Just to give a hint at how far beyond a PIC MP3 has got to be...
A few years back I worked with a company working with MP2, which is
computationally simpler than MP3, and took the time to hack my way through
the overview of how the algorithm works.  It ain't simple at all.

First, a few milliseconds of sample are stored as a block (occupying more
RAM than any PIC has, already).
Then these samples are processed by a digital filter bank that separates the
data into bands of an octave or two. (Several MIPS)
Then a DFT is performed on each band in order to identify maximum-energy
peaks, which are sorted. (A bunch more MIPS)
These peaks are then presented to a "psychoacoustic model", which sorts the
bands in order of importance to the listener, taking into account such
things as the sensitivity of the ear to various frequencies, and, most
importantly, masking effect.  It seems that if there is a loud sound at some
frequency, and a softer one at or close to a harmonic of that frequency, the
human hearing system throws away most of the information about the higher
harmonic, except for its relative intensity.  This information is then used
to assign the available encoding bits to the various bands based on their
importance in conveying the original sound. (Not really all that many MIPS)
Each band is then encoded separately using its allocated bits.  I'm pretty
sure that the lower-frequency bands are resampled to a lower sampling rate
which satisfies the Nyquist requirement for that particular band. (Yet more
MIPS)
For stereo, there are more tricks that take advantage of the redundancy of
the two tracks, and take special care to preserve the phase relationship
between the sounds in the two channels.

As I recall, it took around 20 MIPS to do encoding, and 15 or so to decode
MP2.  MP3 throws more MIPS at the problem to achieve higher compression
while still retaining a good listening quality.

That's why it takes a dedicated DSP or higher-end Pentium to do audio
compression/decompression in real time.

> -----Original Message-----
> From: Herbert Graf [mailto:hgraf@EMAIL.COM]
> Sent: Friday, November 17, 2000 9:06 AM
> To: PICLIST@MITVMA.MIT.EDU
> Subject: Re: [EE]: mp3 player
>
>
> > thanks , byt i'm looking for mp3 decoding alghorithm
> >
> > i want making mp3 decoder whit PIC microkontroler
>
>      How many times are you going to ask this? There is no
> real time mp3
> decoding algorithm for the PIC line of microcontrollers
> because PICs don't
> have enough processing power or memory. You MAY be able to do
> it in non real
> time, I doubt it though. TTYL
>
> --
> http://www.piclist.com hint: To leave the PICList
> mailto:piclist-unsubscribe-request@mitvma.mit.edu
>
>

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.