Hi all,
I was busy doing some tests for my upcoming lossless comparison when I stumbled upon some really weird results. It turned out that the FLAC-files I ripped from a DVD-A I was testing had wasted_bits=2 and wasted_bits=4 on some channels. I was kind of surprised, but as according to the Wiki there is some evidence MLP (the compressor used for DVD-A) searches for these wasted bits, it seems logical: the DVD-A now holds 8,0GB (of 8,5GB max) of data, so without this 'lossy' encoding, they would not have been able to cram all this stuff on the DVD-A. (For the same stuff for The Return of the King, which features even more music, they used a double-sided, DL DVD and don't use these 'wasted bits'-approach as there was enough space)
The DVD-A I'm talking about is from the 3-CD + DVD-A set of the The Lord of the Rings: The Two Towers (Complete Recordings) soundtrack, which was released on November 7, 2006, so it precedes the first HA-post on LossyWAV by about 9 months.
I couldn't find any information on these wasted bits actually being used with MLP, but as the sales of the original soundtracks are over millions, the soundtrack I'm talking about must have been sold at least 10.000+ times, so this stuff is actually being used in the 'real world' as well
I hope someone might find this interesting.
I knew about this. Although it was/is only a possibility (to achieve longer playing time into a given data space), it lead to discussions by some, claiming that DVD-A was not lossless (and SACD was better) (but let's not revive that).
with a quick search I found a document (http://www.thef80.com/w_paper/mlp_jap_aes9_1.PDF) with this quote
Compression adjustment
A producer may wish to save space used by a recording, or to reduce data rate. Lossless compression extends the number of options.
With MLP, data is automatically saved if the incoming precision is reduced. So, reducing (for example) a few or all channels from 24 to 23 or 22-bit will automatically save data: each bit saves about 8%.
Another option for increasing playing time is to lowpass filter some of the incoming channels. Low-pass filtering reduces the entropy in the signal and the MLP encoder provides a lower data rate. This method is preferable to the use of mixed sample rates because the control is flexible and the stream is homogeneous.
As you can see, these are merely options and not standard.
However, I do not think this is quite LossyWav. LossyWav follows the signal and determines where the noise introduced using less bits will be hidden beneath the signal.
BTW. 20 bit is also a valid (native) bit depth for DVD-Audio, but it is never used AFAIK.
Although it was/is only a possibility
So this proves that it is actually used
Compression adjustment
A producer may wish to save space used by a recording, or to reduce data rate. Lossless compression extends the number of options.
With MLP, data is automatically saved if the incoming precision is reduced. So, reducing (for example) a few or all channels from 24 to 23 or 22-bit will automatically save data: each bit saves about 8%.
Another option for increasing playing time is to lowpass filter some of the incoming channels. Low-pass filtering reduces the entropy in the signal and the MLP encoder provides a lower data rate. This method is preferable to the use of mixed sample rates because the control is flexible and the stream is homogeneous.
Yes, this is exactly what has been done: on the stereo files they reduced everything to 22-bit, while on the surround files the front channels are 24-bit and the others are 20-bit.
However, I do not think this is quite LossyWav. LossyWav follows the signal and determines where the noise introduced using less bits will be hidden beneath the signal.
After some reading, I understand what you mean. The only thing I knew about LossyWAV was that is was made to make use of the wasted_bit scheme of FLAC, but the algorithm for choosing when, where and how to lose bits is indeed far a more advanced topic...