INTERACT FORUM
More => Old Versions => JRiver Media Center 20 for Windows => Topic started by: TheLion on November 21, 2013, 10:09:17 am
-
Dear Matt,
some time ago we discussed the possibility of integrating some optional resampling algorithms to choose from. You have asked me to try some and present you with suitable options to compare to JRiver internal SRC. The challenge was to find an openly available library which is suitable for integration in MC.
I am of the opinion that SRC is one of the few fields MC19 can objectively can be "made better regarding sound quality" - bit perfect is bit perfect, but once you start to accept processing audio you can improve on the alogrithms used. I have done lot's of SRC comparisons - as references I used Adobe Audition and Weiss Saracon. There are a few very satisfying options available - a readily available "free" library providing excellent results is the SoX 14.4 resampler - especially in VHQ Linear Phase mode.
If you still consider this an interesting feature please have a look and try it:
http://sox.sourceforge.net/SoX/Resampling
http://vladgsound.wordpress.com/2011/11/12/using-sox-resampler-in-plugins/
Thank you very much! Anybody else having experience with SoX or other available SRC libraries please share your findings vs. JRiver Resampling.
-
Our resampler is based on Shibatch's SSRC. It's well respected and instrument tests really well.
I'm not eager to chase other options unless there's some clear technical benefit.
If you're interested, you can apply resampling during conversion. This means you can feed tones, sweeps, etc. through it and analyze the output.
-
Anybody else having experience with SoX or other available SRC libraries please share your findings vs. JRiver Resampling.
I can't say anything about SOX vs. MC here, but I can say that in Whitebear Media Server I use SOX to down sample High Definition audio tracks down to L16 (linear pcm, 16bit, 2channel, 44100/48000Hz) when the target renderer cannot handle anything better, and I can say that it certainly sounds Ok, and is certainly fast enough to do transcoding on the fly on a mid range computer.
-
Our resampler is based on Shibatch's SSRC. It's well respected and instrument tests really well.
I'm not eager to chase other options unless there's some clear technical benefit.
If you're interested, you can apply resampling during conversion. This means you can feed tones, sweeps, etc. through it and analyze the output.
Thanks for the reply, Matt. SSRC (in high precision mode!) does a good job in my experience. Clear technical benefits over SSRC H P are not really to be expected (see http://src.infinitewave.ca/ for example), it's more a matter of different "flavors". Different SRC algorithms do provide significant changes - it is rather subjective which one is preferred.
If you ever have the interest to make this more flexible in MC adding the SoX SRC library should be possible with little effort/hassle and is a very decent choice. Thanks!
-
+1 for SOX option
-
+1
I'm not eager to chase other options unless there's some clear technical benefit.
Pre/post ringing of various SoX options compared with SSRC:
(http://abload.de/img/rate-44k1-96kkiu2u.png)
Even if you don't want to offer SoX as an option, changing to SSRC High Precision would be beneficial - though SoX still appears to perform better on that SRC comparison site.
SSRC is generally considered to be fine for downsampling, but not the best option for upsampling.
-
+1
-
+1 for SoX in JRiver.
that technical graph on SoX looks excellent with reducing/eliminating the pre-ringing.
-
Izotope SRC looks even better ;D but I think it requires a license....
-
+1 for SoX
-
+1 for SoX in JRiver.
that technical graph on SoX looks excellent with reducing/eliminating the pre-ringing.
That's just one aspect. The hard stop on the ringing causes it's own problems elsewhere (aliasing on the frequency response)
This is an old link and I wouldn't call it accurate today but it does show graphically that there is more to a re sampler than ringing response (I just happened to have it bookmarked but there are some newer links out there). The honest truth is that if measured (and although I've never seen an ABX, I'll bet I know the answer) sox and ssrc are pretty much transparent
http://www.mainly.me.uk/resampling/
Not saying that choice isn't great, mind you...
-
In case anyone missed it, the resampler was changed to use high precision mode for audio only playback.
20.0.67 (2/6/2015)
8. Changed: Switched the resampler to high precision mode (testing appreciated).
20.0.69 (2/11/2015)
3. Changed: Made SRC only use the high precision mode for audio playback (because it was causing timing issues for video playback).
-
That's just one aspect. The hard stop on the ringing causes it's own problems elsewhere (aliasing on the frequency response)
This is an old link and I wouldn't call it accurate today but it does show graphically that there is more to a re sampler than ringing response (I just happened to have it bookmarked but there are some newer links out there). The honest truth is that if measured (and although I've never seen an ABX, I'll bet I know the answer) sox and ssrc are pretty much transparent
http://www.mainly.me.uk/resampling/
Not saying that choice isn't great, mind you...
Agree, it is always a compromise, hence various resamplers out there!
In my personal opinion of listening to various resamplers- A resampler of high quality, eliminating/reducing pre-ringing with low aliasing sounds much better than a high quality resampler of very little aliasing, but with pre-ringing.
A sound coming before a sound, or Pre-ringing is unnatural to the ear. smears the audio.
-
+1 for SoX
-
+1 for SoX
Are you sure you are not a SoX puppet ? ;D
-
I use SOX to upsample offline for my favorites with a script to control the parameters based on sample rate :D Hard to say if it sounds better because it is done offline or 'cause it is SOX ;)
-
+1 for SoX with intermediate phase settings (i need to control pre-ringing)
-
Sox is available in MC22. You can download from the MC22 board.