What the noise floor means
The noise floor is the combined level of unwanted background energy in a recording or audio system. It is usually measured in dBFS for digital files or related electrical units in hardware specifications. A more negative digital value means a lower, quieter noise floor.
The number alone does not identify the cause. A steady broadband hiss, a 50 or 60 Hz hum, air-conditioning rumble and intermittent computer interference can produce similar overall levels but require different solutions.
Signal-to-noise ratio
Signal-to-noise ratio compares the wanted signal with the noise beneath it. If speech averages −20 dBFS and the background sits near −70 dBFS, the effective difference is about 50 dB. A larger difference generally makes the recording sound cleaner.
SNR must be evaluated in context. Quiet classical music needs a lower noise floor than a dense distorted guitar track. Spoken-word recordings expose steady noise during pauses, while loud music can mask the same background level.
Common sources of noise
Microphones and preamps create a small amount of electronic noise. Excessive preamp gain, weak microphone output or recording from too far away can make that noise more obvious. Rooms add ventilation, traffic, reflections and appliance sounds.
Ground loops and power problems often create low-frequency hum and harmonics. USB devices, displays, lighting dimmers and poorly shielded cables can add buzzing or high-frequency interference. In digital files, repeated low-level processing can also expose quantization or codec artifacts.
How to measure the noise floor
Find a section where the intended source is silent but the recording chain remains active. Measure that section with an analyzer, inspect the spectrum and listen on headphones. The spectrum helps distinguish broadband hiss from narrow hum frequencies and low-frequency rumble.
Do not measure a fully muted or digitally silent section, because it may report an unrealistically low value. The useful test is the real background captured by the same microphone, gain and room conditions as the wanted signal.
Preventing noise during recording
Move the microphone closer to the source while maintaining a suitable polar pattern and avoiding plosives. Set gain so the source records clearly with headroom, turn off unnecessary devices, use balanced cables where possible and separate audio cables from power supplies.
Record a short test before the full take. Listen for fans, refrigerators, traffic and hum, then correct the environment. Ten minutes spent fixing the source usually produces a cleaner result than aggressive noise reduction later.
Reducing noise after recording
Start with gentle high-pass filtering for unnecessary low-frequency rumble and narrow notch filters for clearly identified hum. Broadband denoising can learn a noise profile, but excessive reduction often creates metallic, watery or phasey artifacts.
Apply the smallest amount that solves the practical problem. For speech, editing pauses or using an expander can be more natural than removing noise continuously. Always compare with the original at matched loudness and preserve an untouched source file.
Common mistakes and checklist
Do not normalize a noisy recording before diagnosis, because raising the whole file also raises the noise. Do not confuse room ambience with electronic hiss, and do not use a strong noise gate that cuts word endings, breaths or reverb tails.
Before publishing, inspect silent passages, check the spectrum for hum and rumble, confirm that noise reduction has not damaged clarity, and listen on both headphones and speakers. A slightly audible natural background is often preferable to severe processing artifacts.