dconv filterfile < doubles > doublesDCONV(1)
NAME
dconv - fast FFT-based convolution of a double-precision data stream
On this page
SYNOPSIS
DESCRIPTION
dconv convolves a continuous stream of double-precision floating-point numbers, read from standard input, with the filter kernel (impulse response) stored in filterfile, writing the convolved stream to standard output. Both the input data and the kernel file are raw, host-order, double-precision values (see d-f(1) and f-d(1) for converting between double and float streams).
The convolution is computed with a fast Fourier transform (FFT) using the overlap-save method, which yields a true linear convolution rather than the circular convolution produced by a naive block FFT. Because both standard input and standard output must be redirected, dconv refuses to run when either is a terminal.
The number of kernel elements in filterfile must be one less than a power of two (2^i - 1, e.g. 127, 255, 511). The kernel may contain at most BU_PAGE_SIZE elements.
EXAMPLES
Example 1. Low-pass filter a sampled signal
dconv lowpass.kernel < signal.d > filtered.dConvolves the double stream in signal.d with the 2^i - 1 element kernel in lowpass.kernel, producing the smoothed result in filtered.d. A float-format recording can be filtered in a single pipeline by converting on the fly: f-d(1) ` < signal.f | dconv lowpass.kernel | ` d-f(1) ` > filtered.f`.
SEE ALSO
brlcad(1), dfft(1), dmod(1), dstats(1), d-f(1), f-d(1)
AUTHOR
BRL-CAD Team
COPYRIGHT
This software is Copyright (c) 1989-2026 by the United States Government as represented by U.S. Army Research Laboratory.
BUG REPORTS
Reports of bugs or problems should be submitted via electronic mail to devs@brlcad.org