How do I read the live frequency spectrum in Decibel Reader?

What do the two axes mean?

The horizontal axis is logarithmic frequency in Hz. Equal visual distances represent frequency ratios, rather than equal numbers of Hz. Moving from 100 to 200 Hz is an octave, just as moving from 1,000 to 2,000 Hz is. Read the tick labels instead of estimating frequency from the fraction of screen width.

The vertical axis runs from −120 to 0 dBFS. Each bar sums the window-corrected power within that band, with RMS 1 as the digital reference. A full-scale peak sine has total RMS power of −3.01 dBFS. A tone near a band boundary can spread across neighboring bars. Wider logarithmic bands collect more broadband-noise power, so they should not be read as equal-width spectral density. A value closer to zero is stronger relative to the digital reference. It is not a calibrated sound-pressure scale, and the bars do not use the display offset applied to the main Estimated dB number.

Frequency colors locate bands: blue below 250 Hz, green from 250 to below 2,000 Hz, and purple from 2,000 Hz upward. They do not use the sound-level chart’s caution colors or represent safety categories.

Why do bars appear near 20 kHz when my voice sounds low?

The right edge is the display range limit, not the detected main frequency. Decibel Reader displays from 20 Hz to the lower of 20 kHz and half the actual sample rate. A 16,000 Hz sample rate therefore limits the axis to 8,000 Hz. These are range calculations, not collected measurements.

A low-sounding voice can have higher-frequency components, and the microphone can also receive other sound and digital background energy. A bar at the right side only indicates signal energy in that band. The graph does not determine who or what produced it.

The fixed vertical range also makes weak bands visible. For example, −90 dBFS sits one quarter of the way up an axis running from −120 to 0. This is a coordinate example, not a measured noise floor.

Does “Strongest frequency” tell me my voice pitch?

No. The label marks the strongest actual FFT bin inside the displayed range. The fundamental frequency of a voice need not be its strongest spectral component. A harmonic or another sound can be stronger, so the label must not be treated as a vocal-pitch result.

The analyser uses a 16,384-point FFT. Bin spacing is the sample rate divided by 16,384: at 48,000 Hz it is about 2.93 Hz. The analysis window is about 341 ms at that rate; at 16 kHz it is about 1.02 seconds. The spectrum waits for a full window after Start. Fast changes spread over this window, and a Blackman window can spread a single tone over several bins. An approximate integer-Hz label does not imply 1 Hz resolution or a precise voice-pitch estimate. The spectrum sums window-corrected digital power into 32 logarithmic display bands, while the marker refers to the strongest individual FFT bin.

The Web Audio specification describes the analyser’s frequency data and smoothing. That digital analysis does not establish the microphone’s acoustic accuracy.

Why are some bars faint?

Decibel Reader makes bands below −80 dBFS less opaque, at 0.35 opacity. Their coordinates and numerical levels stay on the same scale. This visual threshold does not classify the band as unwanted noise and does not remove it from the input.

A faint high-frequency bar can coexist with a stronger low-frequency component. Changing the display’s calibration offset changes Estimated dB but leaves the raw spectrum unchanged. The spectrum is neither a noise gate nor an automatic voice identifier.

How can I use the spectrum for a comparison?

1. Start measuring and allow microphone access. Keep the device position and input settings fixed.
2. Observe the sound-level history and spectrum for condition A.
3. Change one condition, such as a fan setting, and observe condition B.
4. Return to A to check whether the change repeats.

Do not diagnose a source from a single peak. Repeatable changes provide context, but gain control and microphone frequency response can still influence the result. The room-comparison guide explains how to keep the setup consistent.

What happens when I stop or save a report?

Stopping, interruption or Reset clears the live spectrum and strongest-frequency marker. Decibel Reader keeps only the current derived spectrum frame for display; it does not retain spectrum history, record audio or upload the microphone signal.

A saved report contains the level statistics and full-session overview, sampling method, offset and accuracy status. It does not export audio or a spectrum history. For report interpretation, keep the accuracy limits attached and note the conditions you compared.