Here is what hairy_hen had to say about the cinema DTS when he worked on some of it ....
hairy_hen said:
I was listening to the cinema DTS tracks for the Star Wars special editions recently, and made a somewhat alarming discovery: namely, that theatrical DTS is an inherently flawed system in how it handles the bass content. I haven't yet been able to listen to any of the other DTS preservations that have been done, but based on what I've heard with the SW films, there is a good chance some of the specifics of DTS calibration may not have been set properly when decoding these tracks.
Since it is not a completely discrete channel system, the LFE information is stored in the surrounds and intended to be low-passed out to the subwoofers at 80 Hz; while the surrounds are high-passed at the same frequency. I spent a lot of time reading all the DTS technical papers I could find; but unfortunately DTS gives no specific information about the kind of crossover that is used with their system. So I then did some research about the Linkwitz-Riley crossover technique, which is the only method of maintaining constant level and phase between all frequency components.
Believing this must have been their intention, I applied an 80 Hz Linkwitz-Riley crossover in iZotope Ozone by applying two second-order Butterworth filters in series (this means cascading two 12 dB/octave filters, one after the other, for a total slope of 24 dB/octave and with 80 Hz as the -6 dB point). Once I had raised the resulting LFE signal up to its proper level—which initially had to be done by ear since I didn't know the precise amount of gain that should be applied—I discovered, to my alarm, that the sound of the bass was completely wrong in some scenes. Further listening revealed that in any scene in which the surround channels contained bass at frequencies in the vicinity of 80 to 120 Hz, this content was mixed with the LFE signal and then boosted to a level far beyond that at which it was intended to be heard. I proved this conclusion by listening to the 5.1 AC3 tracks from the SE laserdiscs, which have a true discrete LFE channel, and sure enough this issue of improper bass-management in the surrounds is not present.
The only way I could 'solve' this issue was by using a completely different crossover scheme, one which surely does not measure correctly by any usual standard, but which nonetheless seems to be the only way to make the mix sound like it is supposed to. I'm not the only person on the internet who has noticed this as a potential problem, so after reading some comments on another forum, I arrived at the solution of using a 48 dB/octave low-pass at 80 Hz for the LFE signal, and a 12 dB/octave high-pass at 120 Hz for the surround channels. Later DTS units apparently implemented a steeper filter than earlier versions (which were probably the same wrong-sounding 24 I'd been using), and the surround setting corresponds to the natural rolloff found in the sort of rear speakers found in most movie theatres. My settings duplicate the way a typical DTS installation actually performs acoustically, which sounds 'correct' because it is using both phase shift and speaker limitations to prevent the offending frequencies from becoming too bloated. In other words, theatrical DTS has a design flaw, and I figured out how to cover it up.
I also tackled the issue of subwoofer levels, since going by the usual specifications for level calibration, the numbers given in their papers make no sense. DTS do not specify at what level the LFE signal is to be mixed with the surround channels, nor do they indicate what method is being used to sum the results of the low-passed stereo surrounds into a mono subwoofer—this can result in a net gain change of 0, +3 dB, or +6 dB, depending how it is done—so clearly their unit is programmed to take care of all this without any expectation of the user having to understand what it is specifically doing. Their specified calibration levels only work along with the programming of the hardware device, so for us trying to decode these tracks at home, we're pretty much guaranteed to get it wrong without careful listening tests.
Note also that DTS changed their LFE mix level in 1999, lowering it by 3 dB and requiring a corresponding 3 dB increase in subwoofer output to yield the same acoustic gain. This was done for the purpose of greater compatibility with the way printmasters were being produced, and it definitely affects any projects using a cinema DTS track from these years. All films from 1999 and later must have their LFE signal gained up an additional 3 dB compared to films released from 1993 to 1998, or they will be incorrect. (These calibration oddities only apply to cinema DTS, not the home video version, which follows the more sensible method established by Dolby.)
The numbers I arrived at for correct LFE level came to +17 dB for the earlier discs, and +20 dB for later ones. (This is assuming unity gain when summing to mono, not +3 or +6.) I haven't really 'measured' this per se, but this sounds correct when comparing to the Dolby Digital tracks. I'll probably want to test it some more before calling it 'final', but I'm pretty sure I've got it right, or at least very close.
Since I'm doing this in Pro Tools, my specific method (a combination of iZotope Ozone and the program's own Downmixer plugin) may not specifically apply to what other people are using, but I can try to help in finding settings that yield the correct results. It is very important to get this right, or else we won't be hearing the level balance of these mixes the way they ought to sound. I think I've finally figured it out, and I want to make sure we all know how to do it properly.
![[Image: sNn6jyF.png]](https://i.imgur.com/sNn6jyF.png)
![[Image: 0sPZMBH.png]](https://i.imgur.com/0sPZMBH.png)
