Jim I reember reading these articles of yours (as well as other stuff I can't now recall) and had come away with the impression tha the adaptive filters had never been implemented in HDCD. John W seems to be of a different opinion. I wondered whether John could perhaps identify some HDCDs which had implemented the dynamic coding.
Can you give me an idea Jim how you would see an alterntive adapative filtering (ie one not encoded on HDCD but based on analysis of the ordinary cd) might work.
In the past I wasn't able to tell as I didn't have an 'official' HDCD decoder in any of my kit. (I do now as I bought a CA CX-U a few weeks ago and this claims to have one.) But when I used the decoder software based on code I and someone else found on the web it failed to report any such codes in any of the HDCDs I have. So I've suspected they don't normally actually get used.
If you look at the patents for HDCD they are the usual vague run-around and give no details at all. Claim all, say nowt.
I'd probably approach switching via the kind of method you outlined. i.e. look for transients which have a slew rate that approaches the maximum for the bandwidth, and then use a different reconstruction filter for those parts of the signal.
For testing adding some filter switching, I'd tend to write something like an x4 upsampler to create output files I could then play and measure to experiment.
i.e. Rip an HDCD as is. Then generate an 176.4k upsampled file from it. (You can see some simple x2 upsamplers on my software page, with a choice of filters.)
But TBH I've always felt all this was really a matter of synthesising something 'nice' rather than 'accurate'. The simplest choice would be to go to linear interpolation (a la the old 'legato link' argument that Pioneer trotted out). But take yer pick if you want to do anything like this as it all seemed to go too far to me. This seems to me to fall into the category where the user should be able to choose what suits *them*.
I *suspect* that some modern DACs do this, even with plain CD, but again it tends to be buried under suit-speak and fancy names so you can't easily get to the details.
Now I *do* have a CX-U I plan to do some wideband re-recordings off HDCDs and see if I can detect changes in waveform shape that signify any dynamic alterations in reconstruction filter. But its a challenge, and HDCD is not top of my list. (And I work slowly these days anyway.) So avoid holding your breath. If anyone else wants to try I'd suggest making a 192k recording with a good recorder from the analogue output of an HDCD player, then search for 'out of range' slew rates as a sign. Can't say more until some measurements were done as we're looking for what may be a boojum.
In addition, I found HDCDs that weren't, unlabelled HDCDs, and discs that seemed to use 'HDCD' to peak squash badly. i.e. use as a good old loudness war 'effect'. (Or a disc screwed up when mastered, mangling the codes.) So trying to work out anything from them is a nightmare.
My bottom line was to avoid HDCD as a PITA. The snag being when the music you want is on an HDCD. :-/ So I just used the peak uncompression to get versions I could play *when* they sounded better that the disc.
TBH My main frustration here is that some Joni Mitchell discs are riddled with this problem. Some sound a lot better when decoded. Others sound level crushed or lousy either way. I just wish the people mastering them had made decent, plain, CDs.
Minefield.