DST transform in Matlab Vs FFTW3

Question

I am porting a Matlab/Octave code to C. One of the implementations I need to port is the DST transform. In Matlab/Octave, this is done using

A =[0.00000   4.24264   5.65685   7.07107   8.48528   0.00000]
res = dst(A)
ans =
22.3599   -4.7499   -4.4287    2.9465   -6.8864    1.9383

In C, I have done what the FFTW3 documentation says:

n = 6;
fftw_plan plan = fftw_plan_r2r_1d( n, in, out, FFTW_RODFT10, FFTW_ESTIMATE);
fftw_execute ( plan );

where in is the input array. The result of this is, out:

 42.588457
-9.899495
 0.000000
 2.449490
-11.411543
 5.656854

Why are those results so different? Any ideas of what I am doing wrong here, if any?

Answer 1

If you are looking for exactly the same tranform, you may try the RODFT00 flag instead.

According to docs of fftw, matlab, octave, it seems to fit, except for a 2 factor. Output of fftw may be twice the one of matlab or octave.

http://www.mathworks.fr/fr/help/pde/ug/dst.html

http://octave.sourceforge.net/signal/function/dst.html

http://www.fftw.org/doc/1d-Real_002dodd-DFTs-_0028DSTs_0029.html

Bye,

Answer 2

tried this x=[ 0 -0.041389004581424 -0.049979175343607 -0.057007496876302 -0.062473969179509 -0.066378592253228 -0.068721366097459 -0.069502290712203 -0.068721366097459 -0.066378592253228 -0.062473969179509 -0.057007496876302 -0.049979175343607 -0.041389004581424 0.500000000000000 -0.500000000000000]

matlab's dst gives : dst(x)=[-0.524649937006448 -0.208549082037091 0.175438242940290 -0.201365706563574 0.122632326416342 0.023410695189363 -0.170916792532103 0.404635063593086 -0.557336293807673 0.766827077578399 -0.834391996423464 0.916008576298954 -0.836787527646708 0.743037852780101 -0.522660424688284 0.285317021943821]

using FFTW with "FFTW_RODFT00" flag gives [-1.0493 -0.417098 0.350876 -0.402731 0.245265 0.0468214 -0.341834 0.80927 -1.11467 1.53365 -1.66878 1.83202 -1.67358 1.48608 -1.04532 0.570634]

Basically a factor of two is the difference, even if one uses same flag