Audio/filterbanks/render.cpp

Using optimised filter banks

This example creates a bank of filters using the QuadBiquad class.

The QuadBiquad class is an extension of the Biquad filter class which allows you to process four channels in parallel. To learn more about Biquad filters and how to use the Biquad library, make sure to check Nigel Redmon's fantastic blog: https://www.earlevel.com/main/category/digital-audio/filters/iir-filters/biquads/

In this example we create 150 instances the QuadBiquad filter using bs(150). These instances are connected in series to create some very very long filterbanks, each of which contains lowpass and highpass filters.

In setup() we set the type of filters and their cutoff frequency.

As each QuadBiquad object contains four filters (in parallel), which are completely independent of each other you can set each to a different type of filter, cutoff, Q, etc. Here we give to each of these four filters an increasing cutoff. This will result, in the right channel being brighter than the left channel.

We then process (up to) four audio inputs through the filterbanks and output the result through all available output channels.

/*
 ____  _____ _        _
| __ )| ____| |      / \
|  _ \|  _| | |     / _ \
| |_) | |___| |___ / ___ \
|____/|_____|_____/_/   \_\
http://bela.io
*/
 
#include <Bela.h>
#include <vector>
#include <libraries/Biquad/QuadBiquad.h>
 
std::vector<QuadBiquad> bs(150);
 
bool setup(BelaContext *context, void *userData)
{
        BiquadCoeff::Settings s = {
                .fs = context->audioSampleRate,
                .q = 0.8,
                .peakGainDb = 0,
        };
        // Setting cutoff and type for all the filters
        for(unsigned int b = 0; b < bs.size(); ++b)
        {
                if(b < bs.size() / 2) {
                        // the first half of the filters are lowpass
                        s.type = BiquadCoeff::lowpass;
                        s.cutoff= 2000;
                } else {
                        // the other half are highpass
                        s.type = BiquadCoeff::highpass;
                        s.cutoff = 500;
                }
                for(unsigned int q = 0; q < 4; ++q)
                {
                        // Each QuadBiquad object contains four filters (in parallel), which
                        // are completely independent of each other and you could set each
                        // to a different type of filter, cutoff, Q, etc
                        // Here we give to each of these four an increasing cutoff.
                        // This will result, for instance, in the right channel being
                        // brighter than the left channel.
                        bs[b].filters[q].setup(s);
                        s.cutoff *= 1.3;
                }
                // after updating the filter coefficients of the individual
                // filters, tell the QuadBiquad object to refresh its internal
                // registers
                bs[b].update();
        }
        return true;
}
 
void render(BelaContext *context, void *userData)
{
        unsigned int len = bs.size() & (~3);
        for(unsigned int n = 0; n < context->audioFrames; ++n)
        {
                const unsigned int datasz = 4;
                float data[datasz];
                // the data array is used for input and output of the four parallel filters in each QuadBiquad
                // we fill it with however audio inputs we have
                for(unsigned int c = 0; c < context->audioInChannels && c < datasz; ++c)
                        data[c] = audioRead(context, n, c);
                // we then process it through all the filters in series
                for(unsigned int b = 0; b < len; ++b) {
                         bs[b].process(data);
                         // after each call to process(), data contains the output of the filter.
                         // by processing it over and over, we effectively place the filters in series.
                }
                // last, we output the results to however many audio channels we have
                for(unsigned int c = 0; c < context->audioOutChannels && c < datasz; ++c)
                        audioWrite(context, n, c, data[c]);
        }
}
 
void cleanup(BelaContext *context, void *userData)
{}