Still working on a final cleanup of 'Crime'. Explanation is below... Just completed an attempted final version, and the result is much more clean and airy. However, if you compare conceptually the sound quality with 'drawing a picture', the result was like drawing with too fine a pen. A general explanation and goal is below. The ETA for the next (hopefully final, or close to it) version is 6Hrs from now.
Even though the decodes of 'Crime of the Century' have apparently been relatively good -- I have desperately tired to do one last hail Mary. The correcting filter array has recently (since these tests) been Chebyshev (Q=0.8409), and there were several reasons for that -- mostly that there is full control crafting a frequency response without worrying as much about the shape of the filter itself. 'Butterworth' and 'Bessel' are better filters WRT phase, where the 2nd order approximation for 'Bessel' is a Q=0.500, and the 2nd order approximation for 'Butterworth' is Q=0.7071.
The reason why I have been disappeared for a few days (other than some personal stuff) is the careful crafting of a mix of Chebyshev, Butterworth and Bessel when appropriate. Using one style of filter, the Chebyshev is a bit easier for intimate control.
The important thing is 'frequency response', but phase is a close second. The DHNRDS has a special attribute -- above about 200Hz, it is linear phase straight through. A true DolbyA is NOT linear phase, but is fairly close ("no cigar though"). There COULD be benefit of taking advantage of the linear phase nature of the DHNRDS by using Bessel filters when possible, and avoiding Chebyshev unless really needed. When doing today's test version of 'Crime', 'Breakfast' and 'Quiet' -- the most 'smooth' phase filters were used, but still match the frequency response as close as the original Chebyshev.
This was a VERY, SUPER tedious set of filters to develop, but ended up looking straight forward (that is good that they look very sane.)
Also, I enabled two more super-duper decoding modes on the DHNRDS... It was a very simple set of changes, but really slows down the decode process. The highest quality was used only on 'Quiet' because of the quality of the vocals in the first cut. The other two albums (including Crime) used the new 'slightly higher' quality.
IMPORTANT: we can back off to the previous, more chunky filters. I think that these new filters are *theoretically* higher quality.
My observation: the sound is more light, airier.
Goal: "Sounds ok" means we NEED to do more work, "Sounds good" means we can tolerate status quo. This is a hail Mary to chase down any loose ends -- maybe clean up the results one last step.
Changes: corrective-EQ (same shape, smoother phase), higher quality decode, 'comments' added to .flac file for version ID.
Conceptual difference: 'Draw' the music with a finer pen.
This upcoming version tonight: choose a slightly more broad 'pen' than the just completed test version (uploaded, but don't bother, not worth the download considering the upcoming tonight) so that the sound is more solid, and decoded with absolutely the highest quality.
Will announce when the fully tested version is ready.
John