I want to play MP3 @ 32 bit output, but DTS WAV must be output @ 16. Any idea to avoid change output reslolution every time I chage file format?

May be a dialog to associate bit depths and file types?

Thanks.

But why on earth would you want to do that??!

I mean, aren't you mp3s ripped from 16bit cds anyway? I suppose they could be encoded from 24/32 bit PCMs, but by that you're taking the format out of spec.

If you want your compressed music to sound better using ogg vorbis will make waaay more difference than upsampling mp3.

It´s very easy: I use Kernel Streaming to send to my amplifier decoded PCM data from any encoded file, cause I can ear the diference. If decoded signal is @64 bit (actually for MP3 is 32) the sound must be truncated or dithered to send to output @16 bit, with quality loss.

WAV files aren´t encoded so PCM can be passed directly. A check button as "output WAV files at its bit depth" could be useful.

If WAV is @16, output is 16.
If WAV is @32, output is 32.
...

I guess this has some relevance, so here goes..

MP3 is not an integer format. Whether you encode a 16bit integer PCM wave or a 32bit integer PCM wave, mp3 encoding will result in a floating point mp3 file. Even if the mp3 source was a 16bit integer source (e.g. a CD), you can benefit from decoding mp3 to 24bit or 32bit integer PCM if your soundcard can handle it properly.

It's the same with Vorbis too.

Thanks, sxz, I suspected mp3 to be floating point. I'm still dubious about the advantages of converting fp to >16bit depth integers.

I think I'm safe in saying that (HDCD aside) mp3s ripped from CD can never exceed the quality of it's source - 16bit audio. It's a mater of getting it as close as possible, I'm extremly skeptical that playing a compressed format at 24bit brings it closer to the 16bit original. If the mp3 is ripped from DVD-A there may be soe benefit.

Then again, I can't personally tell 15 from 16 bit in a ABX test, and the jump from 15 to 16 is waaaay more than from 16 to 32. Maybe you have much better hearing than I, but I consider myself to have good ears and have a decent quality hifi. Have you tried comparing 16,24,32 bit mp3 playback in the foobar ABX comparator?

I'm aware of the benefits of upsampling 16bit audio as a DAC tehnique, but that is quite a different thing to sending the DAC raw 24bit audio.

Having said all that, it certainly won't harm the music and if your hardware supports it, why not? In fact, if your DAC is up to it why not just keep the bitrate constantly at 32bit - the increase in data shouldn't trouble any vaguely modern PC.

Hell, playing an mp3 adds barely 1% to my cpu usage!

Jim

@j-bo


I´m agree with you. But, it does not imply this:


32bit MP3 decoded data isn´t better than its 16bit original source. But the problem is that Fb2k decode MP3 to 32bit floatong-point(actually), and, converting that to 16bit, logicaly, we are loosing more and more quality.

Note: 24bit can´t be played through excellent Kernel Streaming.

Note2: With KS I can´t play 64bit. Only 16 and 32 bit.

I suggest that you perform a simple test.

First, decode an mp3 file using "WAV (PCM 16 bit undithered)" setting in the diskwriter.

Then, open the resulting wav and the original mp3 in foobar2000, set output format in playback preferences to 32 bits and see if you can tell them apart in an ABX test.

If you can't, just use 16 bit undithered output for all of your files.

I will do ABX test. But, if quality change between 16 and 32 is so insignificant,

Why does exist "Dither" option?

Probably for those with uber-ears, living in an uber-quiet environment, and with uber-equipment hooked up to their soundcard. I personally have never heard a difference, with my environment as noisy as it is, and with $100+ speakers and $30 headphones.

Because we're working with files without a defined noise floor (mathematically transformed, ie MDCT, subband), converting them to something with a defined noise floor (PCM). Dither is a technique that alters and lowers both the noise floor where we'd be sensitive to it, as well as eliminating aliasing artifacts.

It's more-or-less purely technical, and it's not proven that it's better, but it's available, especially if you need to do something with the output after it's decoded (ie. remixing, editing). In these situations, theoretically inaudible noise floors can creep up to audibility sometimes.

Chances are very good you won't be able to hear the difference dither makes. That's why, IIRC, it's disabled in the default fb2k configuration. Purely technical. And yet, it's the attention to little technical details like this that makes fb2k so great.

Not to mention the level of overhead that noise-shaping adds, especially ATH noise shaping. For fun, try the Speed Meter with float to linear conversion, with and without dithering. The dithering mode setting in playback preferences is global and also controls which mode is used by the speed meter.

Reading this thread, output bit depth above 16bit for MP3's seems makes sense.

But does it make any sense for a 16 bit wav file from a CD? Surely an output bit depth above 16 does not make any sense at all then. Except for cases where the soundcard would've converted to 24 anyway.

I started a similar thread here:

Confused.

If you're not using any DSPs (and that includes volume control and replaygain), it doesn't.

Otherwise, it does. Although you may not be able to hear the difference, of course.

You have an odd definition of fun, kode54.

Correction, to see what I mean about overhead. For best results, use a format which decodes really fast.

Well I can imagine how 24 bit depth could help with DSPs, even for 16 bit files - any rounding errors would be at bits so insignificent so as to be inaudiable.

Having said that, I can't hear the difference anyway. That's on a near silent PC with outboard dac connected to a nice pair of Sennheisers!

If you want 24 bit output for a few files, why not just keep it at 24bit for the whole time?

If you use RG >16 bits are useful.

Thanks to all.

There is a little problem, I will try to explain me:


1.-I understand the mathematical sense of Dither and why it´s so difficult to hear the difference.

2.-All dithers are not the same. When I have to export MP3 to CDs, dither (such Waves dither) makes sense. I can hear the difference in car & home CD players, but ABX isn´t there.

3.-In Fb2k @ 16bit, with Kernel Streaming output, I don´t need to do ABX test, cause I prefer Fb2k undithered output.


4.-But, why downconvert 32bit decoded signal to 16bit if Kernel Streaming supports 32bit?

Moneo said:


I did that ABX test with a normal MP3 and Kernel Streaming, and I didn´t obtained conclusive results. Guess probability wasn´t low. The test wasn´t slow and accurate, but, I didn´t repeat the test with others MP3 because there was a little problem:

In the test I was playing 16bit undithered @32bit, and the test should be done @16bit output. The problem could be solved excluding to WAV files of pipeline, outputting them directly, if it´s posible @ its bit depth (32/64bit floating point are not allowed with my REVO card).


The amplifier (Onkyo 595), wich is connected through SPDIF, logically is prepared to MiniDisc which is @32bit. So, may be my 32bit DAC is better than 16bit DAC?

I think that it isn´t a good idea post large replies like mine, but, for novices, it´s difficult to resume the main concept. Don´t you think so?