Mercurial > hg > qm-dsp
diff tests/TestPhaseVocoder.cpp @ 355:3c7338aff6a8
Drop in kissfft to replace the "old" fft, and add tests for newly-supported sizes
author | Chris Cannam <c.cannam@qmul.ac.uk> |
---|---|
date | Tue, 15 Oct 2013 11:38:18 +0100 |
parents | 04d134031a15 |
children | 2de6184b2ce0 |
line wrap: on
line diff
--- a/tests/TestPhaseVocoder.cpp Tue Oct 15 09:39:58 2013 +0100 +++ b/tests/TestPhaseVocoder.cpp Tue Oct 15 11:38:18 2013 +0100 @@ -164,14 +164,16 @@ // phase of -pi/2 plus advance of 2pi). It has the same measured // phase as last time around, -pi/2, which is consistent with // the expected phase, so the unwrapped phase is 3pi/2. - //!!! - // * Bin 3 is a bit of a puzzle -- it has an effectively zero - // magnitude but a non-zero measured phase. Spectral leakage? + // + // * Bin 3 I don't really know about -- the magnitude here is 0, + // but we get non-zero measured phase whose sign is + // implementation-dependent // // * Bin 4 (Nyquist) has expected phase 4*pi, measured phase 0, // hence error 0 and unwrapped phase 4*pi. - double phaseExpected1[] = { 999, 0, -M_PI, -M_PI/2, M_PI, 0, 999 }; + phase[1+3] = 0.0; // Because we aren't testing for this one + double phaseExpected1[] = { 999, 0, -M_PI, -M_PI/2, 0, 0, 999 }; COMPARE_ARRAY(phase, phaseExpected1); double unwExpected1[] = { 999, 0, M_PI, 3*M_PI/2, 3*M_PI, 4*M_PI, 999 };