Hi Wolfgang,
Nice to hear from you. > I just read on the MDCP the faq about ATRAC3 at > http://www.minidisc.org/mdlpfaq.html and it says for Q#10: "What > bitrates are used?" A:".....In LP2 and LP4 modes 20 bytes of dummy > data per 212 byte soundgroup......" --> those 20 bytes -- are they a > fact or just a wild guess? Fact. I double checked it with Sony. > We know from ATRAC1 that it records at (exact) 292162.5 bits/sec. I > could not find an exact number for ATRAC3 (LP2) -- is it 132300? Yes, and it's easy to calculate: We know that SP mode runs at 172.266 soundgroups/sec (ref. FAQ Q#10), LP2 mode gets precisely twice the capacity of SP mode (contrary to Mr. Rat's freely made up facts and figures), so that's 172.266/2=86.133 soundgroups/sec. Multiply by (212-20=192) bytes/soundgroup = 132300.288 bits/sec for LP2. LP4 should be precisely half that: 66150.144 bits/sec. > I also read somewhere that LP uses huffman coding (fact?) -- only on > computers or on MD decks as well? (Well, it's got to be both or neither to be compatible with all players, right?) In fact it's true (again, contrary to Mr. Rat's free-association-masquerading-as-informed-posting), see http://www.minidisc.org/atrac2_paper/3.html where they show the particulars of the Huffman code table. > And last: at Q#18 this nice picture. What do they mean by > "tonal/non-tonal component decoding"? See again http://www.minidisc.org/atrac2_paper/3.html. They use some measure of tonality to decide when there is a tone in the spectra and then encode it with parameters such as a location and width (seems a bit like run length coding for graphics). This allows them to get a better/more compact representation for tones. > where in this picture takes the joint stereo stuff place? any > thoughts are welcome! Good question. I think the joint stereo stuff is completely left out in all the ATRAC2 literature. Mr Rat writes: > Sony had to do it that way becuase they left no room for expansion in the > MD storage algorithm. Admittedly, they never forsaw a time where anyone > would want to *reduce* recording quality. Sony did in fact provide for future expansion by allocating bits in the UTOC to indicate different encodings (see http://www.minidisc.org/md_toc.html#sec0cstruct), the problem is that non-MDLP players seem to ignore the encoding bits and try and play everyting as SP audio. > Looking at ATRAC in "bits per second" is somewhat inaccurate because it > doesn't work that way. An ATRAC encoder gets a 16 bit wide block of data > (Linear PCM, same as CD-DA). Standard ATRAC removes 4 bits out of every 5 > from the signal resulting in a 5:1 reduction ratio. LP2 is 10:1, removing > 9 out of every 10 bits. LP4 is not 20:1 bitwise reduction, but ~15:1 > reduction (I would have to check for the exact figure) combined with > partial stereo channel merging (joint stereo). ATRAC operates on 512 samples at once (1024 actually if you want to include the frame overlap), ATRAC3 on 1024 at once (again, 2048 including overlap). In ATRAC's case these 512 16 bit samples (1024 bytes) produce 212 output bytes, or a compression factor of 1024/212=4.83. Saying it removes "4 bits out of every 5" makes it sound like it operates without regard to longer periods in the signal, which seems misleading. Rick ----------------------------------------------------------------- To stop getting this list send a message containing just the word "unsubscribe" to [EMAIL PROTECTED]
