Fast FFT filter with float input, float output and float taps. More...
#include <gnuradio/filter/fft_filter.h>
Public Member Functions | |
fft_filter_fff (int decimation, const std::vector< float > &taps, int nthreads=1) | |
Construct an FFT filter for float vectors with the given taps and decimation rate. More... | |
fft_filter_fff (const fft_filter_fff &)=delete | |
fft_filter_fff & | operator= (const fft_filter_fff &)=delete |
fft_filter_fff (fft_filter_fff &&)=default | |
fft_filter_fff & | operator= (fft_filter_fff &&)=default |
int | set_taps (const std::vector< float > &taps) |
Set new taps for the filter. More... | |
void | set_nthreads (int n) |
Set number of threads to use. More... | |
std::vector< float > | taps () const |
Returns the taps. More... | |
unsigned int | ntaps () const |
Returns the number of taps in the filter. More... | |
int | nthreads () const |
Get number of threads being used. More... | |
int | filter (int nitems, const float *input, float *output) |
Perform the filter operation. More... | |
Fast FFT filter with float input, float output and float taps.
This block performs fast convolution using the overlap-and-save algorithm. The filtering is performand in the frequency domain instead of the time domain (see gr::filter::kernel::fir_filter_fff). For an input signal x and filter coefficients (taps) t, we compute y as:
This kernel computes the FFT of the taps when they are set to only perform this operation once. The FFT of the input signal x is done every time.
Because this is designed as a very low-level kernel operation, it is designed for speed and avoids certain checks in the filter() function itself. The filter function expects that the input signal is a multiple of d_nsamples in the class that's computed internally to be as fast as possible. The function set_taps will return the value of nsamples that can be used externally to check this boundary. Notice that all implementations of the fft_filter GNU Radio blocks (e.g., gr::filter::fft_filter_fff) use this value of nsamples to compute the value to call gr::block::set_output_multiple that ensures the scheduler always passes this block the right number of samples.
gr::filter::kernel::fft_filter_fff::fft_filter_fff | ( | int | decimation, |
const std::vector< float > & | taps, | ||
int | nthreads = 1 |
||
) |
Construct an FFT filter for float vectors with the given taps and decimation rate.
This is the basic implementation for performing FFT filter for fast convolution in other blocks (e.g., gr::filter::fft_filter_fff).
decimation | The decimation rate of the filter (int) |
taps | The filter taps (vector of float) |
nthreads | The number of threads for the FFT to use (int) |
|
delete |
|
default |
int gr::filter::kernel::fft_filter_fff::filter | ( | int | nitems, |
const float * | input, | ||
float * | output | ||
) |
Perform the filter operation.
nitems | The number of items to produce |
input | The input vector to be filtered |
output | The result of the filter operation |
unsigned int gr::filter::kernel::fft_filter_fff::ntaps | ( | ) | const |
Returns the number of taps in the filter.
int gr::filter::kernel::fft_filter_fff::nthreads | ( | ) | const |
Get number of threads being used.
|
delete |
|
default |
void gr::filter::kernel::fft_filter_fff::set_nthreads | ( | int | n | ) |
Set number of threads to use.
int gr::filter::kernel::fft_filter_fff::set_taps | ( | const std::vector< float > & | taps | ) |
Set new taps for the filter.
Sets new taps and resets the class properties to handle different sizes
taps | The filter taps (complex) |
std::vector<float> gr::filter::kernel::fft_filter_fff::taps | ( | ) | const |
Returns the taps.