[artazzzzzz]

hello,
As far as I know bit rate = sample rate x bit depth x number of channels.
For example a CD quality song which is 16 bit depth would be at 44,100 khz x 16 bits x 2 channels (stereo) = 1411200 bits or 1378 kbps bit rate. That means a lossless song in WAV format should be at 1378 bit rate or a little bit less in FLAC. Now if you convert lossless quality song which was recorded at 16 bit depth to MP3 (let's say 320 bit rate VBR) its bit depth automatically drops as number of channels and sample rate remain the same.

There are MP3 decoders that offers 8/16/24/32 MP3 decoding.

My questions are:

1. What those decoders actually do?

2. How is it possible to decode MP3 song in 16 bit and higher quality as an MP3 song is usually lower than 16 bit depth?

3. I've heard about accuracy of decoding, will an accurate decoder give me better quality?
 
Thanks

 

[DaBomb77766]

Lossy compression has nothing to do with bit depth.  They don't decrease the bit depth to compress it, they throw away bits and pieces of the audio that the encoder assumes nobody will hear.  It follows a psychoacoustic model.  For example, if a really loud and really quiet sound play at the same time, it will throw away the quiet sound if it is considered too quiet to hear.

 

[Tilpo]

Lossy compression has nothing to do with bit depth.  They don't decrease the bit depth to compress it, they throw away bits and pieces of the audio that the encoder assumes nobody will hear.  It follows a psychoacoustic model.  For example, if a really loud and really quiet sound play at the same time, it will throw away the quiet sound if it is considered too quiet to hear.

The problem then with lossy is that sometimes it throws away information that IS audible. The difference between e.g. FLAC and MP3 V0 isn't going to be big, and there might be cases where you can here a subtle difference in detail or hear a quiet sound that is otherwise omitted.
On the other hand if you compress even further, let's say >192kbps, you can definitely hear the reduced quality as too much information is omitted from the music.

 

[DaBomb77766]

Quote:

Quote:

Lossy compression has nothing to do with bit depth.  They don't decrease the bit depth to compress it, they throw away bits and pieces of the audio that the encoder assumes nobody will hear.  It follows a psychoacoustic model.  For example, if a really loud and really quiet sound play at the same time, it will throw away the quiet sound if it is considered too quiet to hear.

The problem then with lossy is that sometimes it throws away information that IS audible. The difference between e.g. FLAC and MP3 V0 isn't going to be big, and there might be cases where you can here a subtle difference in detail or hear a quiet sound that is otherwise omitted.
On the other hand if you compress even further, let's say >192kbps, you can definitely hear the reduced quality as too much information is omitted from the music.

 
Very true, but about 95% of the people out there probably can't tell the difference between 128kbps MP3s and high-resolution FLAC files.  320kbps or good variable bitrate MP3 or AAC is more than enough for most people, really.  Of course, we here at head-fi accept nothing but the best and would apparently rather have 100 high-resolution songs on their DAP than 1000 lower quality MP3s.

 

[Ham Sandwich]

MP3 decoding is math.  You can do that math at 16-bit integers, 32-bit integers, 32-bit floating point, even 64-bit.  An advantage to doing the math at higher bit levels (higher than 16-bit integer) is that rounding errors can be minimized, special cases like when a peak ends up decoding above 0 dbfs can be adjusted for without chopping off the peak.  There is also an advantage that if you decode the MP3 to something like 32-bit floating point you can keep it at that higher level through the processing chain and do things like apply replay gain or EQ while the music data is 32-bit floating point rather than converting it down to 16-bit integer before doing processing like replay gain and EQ.
 
In the end the higher bit depth MP3 data will get converted to 16 bit integer for listening if that is what playback output is.  What is affected by the higher bit depth in decoding is the internal processing of the number.