Has someone tested the cheby option of soxr lib ?

[lyapounov]

Hi friends

has someone tested the cheby option of soxr lib ?

The parameter in streams.options is the following

convOut=-af aresample=resampler=soxr:precision=28:cheby=1

The impact is (according to ffmpeg web site) :
For soxr only, selects passband rolloff none (Chebyshev) & higher-precision approximation for ’irrational’ ratios

Chebyshev polynomial are supposed to provide better interpolation than zero padding plus passband filter.

(Simon, do you need to restart the server when you change this parameter ?)


Unfortunatly, I had to leave my top end equipment in France while I am temporarly in San Francisco, so I can't really test :-(

Cheers

[simoncn]

Changes to convOut take effect immediately but don't affect any streams currently being played.

[NickD]

I've started using cheby=1 but with precision=33 (that's maximum I think). I like the result so far, very clear and distortion free. I up sample everthing to 24/352.8 or 24/384.

[jaaptina]

I just tried. It works fine but I don't hear a difference. All music is transcoded to WAV24 96kHz and some equalisation is applied:
convOut={-af aresample=resampler=soxr:precision=28:cheby=1,equalizer=f=55:width_type=q:width=​20:g=-15,equalizer=f=170:width_type=q:width=20:g=-10}

[lyapounov]

Cheby=1 and precision=33 does definitively improve a lot !
I transpose to wav24:192 (the maximum the klimax accepts)

Simon+ffmpeg+soxr are the genious of digital music

[NickD]

(14-10-2017 06:26)lyapounov Wrote:  Cheby=1 and precision=33 does definitively improve a lot !
I transpose to wav24:192 (the maximum the klimax accepts)

Simon+ffmpeg+soxr are the genious of digital music

You might want to try adding :cutoff=1.00

Default value is 0.91 so you can experiment with values up to 1.0. Depending on the filtering in your dac it could make a difference.

Ultimately I have removed cheby=1 and replaced with cutoff=0.995