Mercurial > hg > js-dsp-test

    <html>
    <body>
    <p>If 2150 iterations of real-to-complex FFT of size 2048 takes less than 10 seconds, then we may be able to make a high quality real-time phase vocoder (just).</p>

<p>A phase-vocoder of course must use overlapped
windowed FFT (although you can choose the size, within limits), IFFT,
and cartesian-polar conversion to calculate the phase for the
instantaneous frequency.</p>

<p>A reasonable estimate of CPU cost for the whole thing is somewhere
around 10x the cost of simple non-overlapping short-time forward
Fourier transforms across the signal. </p>

<p>2150 iterations corresponds to 100 seconds of audio non-overlapped at
44.1kHz, so if that takes less than 10 second, then in theory we might
be OK.</p>
    <pre>
<script src="nayuki/fft.js"></script>
<script>

/* for a phase vocoder, we probably want 2048-point real-to-complex
 * FFTs (if available) */

function inputReals(size) {
    var result = new Array(size);
    for (var i = 0; i < result.length; i++)
	result[i] = (i % 20) / 10.0 - 1.0;
    return result;
}

function zeroReals(size) {
    var result = new Array(size);
    for (var i = 0; i < result.length; i++)
	result[i] = 0.0;
    return result;
}

function inputReal64s(size) {
    var result = new Float64Array(size);
    for (var i = 0; i < result.length; i++)
	result[i] = (i % 20) / 10.0 - 1.0;
    return result;
}

var iterations = 2150;
var size = 2048;

var start = performance.now();

var total = 0.0;

for (var i = 0; i < iterations; ++i) {
    var real = inputReals(size);
    var imag = zeroReals(size);
    transform(real, imag);
    for (var j = 0; j < size; ++j) {
	total += real[j] + imag[j];
    }
}

document.write("total = " + total + "<br>");

var end = performance.now();

document.write("nayuki fft.js: " + iterations + " iterations at size " + size + " took " + (end - start) + "ms (" + (1000.0 / ((end - start) / iterations)) + " iterations/sec)<br>");

</script>
<script src="fft.js/lib/complex.js"></script>
<script>

var fft = new FFT.complex(size, false);

start = performance.now();

total = 0.0;

for (var i = 0; i < iterations; ++i) {
    var ri = inputReal64s(size);
    var co = new Float64Array(2 * size);
    fft.simple(co, ri, 'real');
    for (var j = 0; j < 2 * size; ++j) {
	total += co[j];
    }
}

document.write("total = " + total + "<br>");

var end = performance.now();

document.write("nockert fft.js: " + iterations + " iterations at size " + size + " took " + (end - start) + "ms (" + (1000.0 / ((end - start) / iterations)) + " iterations/sec)<br>");

    </script>
</pre>
    </body>
author	Chris Cannam
date	Thu, 01 Oct 2015 16:56:01 +0100
parents
children	44f670784d5f