Skip to main content
Topic: Secret Rabbit Code resampler (Read 3587 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Re: Secret Rabbit Code resampler

Reply #25
Thanks Ivqcl!  Seeing as I only resample if the original rate is not compatible with my device (which means a few odd radio streams at 22050;24000;32000;88200) that means I'll stick with SoX.  Yes that is cool though how Case added a solution to SRC for gapless playback issues.

Re: Secret Rabbit Code resampler

Reply #26
Love this resampler, SMOOTHING the song while increasing clarity, other resamplers lack this property. Just want to give some feedback.
I use Cheat Engine to minimize the buffer length to 0ms. When I use other resamplers like SoX or Resampler-V there is no issue, but when I use SRC 1.0.8, some glitches happen just after switching songs. Guessing the plugin is not fast enough, I hope the resampler may be faster. BTW the "exclude rates" may be slow as well, by manipulating the order of DSPs I can hear the differences.


Re: Secret Rabbit Code resampler

Reply #28
from Kohlrabi's signature: "It's only audiophile if it's inconvenient."


Re: Secret Rabbit Code resampler

Reply #30
I use Cheat Engine to minimize the buffer length to 0ms.
:o  Why?
It doesn't matter anyway, as with zero buffer, it wouldn't play at all. So, it programmatically enforces its own minimum buffer size, regardless of what the setting says. Anyone with half a brain knows that the magical "zero" latency is physically impossible. Or maybe someone thinks that throwing around the word "quantum" would magically make it possible to have literally instant output without any buffering or transit in between.

Re: Secret Rabbit Code resampler

Reply #31
extrapolation used in Vorbis and Opus
Is this crosslappping, which is described here - https://xiph.org/vorbis/doc/vorbisfile/crosslap.html?
And, @Case , is there any reason why you didn't publish foo_dsp_amp and foo_hardclip on you site? Also, something about components themselves.
In foo_dsp_amp it is impossible to set some values, for example +0.7 dB and +1.4 dB (or maybe this is just GUI which displays wrong values). foo_hardclip doesn't work without SSE2 (verified on real AthlonXP). Is this by intention?

 

Re: Secret Rabbit Code resampler

Reply #32
Is this crosslappping, which is described here - https://xiph.org/vorbis/doc/vorbisfile/crosslap.html?
Overlapping would require information from two tracks to get the transition smoothed. When the DSP is used in multi-threaded converter there's only information from a single track available. The component uses extrapolation to estimate how the signal was before the track start and how it would continue at the track end. The extrapolated samples are included in the signal fed to the resampler and cut away to retain proper length.

is there any reason why you didn't publish foo_dsp_amp and foo_hardclip on you site?
Just laziness. Putting them there requires for example writing some description texts.

In foo_dsp_amp it is impossible to set some values, for example +0.7 dB and +1.4 dB (or maybe this is just GUI which displays wrong values). foo_hardclip doesn't work without SSE2 (verified on real AthlonXP). Is this by intention?
The first issue was a rounding problem. The second was carelessness. I should know to check settings when upgrading project files to new Visual Studio but apparently made a mistake that year.

Fixed components are up on my site with dynamic 1.4 versions and larger legacy 1.1 versions.

 
SimplePortal 1.0.0 RC1 © 2008-2018