The noise floor is the total level of unwanted background noise present in an audio system when no intended signal is passing through it. In recording, it marks the lower boundary of usable dynamic range: any sound quieter than the noise floor is masked or buried in hiss, hum and room noise.
Every stage of the signal chain contributes to it, from the acoustic background of the room through the microphone’s self-noise and the preamp to the converter. The recording’s effective noise floor is set by the loudest of these contributions once they are referred to the same level, which is why one weak stage can spoil an otherwise quiet chain.
Where the noise floor comes from
The noise floor comes from a mix of physical, electrical and digital sources. Some are unavoidable consequences of physics, such as thermal noise in resistors and transistors, while others are faults that careful setup can remove, such as ground loops.
| Source | Character | Typical cause |
|---|---|---|
| Room and ambient noise | Broadband rumble, hum, traffic | Ventilation, appliances, outside sound |
| Microphone self-noise | Steady hiss | Capsule and internal electronics |
| Preamp noise | Hiss rising with gain | Thermal noise of input stage |
| Mains hum and buzz | 50 or 60 Hz and harmonics | Ground loops, unshielded cables |
| Interference | Clicks, whine, chirps | Radio, phones, switching power supplies |
| Quantization noise | Low-level grain or distortion | Too few bits for the signal level |
In a typical home or project studio, the acoustic background usually dominates. A quiet room may sit around 25 to 35 dBA, while a well-designed studio aims below 20 dBA. A microphone with a self-noise of 7 dBA cannot reveal its quietness in a room that is 20 dB louder.
Independent noise sources add by power, not by level. Two equal sources raise the combined floor by about 3 dB, while a source 10 dB quieter than the dominant one adds less than half a decibel. This is why improving any stage other than the loudest contributor changes almost nothing.
How the noise floor is measured
The noise floor is measured as a level relative to a reference, and the units differ by stage. Comparing figures only makes sense when they are expressed in the same unit and weighting.
| Stage | Unit | Typical range |
|---|---|---|
| Microphone self-noise | dBA equivalent SPL | About 3 to 20 dBA |
| Preamp equivalent input noise | dBu, A-weighted or unweighted | About -125 to -130 dBu |
| 16-bit digital audio | dBFS | About -96 dBFS theoretical |
| 24-bit digital audio | dBFS | About -144 dBFS theoretical |
| Real 24-bit converters | dBFS | About -110 to -125 dBFS |
A-weighting de-emphasizes low and very high frequencies to approximate human hearing at low levels, so A-weighted figures look several decibels better than unweighted ones. Preamp equivalent input noise (EIN) is meaningful only with the stated gain and source impedance, commonly a 150-ohm source at high gain.
The difference between the loudest undistorted level and the noise floor is the dynamic range, closely related to the signal-to-noise ratio. In practice, the theoretical 144 dB of 24-bit audio is never reached, because analog circuitry limits real converters to about 110 to 125 dB.
Noise floor, headroom and gain staging
The noise floor and headroom are the two edges of the same window. Gain staging places the signal inside that window: high enough above the noise floor to stay clean, low enough below clipping to absorb peaks.
| Recording level | Distance above noise floor | Risk |
|---|---|---|
| Peaks near -40 dBFS | Small at 24-bit after analog noise | Preamp hiss audible when boosted |
| Peaks around -18 to -12 dBFS | Large | Balanced; generous headroom |
| Peaks around -6 dBFS | Very large | Little room for unexpected peaks |
| Peaks touching 0 dBFS | Maximum | Digital clipping, irreversible distortion |
At 24-bit, the converter’s own noise is rarely the limit. Raising the gain at the preamp, not in software, is what keeps quiet sources clear of the analog noise floor. Turning up a weak recording later raises the signal and the noise together, so the ratio between them does not improve.
Lowering the noise floor in practice
Lowering the noise floor means addressing the dominant source first. Buying a quieter microphone gives no benefit if the refrigerator in the next room is louder than the microphone’s self-noise.
- Switch off or isolate ventilation, computers and appliances during takes.
- Place the microphone closer to the source, which raises the signal without raising the room noise.
- Use balanced XLR cables and a common power point to avoid ground loops and hum.
- Choose a low self-noise microphone for quiet sources such as voice-over, Foley or classical guitar.
- Set preamp gain so speech or instruments peak near -12 dBFS.
- Use inline gain boosters for low-output dynamic and ribbon microphones when the preamp runs out of clean gain.
For field recording, the noise floor often sets the whole approach. Recordists capturing quiet nature sounds choose microphones with self-noise below about 10 dBA and recorders with clean preamps, because the environment itself may be only slightly louder.
Common mistakes with the noise floor
The most frequent mistake is relying on noise reduction to repair a recording made too quietly. Modern tools can remove steady hiss surprisingly well, but they also remove low-level detail and room tone, and heavy settings leave watery, metallic artifacts.
- Noise floor ignored during monitoring: listening at high volume in quiet passages reveals hiss that loud playback hides.
- Compression after recording: heavy compression and limiting raise the noise floor along with quiet parts.
- Noise gates as a fix: gates hide noise in pauses but make it jump in and out audibly.
- No room tone recorded: editors need a clean sample of the background to fill gaps and train noise reduction.
The simple explanation
Every recording has a little background sound, even when nobody is making noise. It might be a faint hiss, a hum from electricity or the air conditioning. This constant background is the noise floor.
Think of it as the water level in a pool. Loud sounds stand high above it and stay clear. Quiet sounds can sink below the surface and disappear. A good recording keeps the water level low and the voice well above it.
How it is used in practice
An audiobook narrator works in a small closet booth with a large-diaphragm condenser microphone. The publisher requires a noise floor below -60 dBFS in silent passages, but the first test files measure about -52 dBFS, with an audible low hum.
The narrator finds that the computer’s fan and a laptop charger on the same power strip cause most of the noise. Moving the computer outside the booth, replacing the cable with a balanced XLR run and raising preamp gain so speech peaks near -12 dBFS brings the floor down to about -68 dBFS. The files now pass review without any noise reduction.