Gain structure
Setting levels so every stage sits well above noise and below clipping.
Every audio system is a chain of stages — preamp, converters, DSP, outputs, amplifier — and each one has a floor of noise underneath and a ceiling where it clips. Gain structure is the practice of setting each stage's level so the signal sits comfortably between the two, all the way through. Get it right and the system is quiet and clean. Get it wrong and no amount of EQ will save it.
Noise floor, headroom and dynamic range
| Term | Meaning |
|---|---|
| Noise floor | The hiss and hum a stage produces with no signal at all |
| Clip point | The highest level the stage can pass before the waveform is flattened |
| Headroom | The gap between the signal's peaks and the clip point |
| Signal-to-noise ratio | The gap between the signal's average level and the noise floor |
| Dynamic range | The gap between the noise floor and the clip point — the space you have to work in |
- Noise floor
- The hiss and hum a stage produces with no signal at all
- Clip point
- The highest level the stage can pass before the waveform is flattened
- Headroom
- The gap between the signal's peaks and the clip point
- Signal-to-noise ratio
- The gap between the signal's average level and the noise floor
- Dynamic range
- The gap between the noise floor and the clip point — the space you have to work in
Speech isn't a steady level. Its peaks typically sit well above its average — roughly 10 to 15 dB — so a channel that looks fine on a meter can still clip on plosives and laughter. Headroom is what absorbs those peaks.
Gain early, then leave it alone
The golden rule: get the signal to its operating level at the first gain stage, and then run every stage after that close to unity (0 dB).
- Too little gain at the preamp leaves the signal close to the noise floor. Boosting it later raises the noise by exactly the same amount — SNR lost at the start is lost for good.
- Too much gain at the preamp clips the converters on peaks. Clipping can't be undone either; turning down later just makes quieter distortion.
-42 dBFS
Aim for peaks well clear of 0 dBFS but not buried — many integrators target roughly −20 to −10 dBFS.
60 dB
Once noise is amplified, no later stage can remove it.
4 dB
Peak level below the amplifier's full-power input.
Teaching model with typical values: speech at −50 dBu average from the mic with 12 dB peaks, preamp EIN −126 dBu, converters aligned so 0 dBFS = +24 dBu with about 114 dB dynamic range, amplifier full power at +4 dBu. Real devices differ — read their specs.
dBFS and alignment
Inside a DSP, levels are digital and measured in dBFS — decibels relative to full scale, where 0 dBFS is the largest value the converters can represent. Every DSP defines how analog levels map to digital ones; a common alignment puts 0 dBFS at somewhere between +18 and +24 dBu at the converters. Check the spec sheet: that figure tells you how much headroom a +4 dBu signal has.
A practical procedure
- Set every fader and output in the chain to unity (0 dB) and amplifiers to minimum.
- Have someone talk at a normal level, at the normal distance from the mic. Raise the preamp until speech peaks land comfortably below full scale — many integrators aim around −20 to −10 dBFS.
- Apply processing (EQ, dynamics, AEC). Keep the overall channel level near where it started.
- Set DSP outputs to drive amplifiers at an appropriate level, then bring each amplifier up until the room reaches the target level.
- Check the quietest and loudest talkers, and listen for noise with all mics open and nobody talking.
Key takeaways
- 01Every stage has a noise floor and a clip point; gain structure keeps the signal comfortably between them.
- 02Set the signal to its operating level at the first gain stage, then run later stages near unity.
- 03Noise added early is amplified with the signal; clipping can't be undone by turning down later.
- 04Speech peaks sit roughly 10–15 dB above its average, so leave headroom.
- 05Know your DSP's alignment — how many dBu equal 0 dBFS — to judge headroom.
Check your understanding
Q1A mic channel has low preamp gain and the DSP output is turned up to compensate. What's the likely result?
Q2Speech peaks are clipping at the A/D converter. Where should you fix it?
Q3A DSP is aligned so that 0 dBFS = +24 dBu. How much headroom does a +4 dBu signal have before clipping?