"Core" DTS is lossy as we know. DTS 96/24, where the source has been split into two 'streams' ('core' +
extension') is back-compatible with older DTS because in the absence of DTS 96/24 decoding specifically, a 'regular' DTS decoder can still decode the core (lossy) stream. My question is, using a DTS 96/24 decoder, in which both streams are recombined and decoded, is the result still lossy, or is is lossless, compared to original?

I would think it's still lossy, because DTS has been touting their more recent DTS-HD as lossless. But it's actually unclear to me how to classify a bona-fide DTS 96/24 encode/decode -- especially as lossy files typically don't have a 'bitdepth' associated with them.

I've never come across a bitstream, but my understanding is that it's lossy. IIRC the target bitrates mean that is has to be.

It's a long time since I was at the AES where they announced it - someone in the audience asked "what psychoacoustic model are you using above 20kHz?"

Cheers,
David.

DTS 96/24 is also lossy, of course.
DTS-HD Master Audio supports up to 7.1 96/24 lossless reproduction of the original studio master.

OK , lossy, but how so? (It's kinda the same question that AES member asked). What's still 'left out' of the 96/24 audio bitstream?

Think of it as HDCD. If decoder can decode the "extension", you get higher quality output than just the "core", but it's still lossy. I haven't seen the specs of 96/24, so I dont't know how it works. Even the DTS-HD Master Audio contains backward-compatible DTS Digital Surround 5.1 or 6.1 channel "core". But DTS-HD Master Audio is not an extension to the core. In fact, the core is embedded in DTS-HD stream.

I understand the core + extension strategy for back compatibility and I understand that DTS-HD implements back compatibility differently. What I don't understand is what data is still 'lost' when you have core + extension playback of a DTS 96/24 stream.

Basically, how 'lossy' is DTS 96/24 when used in its 'best' mode?

Don't let the name fool you ... even if the input files can be 24/96 and decoding can be done to 24/96, that does not mean that it is bit for bit exact (compare 16/44.1 wav to mp3 back to 16/44.1 wav).
So what is lost will depend on the material to be compressed. Consumer DTS, as we know it, is CBR (in the case of DTS-96/24 always 1.536Mbit/s) so .. to keep the bit rate constant the quality has to vary. I have read somewhere that (normal 48/24) DTS gives at best 18 to 19 bit resolution, No idea about worst case.

As far as I understand (DTS is a bit vague in some details) the "core" is used below 24kHz and the 96/24 extension adds some spectral replication of the (few) higher frequencies and outputs at 96 kHz.

Well, 18/19 bits , if true, is not too shabby for best case -- given that CD is 16 bit. But not knowing 'worst case' it's hard to compare them by the numbers.

Sorry for for bringing this thread up, here is the link to the document where you can find out much about the DTS 96/24.