I want to convert a bunch of .wav files from 24Bit to 16Bit so I can compress them more efficiently with FLAC. What's the best was to do this (quality wise), and will there be _any_ audible difference? Some one said that I should do it with Foobar2000 becuase it has a nice dither. Is that true? I do have access to more professional gear at work (Waves plugins etc), but there's no need to do that if Foobar can do the job just as well.

Tony

No, there will be no audible difference if all you are going to do to the files is listen to them. (Yes, I know this is a contentious issue, but there really is unlikely to be an audible difference). If you are using the files for mixing or other processing, then keeping the 24bit is worth while.
Foobar does have nice dither features. Dither, as you may know, reduces the noise floor in the audible band by moving the noise energy up into high frequencies. This probably won't be audible in most environments, but is still worth doing (for listening, if you are going to apply processing later then don't dither).

Dithering and noise shaping are two different things. Dithering makes sure, that the error is not correlated to the signal. Noise Shapings spectrally shapes the noise (moving it to places we are less sensitive to).

The standard conversion dithering in foobar uses noise shaping and adds inaudible noise to the high frequency area. The extra noise is clearly visible on a spectrum analyzer display at over 18 kHz. This has a slight negative effect to the compression ratio that can be achieved with lossless encoders and also lossy encoders with the highest quality settings.

I tried the bitrate behavior with an uncompressed 24/44.1 test track:

conversion to 16/44.1 without dithering vs dithered with fb2k 0.83 default dithering

FLAC 5: 653 kbps vs 685 kbps
LAME V0: 222 kbps vs 227 kbps
Vorbis Q7: 200 kbps vs 207 kbps

You may want to try to ABX the effect of dithering. Foobar has fine dithering, but I guess you cannot hear any difference.


Edit: changed the first sentence a bit after Sebastian's reply.

This might be a "contentious issue" but mainly because not everyone is looking in the right place. Dither only effects the least significant bit so it has no effect, audible or otherwise, on most of the audio we can hear. Its purpose is to reduce distortion and extend the dynamic range. It has been demonstrated to make a very measurable improvement on very low level signals, but until the audio gets down to -90dB and lower, dither is just a little extra noise.

Ok, since my recordings have little dynamic range (they are from vinyl), I will probably not dither and just leave them at 24Bit.

I am going to normalise some of them, so is there any steps I need to take before / after I do this? Or do I just hit 'normalise' then save?

And thanks for all the input so far, I learn a LOT of very useful stuff here and it help me everyday. It's good to know there's a forum where people argue facts and figures rather than belief and hearsay.

Tony

Just to nitpick a little: recordings made from vinyl typically have *more* dynamic range than modern CDs, (due to the incompetent mastering currently in vogue). This is why a vinyl recording sounds quieter than a CD rip, even after normalisation. What vinyl has is a high *noise floor*, which effectively self-dithers. I frankly can't see any point in adding dither when converting 24 to 16 bit if the source was vinyl. (But then I can't see any point in recording vinyl at higher than 16 bit in the first place, unless you plan on applying loads of DSP operations (like 20 or more passes), in which case the rounding errors might conceivably build up enough to encroach into the vinyl noise floor).


No, just hit "normalise" - all it does is apply a linear gain so that the highest peak hits a defined maximum. You might consider setting the maximum to slightly less than 0dB, because some DACs and lossy compressers can clip when faced with 0dB inputs. I usually normalise to -0.1dB.

Hmm, don't you measure dynamic range against a noise floor? I mean, regardless of what music / sound is pressed to a vinyl or CD, a CD still has a greater dynamic range possibility in it's 16Bits than a vinyl will ever have... ? No?

And thank for the help ;-)

Correct. The CD format has greater dynamic range *possibility* than vinyl can ever have. But in practice, virtually every pop/rock CD mastered since about 1990 hasn't taken advantage of this possibility. Most classical CDs are still done properly.