Smoothing And Resolution
Canonical Definition
Smoothing reduces fine variation in displayed or analysed data over a defined frequency bandwidth, while resolution describes the measurement's ability to distinguish separate features. Smoothing can hide detail but cannot create resolution or improve the underlying measurement.
Three Different Ideas Often Mixed Together
| Setting | What it controls | Typical misunderstanding |
|---|---|---|
| Smoothing | How much local frequency detail is averaged in the displayed trace | Believing it improves the raw measurement |
| Data spacing or points per octave | How densely frequency points are stored or displayed | Assuming dense points guarantee independent resolution |
| Time or FFT resolution | How well the analysis separates close frequency components | Ignoring window duration, sample rate, and record length |
Common Smoothing Views
- No smoothing: preserves displayed fine structure and noise; useful for diagnostics when the measurement quality supports it.
- 1/24 or 1/12 octave: common for crossover, resonance, and detailed response work.
- 1/6 octave: useful for general tuning and tonal-shape review while retaining moderate detail.
- 1/3 octave: broad presentation or RTA-style tonal overview; often too coarse for narrow crossover and cancellation diagnosis.
- Psychoacoustic smoothing: variable bandwidth intended to produce a perceptually oriented response view; useful for presentation, not a universal pass/fail rule.
- ERB smoothing: variable bandwidth based on equivalent rectangular bandwidth; it can be very broad at low frequencies and settles near a finer fractional-octave view at higher frequencies.
Choose The View By The Question
| Question | Useful starting view | Also inspect |
|---|---|---|
| Overall tonal balance | 1/6 octave or psychoacoustic | individual positions and target context |
| Crossover summation | 1/24 or 1/12 octave | phase, polarity, individual drivers, nearby positions |
| Narrow resonance | little or no smoothing | decay, repeatability, distortion, nearby positions |
| Live RTA adjustment | 1/12 to 1/3 octave depending task | averaging, noise floor, source spectrum |
| Public summary graph | clearly labelled smoothed view | link or retain raw measurement data |
Window Length Sets A Frequency Limit
When frequency response is derived from a gated impulse response, a short window cannot support valid low-frequency detail. A useful orientation is that a 10 ms effective window cannot resolve behaviour below roughly 100 Hz. The exact validity depends on window placement and processing, so use the software's stated limit and display invalid lower data as such.
Graph Scale Is Part Of The Presentation
The same data can appear smooth or terrible depending on the vertical range. Keep before-and-after graphs on the same axes. Record the level span, frequency range, smoothing, and normalisation so the visual impression can be audited.
Reporting Rules
- name the smoothing type and amount
- name the frequency spacing or points per octave when exported data is compared
- name the impulse window and valid frequency limit for gated data
- keep axes identical for comparison graphs
- retain the unsmoothed measurement even when publishing a smoothed view