DFFT(1)

NAME

dfft - compute fast Fourier transforms of a stream of doubles

On this page

SYNOPSIS
DESCRIPTION
OPTIONS
EXAMPLES
SEE ALSO
AUTHOR
COPYRIGHT
BUG REPORTS

SYNOPSIS

dfft [-clpLAN] [-d _dB_] [_width_] < doubles > spectrum

dfft reads a stream of double-precision floating-point samples from standard input, computes the real fast Fourier transform of each 1024-sample record, and writes the resulting spectrum to standard output. Both standard input and standard output must be redirected; if either is a terminal, dfft prints a usage message and exits.

By default the output is the periodogram-scaled log-magnitude in decibels, written as 512 double-precision values per 1024-sample record. Options select phase output instead of magnitude, a linear (non-dB) magnitude scale, a logarithmic frequency scale, a third-octave critical band filter, normalization to the peak magnitude, or human-readable ASCII text.

The optional width argument is accepted for historical compatibility but is ignored; the transform length is fixed at 1024 samples, yielding 512 output values. Input is normally produced by a tool such as dconv(1) or dmod(1).

OPTIONS

-d dB: Set the minimum magnitude in decibels (a positive value, applied as its negative). Magnitudes below this floor are clamped when output is normalized. The default is 120, i.e. -120 dB.
-l: Interpolate the magnitudes onto a logarithmic frequency scale.
-c: Apply a third-octave critical band filter to the magnitudes.
-p: Output the phase, in units of pi, instead of the magnitude.
-N: Normalize the power spectral density so the peak is 0 dB (or 1.0 in linear mode).
-L: Output linear magnitude rather than log-magnitude in decibels.
-A: Write human-readable ASCII text instead of binary doubles. Each line holds a frequency fraction and its magnitude separated by a space.
-h, -?: Print a usage message and exit.

EXAMPLES

Example 1. Plot a power spectrum

dfft -A -N < signal.doubles > spectrum.txt

Computes the FFT of the samples in signal.doubles, normalizes the power spectral density to a 0 dB peak, and writes (frequency, magnitude) pairs as ASCII text suitable for feeding to a plotting program.

Example 2. Critical band magnitude on a log frequency scale

dmod -m 32768 < raw.doubles | dfft -c -l > bands.doubles

Scales the raw samples with dmod(1), then computes a third-octave critical-band-filtered spectrum interpolated onto a logarithmic frequency scale, producing binary doubles for further processing.