Overview

Vector convolution

Discussion

This module convolves the signals on its input pins a block at a time. It differs from the FIR module because it doesn't maintain any history between blocks. The output is computed by $s[n] = \sum_{k=0}^{M-1} x[n-k] y[k]$, where $x[n]$ is on pin 1 with length $M$ and $y[n]$ is on pin 2, also of length $M$. This operation could be generalized to allow different block sizes if needed.

The parameters shape, skip and length allows control of the range of $n$ that gets calculated. Shape is for quick setting of common ranges. shape = 0: sets skip=0, olen=2*M-1 to compute the full convolution shape = 1: sets skip=M/2, olen=M to compute the center part that has the same length as the inputs. shape = 2: leaves skip and olen to be user-defined. skip: starts storing output at $n = \mathit{skip}$ olen: writes only this many output samples, so $\mathit{skip} <= n < \mathit{skip} + \mathit{olen}$.

Type Definition

typedef struct _ModuleConvolve
{
    ModuleInstanceDescriptor instance;            // Common Audio Weaver module instance structure
    INT32 shape;                                  // Shape of subsection
    INT32 skip;                                   // Number of samples of full convolution to skip at output
    INT32 olen;                                   // Length of output
} ModuleConvolveClass;

Variables

Properties

Pins

Input Pins

Name: in0

Description: Input signal 0

...

Sample rate range: Unrestricted

Complex support: Real

Output Pins

Name: out

Description: Convolution output

Data type: float

MATLAB Usage

File Name: convolve_module.m

...