mp3 decoding

If you look at the internet radio app/demo PCB schematics, you will find that other than being short on RAM, a PIC18 w/Ethernet can stream live Internet MP3...

The reason being is the PIC18 uses an external MP3 decoder IC.

Check it out...

A PIC32 (if doing the decoding in external hardware) is overkill.

 -Ben

Also have a sneak preview here at [link]http://www.microchip.com/forums/fb.ashx?m=339169]E16 Audio and Speech/link] from the E16 Gallery real cool work

Hi all,

  it is look that int32 IMDCTInfo.outBuf[MAX_CHAN/*2*/][BLOCK_SIZE/*18*/][NBANDS/*32*/] can be used to store output pcm samples. Total size of this buffer is 4608 bytes and if output pcm samples are 16-bit this buffer will be enough to store samples ( 2 ch x 1152 samples x 2 bytes each ). 
The sequence of functions that manipulate the buffer is:
- IMDCT();
- HybridTransform();
- Subband();
- FDCT32();

After FDCT32() Helix call PolyphaseStereo() or PolyphaseMono() as first parameter is supplied from MP3Decode() pointer to pcm  outbuf. Instead of separate pcm buffer defined somewhere in main.c may be it is possible to supply MP3Decode with pointer to IMDCTInfo.outBuf ( through MP3DecInfo->IMDCTInfoPS->outBuf ). MP3Decode() works with input MP3 data in granule, then in channels ( one for mono, two for stereo ), blocks ( up to 18 ) and finally bands ( 32 ). After every step it is look that data remained from previous step in IMDCTInfo.outBuf are no more used, so we can use it to store pcm samples. This will save 4608 bytes RAM. I did not test yet this idea, probably in next few days will be able to test it.

I think MAD library use similar technology because MAD invoke user call back function with pointer to output pcm samples.

Regards,
Pirev