The largest arrival the fit can express, in ms, for a window of fftSize at sampleRate.
THE WHOLE ANALYSIS WINDOW, since the fit auto-ranges (smart-align.ts's anchor): the delay
is the CORRELATION peak, which names a lag over the full fftSize samples, and §9b declares
the arrival physical so all of it is arrival rather than half of it being a lead. About
170.7 ms, 58.5 m of air — and at EVERY graph rate, because omx_rta_fft_size_for derives
the length from the rate to hold the window time fixed (spec mix_dsp.h). The number this
replaces, +/- fftSize / 2, was the phase SLOPE's reach: half as far, and it answered a
30 m flight with -28.5 m rather than refusing.
This is not a tuning knob — it is what the arithmetic can say — and it is not a promise that
a reading out there will be BELIEVED in one pass. See speakerAlignOverlapReachMs
and speakerAlignAim.
The largest arrival the fit can express, in ms, for a window of
fftSizeatsampleRate.THE WHOLE ANALYSIS WINDOW, since the fit auto-ranges (
smart-align.ts's anchor): the delay is the CORRELATION peak, which names a lag over the fullfftSizesamples, and §9b declares the arrival physical so all of it is arrival rather than half of it being a lead. About 170.7 ms, 58.5 m of air — and at EVERY graph rate, becauseomx_rta_fft_size_forderives the length from the rate to hold the window time fixed (specmix_dsp.h). The number this replaces, +/-fftSize / 2, was the phase SLOPE's reach: half as far, and it answered a 30 m flight with -28.5 m rather than refusing.This is not a tuning knob — it is what the arithmetic can say — and it is not a promise that a reading out there will be BELIEVED in one pass. See speakerAlignOverlapReachMs and speakerAlignAim.