I finally decided to have my vinyl collection "digitized". A friend of mine compressed about 100 of my vinyls and now I have it in single wv files 24bit/192kHz so "it's as close to vinyl as possible" At least that's what I was told. When I use my PC to play it it's fine, but when I want to copy a bunch of em to my mobile I quickly run out of space. So... my idea is to resample albums to 44kHz whenever I want one on my mobile. That works fine and I get an acceptable file size but there's one thing I'm not sure about. Do I need to (I thing you call it) dither it down to 16bits as well or is it OK to have a wavepack 24/44khz or even an AAC 24/320kbps? Does it make any sense? I have XperiaZ3 if that's important. I know it can handle files up to 96kHz but I'm afraid it sounds as 44 to me.

It's perfectly fine to convert your 24 bit rips to 16 bit, there's still more than plenty enough dynamic range ("space" in the signal) for everything that's on an LP.

It's the same thing with resampling, a 44.1KHz sample rate is more than plenty to contain all the useful signal that's on an LP.

I'm afraid it sounds as 44 to me.

What is this supposed to mean?

It means 192 and 44 sound the same to me. I have read about HR releases and even bought a Led Zeppelin album but after a conversion to 44kHz it's the same to me.

It seemed as if you were saying 44 had a sound and that it was different from 192, perhaps because of the software or some other reason.  Thanks for the clarification.

Do I need to (I thing you call it) dither it down to 16bits as well or is it OK to have a wavepack 24/44khz or even an AAC 24/320kbps? Does it make any sense?

Yes, it's called dither, and it's generally recommended any time you convert to a lower bitdepth. However, since your source is ripped from LPs, the noise level should always be high enough that dither would not be necessary (it only matters at very low levels), but you will not be able to hear a difference either way.

You don't want to convert to 16-bit (or dither) if you encode to AAC. It won't make any difference in the file size, and in theory leaving the source 24-bit will give the AAC encoder more to work with.

Two possible considerations:
1) peak-normalize. In case you happened to be record at tens of dBs too low volume.
(Possibly, I don't know, the noise from the vinyl will hit full volume anyway?)

2) 44.1 vs 48. My DAC makes annoying noises when switching from one track at one samplerate to one at another. If most of your other music is CD, that is a case for 44.1. If most of your other music is 48, then you might want to consider that.

AAC 24/320kbps

There is no such animal.  AAC has no inherent bit depth as it relates to PCM.

You should have said AAC 320 sourced from 24-bit material if that was what you meant.

Thank you guys! I think I was not clear enough about one think. When I was talking about bit depth I had in mind output bit depth. When I convert 24/192kHz to xx/44kHz I simply use foobar2000 converter. So should I keep the bit depth as it is or (24/44kHz) or set it to 16 (16/44kHz)?

16 would generally suffice, and especially when the source is recorded from a medium that can not handle 96 dB dynamics.

But again: if the peak is at (say!) -60 dB, you need more volume.  (Volume is one of the reasons to use more bit depth when recording. And the RIAA eq. will attenuate the extreme treble by 20 dB and boost the bass by 20 dB - it is not a bad idea to record at an overkill bit depth (may not be needed, but why not?) and then convert afterwards.) That said, if the peak is really low, then I would suspect something wrong in the process (not used the phono input?). And I actually have no clue how much a pop/crackle can go above the "peak of the music" ... anyone?

And I actually have no clue how much a pop/crackle can go above the "peak of the music" ... anyone?

It would have to be greater than 8x before I would lose any sleep at the thought that I might hear quantization noise because of a loss of "precision" using 16 bits after peak-normalization.

I have a few more questions related to down sampling. I use FB2000+SoX and I'm not sure about some settings. My destination format is WV 16bit/44kHz. Source files are 24/192kHz. When it comes to conversion to WV I have set output bit depth to 16bit, dither is set to never.



 SoX re sampling settings:

• Target samplerate: 44100
• Quality: best
• Passband: 95,0%
• Allow aliasing/imaging: unchecked
• Phase response: 0%

Do you thing these are optimal settings? I was looking for information about aliasing/imaging and dither but looks like that's too complicated to me. Do I need it when I convert my records?

I recommend for SoX

passband 92 (or default 95 as second best recommendation)
aliasing on (checked)
phase linear (50 %)
best quality

or on command line rate -v  -b 92 -a 44100 dither
or as second best rate -v -a 44100 dither

It is just a suggestion based on subjective experience. Others experience may vary.

Do you thing these are optimal settings? I was looking for information about aliasing/imaging and dither but looks like that's too complicated to me. Do I need it when I convert my records?

TPDF dither or low noise shaped dither like low-shibata or modified-e-weighted (or eventually foobars internal dither if you do not use SoX command line) is recommended for converting to 16 bit.
Aliasing occurs with SoX only above the passband and if you would like to use gentler filtering it can be enabled. Personally I have enabled it on most conversions. If you do not like to have it enabled then I would suggest to use passband 92 for that option, not 95.

If you would like to experiment more you can try also finalcd latest version http://www.sonicillusions.co.uk/downloads.htm supports 192 kHz as input. Again I recommend using the option /f1 e.g. 192->44.1 Gentle Filter bm40 /d1 very light shaped TPDF dither.

For those who prefer steep brickwall filtering with higher ringing the SoX options -s can be used (passband 99 in the UI), FinalCD /f0 100 Hz transition works and also SSRC is by design very good steep filtering resampler.

No claims about audibility or audible quality are made/intended with the above.

TPDF dither or low noise shaped dither like low-shibata or modified-e-weighted (or eventually foobars internal dither if you do not use SoX command line) is recommended for converting to 16 bit.

Do you mean dither option available when setting up Converter?

Yes that should be internal foobars dither. It is not as light-shaped as low-shibata or modified-e-weighted, nor it is plain TPDF, but for most material it should be OK if you do not want to use sox/finalcd/ssrc command line.

Thank you!

SoX works with both FLAC and WAV no need to convert. FinalCD works with WAV only.

SoX Foobar plugin works the same as the command line. The difference is only with dithering. For simplicity you can stick with foobar UI conversions.

If you decide later to try SoX command line try plain TPDF dithering (in command line posted -b 16 option before the filename for 16 bit should be then added for complete conversion, you can look into manual), especially if there is any possibility of changing the output later. Shaped TPDF is recommended only for one-way conversion to final 16/44.1 version. Or try FinalCDs algorithms, dither shaping is very low there and still improves resolution.

I'll definitely try the command line. Thanks again!

Just to make my previous post clear

sox.exe  -V2 infile -b 16 outfile rate -v  -b 92 -a 44100 dither

works for me for 16/44.1 conversion with plain tpdf dither, other modifications apply as discussed. But the resampling itself is the same as with foobar's UI.

as for FinalCD example I use most often this

finalcd.exe /f1 /d1 "infile.wav" "CDoutfile.wav"

Both could be modified for use with batch files and file lists e.g. finalcd.exe /f1 /d1 "%~n1.wav" "%~n1CD.wav"