problems converting MP3s to WAVs

[Klaus-Peter Schrage]

Net Llama! wrote:
> On Fri, 27 May 2005, Klaus-Peter Schrage wrote:
> 
> 
>>Net Llama! wrote:
>>
>>>On Fri, 27 May 2005, Klaus-Peter Schrage wrote:
>>>
>>>
>>>>Sorry, Lonnni, there seems to be consistency in the way the line from
>>>>mp3 to audio cd works:
>>>>cdrecord needs an input which is 16 bit, stereo, 44100 Hz, uncompressed
>>>>- usually a wav file. If it isn't, you have to make it like that. So, if
>>>>your mp3 is already 44100 Hz, you're nearly done with a simple mpg321
>>>>command. Otherwise, use a tool like sox or sndfile-resample (in the
>>>>libsamplerate package) to fix the wav.
>>>>Klaus
>>>
>>>
>>>That seems to be my problem.  All of the MP3s are 32bit, instead of 16.  I
>>>just scanned through the sox man page and didn't see anything obvious on
>>>how to convert from 32 to 16 bits.  grrr.
>>
>>Can you put a sample mp3 somewhere so that I may have a look at it
>>(can't promise anything, I'm not a sound expert)?
>>Klaus
> 
> 
> Sure, try this one:
> http://netllama.linux-sxs.org/1.mp3
> 

This sequence of commands turnes the initial file 1.mp3 (8.6 MB) into an 
one track audio cd:
1) mpg321 -v -w 1_mo.wav 1.mp3
2) sox -V 1_mo.wav -r 44100 -c 2 1_fin.wav resample
3) cdrecord -v dev=/dev/hdc -pad 1_fin.wav

EXPLANATION short (a longer one might be worth a SxS ...)

1) Using the -v option in mpg321 tells you what the initial file is all 
about:

Freq: 22050
mode: Single-Channel
encoding: signed 16 bit
Bitrate: 32 Kbits/s

What these figures mean:

- Frequency (aka: sample rate):
Indicates how often a sample has been taken from the sound, it describes 
the horizontal resolution.

- Mode:
Of course, single-channel is otherwise called mono.

- Encoding:
Describes the fineness of each of the samples, ie the vertical 
resolution, in this case (signed) ranging from about -32000 to 32000.

- Bitrate:
the overall data flow per second. Eg in CD audio, we have 16 bit, stereo 
44100 Hz, resulting in a bitrate of
44100 x 16 x 2 = 1411.2 Kbits/s.
Common mp3 compression modes yield bitrates of 96, 128, 192 Kbits/s.

BTW: the filesize of the output wave file is 94.8 MB

2) 1_mo.wav still is mono and has a horizontal resolution of 22050 Hz. 
The sox command lifts this up to what cdrecord expects as input format:
44100 Hz, 16 bit, stereo.
The 'resample' effect is not obligatory, but it is recommended to 
enhance the rate conversion result.
The output 1_fin.wav is, not unexpected, nearly exactly four times as 
big as the input (1_mo.wav).

3) The only remarkable option is -pad. It enlarges the input file to a 
size which is a multiple of 2352, as data on an audio cd are stored in 
chunks of 1/75 of a second:
44100/75 x 16 x 2 = 2352 Byte.

Klaus