View Single Post
Old 20 August 2019, 07:03 am   #2
Shy's Avatar
Join Date: Jul 2004
Posts: 372

Originally Posted by Marillion78 View Post
Please give more information about using of --nmt --tmn --pns x switches.

'--longhelp' shows:

--nmt x set NMT value to x dB (dflt: 6.5)
--tmn x set TMN value to x dB (dflt: 18.0)
--pns x set PNS value to x dB (dflt: 0.0)

And what is the full able range of 'x dB' values (minimum and maximum) for these swithes?

One person recommended me in case of using '--quality 10' switch to add '--nmt 18 --tmn 28' for more better quality, and it this the best possible quality, or any other values will give more better quality? And what about '---pns' switch?

I also found that in '--longhelp' manual swithes '-xlevel' and '--noxlevel' are abscent. Are there any other forgotten switches in '--longhelp' manual? Is there full detailed list of swithes anywhere?
xlevel and noxlevel were removed with the release of SV8. They're not needed, since there's no internal clipping anymore. (this was mentioned in the changelog, by the way)

There's no independent "best possible quality" when talking about a perceptual, lossy audio format, there's only "best possible quality per defined bit-rate range", and the definition of "quality" in the case of a lossy format is "how imperceptible from the source the result is at a given bit-rate" - which depends on many factors, and "absolutely imperceptible", which is usually the result that Musepack provides, could easily count as "best possible quality", but nevertheless, it's not independent. If you want a truly independent and absolute best possible quality, you need to use a lossless format. I don't remember the minimum and maximum for those 3 parameters right now, but the min and max that are used in the quality settings are: "--quality 0" - PNS: 1.09, NMT: -1, TMN: 3, "--quality 10" - PNS: 0, NMT: 14, TMN: 33.

Those and various other parameters are not meant to be touched at all by any user to "improve the quality", the optimal quality balance for each quality preset is already provided by the presets. It had been reached after years of serious, extensive testing. You can mess with stuff like NMT (noise masks tone ratio) and TMN (tone masks noise ratio) if you feel like it, but if by chance you get a bit better quality with some audio by changing the defaults of --quality 5 for example (while your bit-rate increases, of course), don't expect this setting to be good for most other things you encode, it could be (and often is) actually worse than the default, and there are very many ways to unexpectedly ruin the sound with this encoder just by changing one setting a tiny bit, and you could notice it later and regret that you even touched it.

Simply, there's usually a very high chance that using --quality 6 for example would give you better quality at a similar average bitrate to the one you got with the manually "tweaked" --quality 5.

Ask yourself whether you actually need a "better" setting than a setting that already has absolutely no perceptible difference from the original audio. Good luck noticing any artifact whatsoever at --quality 10.
Shy is offline   Reply With Quote