minidsp_fir.c File Reference

FIR filtering and convolution: direct time-domain methods and FFT overlap-add fast convolution. More...

#include "minidsp.h"
#include "minidsp_internal.h"

Go to the source code of this file.

Functions
static unsigned	next_pow2 (unsigned n)
	Return the next power-of-two >= n (n must be > 0).
unsigned	MD_convolution_num_samples (unsigned signal_len, unsigned kernel_len)
	Compute the output length of a full linear convolution.
void	MD_convolution_time (const double *signal, unsigned signal_len, const double *kernel, unsigned kernel_len, double *out)
	Time-domain full linear convolution (direct sum-of-products).
void	MD_moving_average (const double *signal, unsigned signal_len, unsigned window_len, double *out)
	Causal moving-average FIR filter with zero-padded startup.
void	MD_fir_filter (const double *signal, unsigned signal_len, const double *coeffs, unsigned num_taps, double *out)
	Apply a causal FIR filter with arbitrary coefficients.
void	MD_design_lowpass_fir (double *coeffs, unsigned num_taps, double cutoff_freq, double sample_rate, double kaiser_beta)
	Design a Kaiser-windowed sinc lowpass FIR filter.
void	MD_convolution_fft_ola (const double *signal, unsigned signal_len, const double *kernel, unsigned kernel_len, double *out)
	Full linear convolution using FFT overlap-add (offline).

Detailed Description

FIR filtering and convolution: direct time-domain methods and FFT overlap-add fast convolution.

Definition in file minidsp_fir.c.

Function Documentation

MD_convolution_fft_ola()

void MD_convolution_fft_ola	(	const double *	signal,
		unsigned	signal_len,
		const double *	kernel,
		unsigned	kernel_len,
		double *	out )

Full linear convolution using FFT overlap-add (offline).

Produces the same output as MD_convolution_time() but is faster for longer kernels by processing blocks in the frequency domain.

Parameters

signal	Input signal of length signal_len.
signal_len	Number of input samples (must be > 0).
kernel	FIR kernel of length kernel_len.
kernel_len	Number of FIR taps (must be > 0).
out	Output buffer of length signal_len + kernel_len - 1.

Definition at line 124 of file minidsp_fir.c.

MD_convolution_num_samples()

unsigned MD_convolution_num_samples	(	unsigned	signal_len,
		unsigned	kernel_len )

Compute the output length of a full linear convolution.

For input length N and kernel length M, full convolution length is N+M-1.

Definition at line 18 of file minidsp_fir.c.

MD_convolution_time()

void MD_convolution_time	(	const double *	signal,
		unsigned	signal_len,
		const double *	kernel,
		unsigned	kernel_len,
		double *	out )

Time-domain full linear convolution (direct sum-of-products).

Computes: out[n] = sum_{k=0}^{kernel_len-1} signal[n-k] * kernel[k] with out-of-range signal samples treated as zero.

Parameters

signal	Input signal of length signal_len.
signal_len	Number of input samples (must be > 0).
kernel	FIR kernel of length kernel_len.
kernel_len	Number of FIR taps (must be > 0).
out	Output buffer of length signal_len + kernel_len - 1.

Definition at line 25 of file minidsp_fir.c.

MD_design_lowpass_fir()

void MD_design_lowpass_fir	(	double *	coeffs,
		unsigned	num_taps,
		double	cutoff_freq,
		double	sample_rate,
		double	kaiser_beta )

Design a Kaiser-windowed sinc lowpass FIR filter.

Generates a linear-phase lowpass filter with the specified cutoff frequency and Kaiser window shape. The coefficients are normalized to unity DC gain (they sum to 1.0).

\[ h[i] = 2\,f_c\;\mathrm{sinc}\!\bigl(2\,f_c\,(i - (N-1)/2)\bigr) \;\cdot\; w_{\mathrm{Kaiser}}[i] \]

where \(f_c = \mathrm{cutoff\_freq} / \mathrm{sample\_rate}\).

Parameters

coeffs	Output buffer of length num_taps (caller-allocated).
num_taps	Filter length (must be > 0).
cutoff_freq	-6 dB cutoff frequency in Hz (must be > 0 and < sample_rate / 2).
sample_rate	Sampling rate in Hz (must be > 0).
kaiser_beta	Kaiser window \(\beta\) parameter (e.g. 10.0).

double h[65];
MD_design_lowpass_fir(h, 65, 4000.0, 48000.0, 10.0);  // 4 kHz LPF

Definition at line 88 of file minidsp_fir.c.

MD_fir_filter()

void MD_fir_filter	(	const double *	signal,
		unsigned	signal_len,
		const double *	coeffs,
		unsigned	num_taps,
		double *	out )

Apply a causal FIR filter with arbitrary coefficients.

Computes: out[n] = sum_{k=0}^{num_taps-1} coeffs[k] * signal[n-k] with out-of-range signal samples treated as zero.

Parameters

signal	Input signal of length signal_len.
signal_len	Number of input samples (must be > 0).
coeffs	FIR coefficients of length num_taps.
num_taps	Number of FIR taps (must be > 0).
out	Output buffer of length signal_len.

Definition at line 68 of file minidsp_fir.c.

MD_moving_average()

void MD_moving_average	(	const double *	signal,
		unsigned	signal_len,
		unsigned	window_len,
		double *	out )

Causal moving-average FIR filter with zero-padded startup.

Computes: out[n] = (1/window_len) * sum_{k=0}^{window_len-1} signal[n-k] where out-of-range samples (n-k < 0) are treated as zero.

Parameters

signal	Input signal of length signal_len.
signal_len	Number of input samples (must be > 0).
window_len	Moving-average window length (must be > 0).
out	Output buffer of length signal_len.

Definition at line 48 of file minidsp_fir.c.

next_pow2()

unsigned next_pow2 ( unsigned n )

static

Return the next power-of-two >= n (n must be > 0).

Definition at line 11 of file minidsp_fir.c.