Gain staging is the single most neglected step in desktop audio, and it is also the one that separates clean vocal recordings from hiss-and-clip disaster reels. The idea is simple: set each gain stage so your signal is loud enough to bury the noise floor but far enough from clipping that an unexpected loud syllable does not wreck the take. Too little gain and you fight hiss in post; too much and you get digital clipping that no plugin can repair. In 2026, most interfaces give you more clean gain than the average setup actually needs, so the practical problem has flipped — the mistake is usually too much, not too little. This guide gives you concrete starting values, explains the trade-off behind each dial, and shows how the advice changes depending on whether you own a $99 unit or a flagship converter.
What to prioritize, in order: correct input type (XLR combo jack on mic mode, not line/instrument), proper gain so peaks land around −18 dBFS to −12 dBFS in your DAW meter, phantom power only when the mic needs it, pad and high-pass decisions made before you record, and buffer settings that keep latency tolerable for monitoring. Everything else is secondary.
Quick Recommended Settings
| Setting | Recommended value | Why |
|---|---|---|
| Input type | Mic (XLR combo jack, mic-level mode) | Line/instrument mode bypasses the preamp and leaves a dynamic mic far too quiet. |
| 48V phantom power | On for condensers, off for dynamics like SM58/SM7B | Phantom on a ribbon mic without protection can damage it; on dynamics it is usually harmless but pointless. |
| Gain level | Peaks between −18 and −12 dBFS in the DAW | Leaves ~12 dB of headroom for plosives and laughter while staying well above the noise floor. |
| Pad | −10 or −20 dB for loud sources or close-miked shouty delivery | Lets you run the preamp in its sweet spot without input clipping. |
| High-pass filter (on interface, if present) | 75–80 Hz engaged for voice | Removes rumble and desk thumps before they eat headroom. |
| Air/transformer/inst bulk buttons | Off until after a clean reference take | Coloration is easier to add in post than to remove. |
| Direct monitor | On for tracking, off when using DAW monitoring | Avoids hearing both paths at once or double-monitoring with latency. |
| DAW buffer size | 128–256 samples while tracking; 512–1024 for mixing | Lower buffers cut latency; higher buffers give plugins more headroom to avoid crackle. |
Setting-by-Setting
Gain level: target −18 to −12 dBFS peaks
Exact value: Speak or play at your real tracking level, then turn the gain knob up until the interface’s clip LED flashes only on your loudest moments, and back it off until it never lights. In the DAW, peaks should sit between −18 dBFS and −12 dBFS on the track meter.
Trade-off: Every dB of extra gain raises the noise floor along with the signal. Every dB of shortfall tempts you to boost in post, which amplifies preamp hiss and room noise equally. A 24-bit interface has enormous theoretical dynamic range, but only if the analog front end is set right — no amount of bit depth rescues a signal recorded 30 dB too quiet. If your mic is a low-output dynamic (an SM7B needs roughly 60 dB of gain for quiet talkers), a bare-bones interface may genuinely run out; that is a hardware problem an inline booster like a Cloudlifter-style active gain device solves, not a knob problem.
Pad: engage at −10 to −20 dB for hot sources
Exact value: If the clip LED flashes even with the gain knob below the 9 o’clock position, switch on the pad (labeled “PAD −10 dB” or “−20 dB” on the rear panel or top of many units) and rebuild your gain setting from silence.
Trade-off: The pad reduces the signal before it hits the preamp, trading sensitivity for headroom. With the pad on, you will need more knob travel, which is fine — most preamps are cleanest between 10 and 40 percent of rotation anyway. The downside is slightly less signal-to-noise ratio if you over-pad and then crank gain to compensate. Check with a simple test: record ten seconds at your loudest realistic delivery and look at the waveform in your DAW — if peaks touch −6 dBFS or higher, pad down.
High-pass filter: 75–80 Hz for voice
Exact value: Flip the HPF/low-cut switch on the interface channel (often a small recessed button labeled “80 Hz” or “LOW CUT”). If your interface lacks one, add your DAW’s stock EQ high-pass at 75 Hz with a 12 dB/octave slope as the first insert.
Trade-off: Cutting below 75 Hz removes rumble, HVAC hum, and desk vibrations that eat headroom without harming vocal intelligibility — male fundamentals sit around 85–180 Hz. Go higher (100–120 Hz) only for aggressively bright broadcasts where boominess is a known problem, and never for bass instruments or full-range sources.
Direct monitoring and buffer size
Exact value: Set the interface’s “DIRECT MONITOR” switch to ON (or in the control panel app, e.g. Focusrite Control / MOTU Pro Audio control panel, enable “Loopback: Off, Direct: On”) while tracking. In your DAW’s audio settings (Preferences > Audio > Buffer Size, or ASIO Control Panel on Windows), set 128 or 256 samples for tracking; raise to 512–1024 for mixing.
Trade-off: Direct monitoring is latency-free because you hear the raw input before the computer round-trip. The cost is that you cannot hear your DAW’s processing (compression, reverb) while tracking. If you prefer monitored plugins, keep direct monitor off and drop the buffer to 64–128 samples, accepting higher CPU load and the risk of clicks if you overload the machine.
By Hardware Tier
Low tier (budget 2-in/2-out units, roughly $99–$180)
These often deliver 50–56 dB of gain, which is marginal for low-output dynamic mics. Practical settings: run the gain knob between 60 and 85 percent for an SM7B-class mic and accept you may need an inline booster (adds ~+25 dB clean gain via XLR, requires phantom power). For condensers, keep the gain at 30–50 percent — their hot output means less knob travel is needed. Avoid engaging any “Air” or “Boost” style coloration circuit until after you have a clean baseline take, since these circuits also raise the noise floor on cheaper preamps. Keep the buffer at 256 samples if you hear crackle; the USB bus on budget units is less forgiving.
Mid tier ($300–$700 units with DSP or better preamps)
With 60–70 dB of gain available, a dynamic mic sits comfortably at 40–60 percent knob rotation. If your unit has a software control panel (check the manufacturer’s utility, usually installed to C:\Program Files\[Manufacturer] on Windows or /Applications on macOS), lock the sample rate to 48 kHz there and in the DAW — mismatched rates between panel and DAW are the top cause of distorted or muted inputs at this tier. Use the panel’s built-in input metering as your reference: aim for the meter to average in the lower third with peaks in the middle third. Enable any onboard DSP compression at a mild 2:1 ratio for monitoring only if your DAW latency budget allows.
High tier ($700+ converters/preamps, rack or flagship desktop)
Premium units often expose trim in precise steps (e.g. 1 dB increments via a rotary encoder or control-panel slider). Set input trim so a reference test recording peaks at −18 dBFS, then use any analog output trim to calibrate your monitor path separately. If the unit offers selectable input impedance (e.g. 1.5 kΩ vs 10 kΩ), the higher setting tames bright condensers slightly; the lower setting fattens some ribbon mics. Keep sample rates at 48 kHz for voice work unless a project specifically requires 96 kHz — the files double in size for no audible benefit in podcasts or streaming. Document your trim settings in the unit’s onboard scene/preset memory so recall is instant.
Common Mistakes
Recording 20+ dB too quiet, then boosting in post
What causes it: Trusting the knob’s visual position rather than the meter, or assuming “gain at 50 percent” is a universal setting.
How to check: Record 15 seconds at your real level. In the DAW, look at the clip waveform: if peaks are below −24 dBFS, you are too quiet. Also zoom into the waveform’s vertical scale — if it looks like a flat line with tiny wiggles, same problem.
What to do: Raise gain until peaks land between −18 and −12 dBFS during a real delivery, not a gentle test sentence. If the knob maxes out before that on a dynamic mic, add an inline booster or move the mic 2–4 inches closer to your mouth.
How to undo: If you already recorded quiet takes, use a gain plugin in the DAW to add digital gain first, then apply noise reduction — or re-record. Avoid stacking multiple boost plugins, which compounds quantization noise.
Clipping because the pad was never engaged
What causes it: Close-miking a loud source (shouted commentary, screaming, drum amps) with a hot condenser and no pad.
How to check: The interface’s clip LED flashes during normal delivery, or the DAW waveform shows hard-clipped flat tops.
What to do: Engage the −10 dB or −20 dB pad, then re-dial gain from zero. Re-test with your loudest realistic take.
How to undo: Flip the pad off for quiet sources and rebuild gain. Clipped audio itself cannot be repaired — re-record the take.
Phantom power left on into the wrong mic
What causes it: Leaving 48V engaged when swapping between a condenser and an unbalanced-output dynamic or vintage ribbon.
How to check: Look at the 48V LED on the interface front panel. If lit and you have just connected a ribbon mic, power down before continuing.
What to do: Turn phantom off (front-panel switch, or in the control panel app under Inputs > Phantom). Hot-swap only with phantom off.
How to undo: Turn phantom back on when reconnecting a condenser, and give it 10–15 seconds to stabilize before tracking.
Mismatched sample rate between interface panel and DAW
What causes it: The interface control panel is set to 44.1 kHz while the DAW project is 48 kHz, or Windows shared-mode audio is at a different rate.
How to check: On Windows, right-click the speaker icon > Sounds > Playback > [your interface] > Properties > Advanced > Default Format, and compare to the DAW’s project sample rate. On macOS, open Audio MIDI Setup, select the device, and check the format dropdown.
What to do: Set panel, OS, and DAW all to 48 kHz (or all to 44.1 kHz), then reconnect the device or restart the DAW.
How to undo: Change any single setting back and restart the audio engine; rates only need to match, so reverting one is safe as long as you keep them consistent.
FAQ
What dBFS should my voice peak at when recording?
Peaks between −18 and −12 dBFS. This leaves roughly 12 dB of headroom so a laugh or plosive does not clip, while keeping the signal far enough above the noise floor that you never need aggressive post-boosting. Loudness for delivery (podcast platforms, streaming) is handled later in mastering, not at the interface.
Do I need 48V phantom power for my microphone?
Only if it is a condenser or an active ribbon/booster device. Standard dynamics like an SM58 or SM7B do not need it. Check the mic’s spec sheet: if it lists “phantom power requirement: 48V ±4V,” it needs phantom; if it lists only an output impedance, it does not.
Why does my recording sound noisy even at moderate gain?
Three usual suspects: the mic’s output is too low for your preamp (fix with an inline booster or a closer mic position), a gain-after-conversion mistake (raising DAW fader volume on a quiet recording just amplifies the recorded hiss — raise interface gain instead), or a ground-loop/hum problem, which shows up as a steady 60 Hz tone rather than broadband hiss. Diagnose by recording 10 seconds of silence at your working gain and inspecting the noise floor level in the DAW.
What buffer size should I use while tracking?
128 or 256 samples is the sweet spot for most systems — low enough that direct or DAW monitoring feels instant, high enough to avoid CPU overload clicks. Drop to 64 only if you monitor through DAW plugins and your machine handles it; raise to 512 or 1024 when mixing, when latency no longer matters and plugin headroom does.
