shuffle.
· louder = fewer dynamics — avoid the extremes: −5 is bricked, −20 disappears.
· the loudest genres push toward −6; more dynamic music sits at −12 or lower.
TRUE PEAK
the single tallest instant.
RMS
the average energy of the signal over time.
LUFS
the average signal weighted by frequency — perceived loudness to the human ear.
Buffer size ↔ latency
how fast you hear yourself vs. how hard your CPU works — one dropdown, one tradeoff.
#buffereach tap returns after the round trip for the buffer you picked — try tapping a steady rhythm at 2048.
· buffer = how many samples the computer collects before you hear sound — lower = less delay, harder on your computer (glitches)
· ms shown are one-way at 48 kHz — the round trip you feel is usually a bit more than double
RECORDING & OVERDUBS — latency matters: you’re performing while monitoring yourself.
The round trip — the mic/guitar → interface → DAW → back to your headphones — runs about 5–10 ms at a 64–128 buffer.
Up to about 10 ms feels instant to most players; vocalists are the pickiest — their own voice combs at even a few ms.
64–128 is the sweet spot, 32 if your rig can handle it.
MIXING — a little latency is fine, you’re just listening.
You may notice a lag when you hit the spacebar, or the playhead running slightly ahead of what you hear.
512–2048+ for big sessions and/or slower machines — to avoid audio dropouts and glitches.
Record low (32–128). Mix high (512–2048+).
Know your mics
the three major families — mic & preamp choice are technical as well as artistic choices that shape the tone of a recording.
#micsDYNAMIC
tough, likes loud signal, no power needed: stage vocals, drums, amps.
SM57 (snare/amps) · SM58 (stage vox) · SM7B (podcast) · RE20 (broadcast) · MD421 (toms, broadcast) · D112 (kick)
Some dynamics are very low-output and benefit from clean gain — an inline Cloudlifter / FetHead.
CONDENSER
sensitive & detailed, needs 48V phantom: studio vocals, acoustic.
Neumann U87 · AKG C414 · Sony C-800G · KM184 (small-diaphragm: acoustic gtr, overheads) · NT1 (budget door-opener)
Most are FET (field-effect transistor) — solid-state, run on that phantom. Tube condensers bring their own power supply.
RIBBON
smooth & dark, vintage sound, fragile: horns, rooms.
Royer R-121 · Coles 4038 (drum overheads, brass) · beyerdynamic M 160 · RCA 44 (classic)
Phantom OFF unless it’s an active ribbon that needs it — never hot-patch with 48V live.
POLAR PATTERNS — which directions the mic hears
CARDIOID — hears the front, rejects the back
OMNI — hears all around
FIGURE-8 — front & back, dead at the sides. Ribbons are natively fig-8.
THE INTERFACE SWITCHES
48V PHANTOM — power for condenser mics, sent down the XLR cable. Mute before switching it.
PAD −10/−20 dB — calms a too-hot source (drums, horns).
Hi-Z / INST — for a guitar or bass plugged straight in — high impedance. Mics & line gear are Lo-Z.
XLR (3-pin)
mono — for mics & speakers; carries 48V phantom
BALANCED
¼" TRS — 2 stripes
balanced mono into line gear OR stereo (unbalanced) in headphones — L on tip, R on ring
BALANCED
¼" TS — 1 stripe
mono, unbalanced — guitars & pedals. Keep it short.
UNBALANCED
RCA
mono, unbalanced — consumer / hi-fi gear
UNBALANCED
3.5mm mini
stereo (a tiny TRS) — phones & laptops
UNBALANCED
HOW “BALANCED” WORKS — the gear at both ends does it, not the cable.
The output sends the signal down two conductors; hum hits both, and the far end cancels it.
That’s why XLR & TRS run long without buzz — when both ends are balanced. A TRS cable can’t balance an unbalanced output (pedals, most synths, RCA) — keep those short.
Count the stripes: 1 stripe → TS · 2 stripes → TRS.
MIC vs LINE INPUT — mic level is a tiny signal that needs preamp gain.
Line level is full strength — synths, drum machines, outboard gear (TRS input).
Guitars are neither: instrument level → the Hi-Z / INST jack.
TO THE COMPUTER — USB (Focusrite Scarlett…) · Thunderbolt (UA Apollo…) · Ethernet audio-over-IP: Dante, AES67/Ravenna (Focusrite RedNet, Merging…)
PREAMP — mic signals come out tiny; the preamp lifts them up to line level. your gain knob drives input volume.
Good preamps do more than lift: they color the sound — a big part of why the classics are loved.
Inline preamps (Cloudlifter, FetHead) sit on the XLR line and add clean gain for quiet mics.
A/D D/A CONVERSION — this is your interface’s job.
A/D: mic or guitar (analog) → computer (digital).
D/A: computer (digital) → speakers (analog).
Converter + clock quality = a big part of what you pay for in an interface.
The clock times every snapshot — steadier timing (less jitter) means a cleaner, more precise capture.
In big rooms, conversion is its own rack unit. It sums all the analog inputs and converts them to digital for the DAW. Multi-channel converters (Burl Mothership · Lynx Aurora · Apogee Symphony · Merging Hapi) connect to the PC/DAW — Pro Tools · Logic · Ableton · Reaper — by Thunderbolt/Ethernet (AES67/Ravenna).
When multiple converters are connected together to grab many inputs at once, one acts as the lead master clock, keeping every converter in sync.
INTERFACES YOU’LL HEAR ABOUT — UA Apollo (Thunderbolt) · Focusrite Scarlett (USB) · RME · Merging Hapi/Anubis (AES67/Ravenna) · Focusrite RedNet (Dante) · Audient
BOARDS (consoles) — SSL · Neve · API: big summing mixers full of transformers, high-end preamps and EQs — with legendary sound-sweetening reputations.
Much of what “analog warmth” means lives here.
A FAMOUS CHAIN — “Hello” — Adele, 2015
performance · room · mic · preamp · compression · conversion · capture — every stage shapes the sound.
engineer/producer: Greg Kurstin — mixed by Tom Elmhirst. Chain quoted in SonicScoop.
DITHER — a whisper of noise that smooths the cut down to 16-bit.
At 24-bit it’s inaudible either way — most engineers skip it.
Forced to 16 (CD)? Add it on export — never twice.
The audible frequency spectrum
humans hear 20 Hz to 20 kHz — here’s roughly where everything sits.
listen, sweep, and cut to identify where elements sit in the mix.
this is a loose visualization — always use your ears and EQ sweeps to identify elements
Your ears hear pitch in octaves — doublings — not in even Hz steps.
So frequency charts are drawn logarithmically, spreading the low frequencies across more of the range to even it out: 20–640 Hz fills the left half and 640–20,000 Hz fits in the right half.
SOME COMMON EQ MOVES
Muddy → roll off 20–400 Hz · Boxy → cut 300–500 Hz · Honky → cut 500 Hz–1 kHz · Harsh → cut 2.5–4 kHz · Dull → shelf up 10 kHz+
THE SHAPES YOU’LL REACH FOR
Many engineers advocate subtractive EQ.
Using a dip/cut at one frequency makes others stand out; boosting can cause issues in the mix and cost headroom.
TL;DR Ref Sheet
#recapbuffer per task — low to track, high to mix
48 kHz / 24-bit — record it AND bounce it
avoid clipping — keep peaks below 0 dB when recording
limiter on — deliver −12 to −8 LUFS · ceiling −1 to −0.1 dBTP
use your ears — numbers can’t tell you how the music sounds
.wav full size, full quality · .mp3 smaller, compressed
Follow for the VIDEO edition