How to decode 6-track APTX-100 (cinema DTS) with the correct channel levels

[stwd4nder2]

Recently did a BD Sync for the Speed Racer Cinema DTS and I just kinda winged it with the LFE; left it alone in Foobar and instead went into Audacity, copied the surround channels and mixed them down to mono, put an 80hz 48dB rolloff LPF on the mixdown and corresponding HPF on the surrounds. Is this the way it's typically done here? Because I noticed the LFE sounded a bit different from the BD 5.1 this way.

I've always used the Foobar plugin to extract the LFE, then make the gain adjustments in a DAW.

[Frontino]

For the sake of science though, I did convert the Tenet R6 AUD to both 16-bit and 64-bit Float to see what'd happen, and both seem to report back with the exact same results in Reaper when doing a dry pass with the original sample rate preserved (resampling to 48000 causes it to peak by +0.2, so I'm guessing it's the r8brain algorithm to blame).

Hi.
Can you confirm if the UHD BD of Tenet has the same mix given to cinemas, besides surrounds and LFE mismatch?
Like, are the L, R, C channels identical?
Thanks.

[little-endian]

So because a +6 DB boost in the LFE for mixes like Tenet would clip the peaks, one could conservatively instead lower all other tracks by -6 DB to preserve the LFE as-is (turning the mix from 16-bit to 17.5 bit to account for the -9 DB in the surrounds).

Which technically - audible or not - is the better way of doing it. Clipping should be always avoided.

Would this do a better job?

Apparently, yes.

Welp, I just tried a sample of a track that had known clipping in the center when decoding straight to 16-bit.  I did 32-bit and it's not there.

Exactly my observation with the Cinema DTS of "Fight Club" with the notorious "everytime the plane banked too sharply for takeoff or landing, I prayed for a crash or a mid air collison" scene on reel 2 - the surround channels reverely clip when decoded at 0dB setting in the plugin and 16 bit output:



Attenuating it, doesn't change anything and it remainds clipped:



When using 32-bit floating point however, while it nominally clips as well:



Peaks can be nicely recovered, confirmed also by the fact that a reducation to nominally -5dBFs in this case leads to still a higher peak of about -1.1767 dBFS.

For the sake of science though, I did convert the Tenet R6 AUD to both 16-bit and 64-bit Float to see what'd happen, and both seem to report back with the exact same results in Reaper when doing a dry pass with the original sample rate preserved (resampling to 48000 causes it to peak by +0.2, so I'm guessing it's the r8brain algorithm to blame).

I guess with "blame" you simply mean "caused by" as level rises are to be expected with fundamental operations such as sample rate conversions or when dealing with lossy sources - which is exactly the reason why better DACs such as from RME or Benchmark nowadays attenuate the levels to create additional headroom for mangling around the audio before final conversion. It shall be meantioned though that the audibility is debated as in most such critical cases, the source material is already (too) hot to begin with and hopelessly clipped anyway.

Something to note, the DTS-HD MA on the UHD (which seems to be a direct port of the Cinema DTS) also contains clipping on the LFE channel:


Some other titles that are also close ports of the Cinema DTS show the same.

Indeed, however them having baked in clipping (too), doesn't mean it's right and one should follow their example. If there is an option to have less or no clipping, then this is technically the correct way to go. Or in other words: proper decoding and level adjustment of the Cinema DTS sources then gives us the chance to create a better version than the ones used for the (UHD) BDs.

Which is also the reason why I didn't decode the Fight Club Cinema DTS AUD files with +6dB for the LFE and -3 dB for the surrounds straight away as the LFE would already clip in the "mid air collision" scene as becoming about +2dBFS then, so I opted for lowering L,R,C by 2dB, raising the LFE by 4dB and the surrounds by -5 dB. No (additional) clipping this way, although the L,R,C channels for Fight Club look a little "hot" on their own, even when decoded to floating point which makes me wonder whether the APT-X100 is really using it for all channels (on the other hand, why not, right?).

[little-endian]

Alright, after I contacted the author of the Foobar APT-X100 plugin for the ignored LFE/Surround gain settings for versions > 0.3.5, he fortunately fixed it quickly (still a mystery to me how such an elemental bug can survive through several updates, but anyway).

So now we have version 0.3.11 whereas the newer ones also seem to decode more than 6 channels.

Leading to another observation because where version 0.3.4 of the plugin produced 5.1 (6 channels) for e.g. the Fight Club Cinema DTS, this one outputs 6.1 (7 channels) whereas both plugin versions report "7 ch" either way to make the confusion perfect. I find it especially confusing as either that is yet another bug in e.g. 0.3.11 now (the additional channel 5 after the LFE seems to be more or less simply the sum of channnel 6 and 7) or we missed the 6.1 presentation all along for Fight Club because this sync project here also only shows 6 channels.

[FoLLgoTT]

I'm struggeling with "Fight Club", too. A new release of the decoder is online with a new front channel gain setting and a fix for the surround channel gain.

Only the curiosity of the back surround channel remains. I don't know if "Fight Club" was encoded with DTS ES Matrix or Discrete. The channel looks like it is a mix of the two surround channels, but this is hard to verify...

[little-endian]

Not sure if resampling can produce clipping, have to think, maybe.

Just to pick up that remark since I'm scrubbing through the thread once again as these levels and (buggy) decoding thing is quite a rabbit hole -

Yes, it can and it's what intersample peaks are actually all about. Many DACs nodaways perform upsampling for performance improvement and if there is not enough headroom to account for the reconstructed waveform which may be higher in level than what the sample points suggest (Cool Edit Pro / Audacity traditionaly being one of the few to draw the waveforms correctly by the way), clipping occurs.

[little-endian]

Oddly the decoder plug-in, all versions beyond the very original 1.1, crash out, give negative starting point and don't work or play back garbage when trying to decode the original JP Cinema DTS discs although the later re-formatted updated discs decode the same as anything else with all the versions.

Since currently lively in contact with the author anyway, I've reported that bug as well as I guess someone has to do it. 😛

[little-endian]

@schorman,

Your name was mentioned by Soundman who you have been in contact with regarding Cinema DTS. Since this whole thing is interesting for everyone, I post it here.

Now I recently got in contact myself with the author of the famous APT-X100 plugin for Foobar. Since he also makes use of libraries, many details still seem to be vague and kind of a black box.

Meanwhile upon my input and bug reports, we arrived at v0.3.13 but I still have my doubts if the decoding and filtering now really is proper or not.

Oddities which still persists and should be sorted put from my point of view:

- the older version of Jurassic Park with its .R1, .R2 files are only peoperly decoded (at least in a way of not entirely be screwed up) by v.0.1.1. According to the author, this is due to missing header information of that particular Cinema DTS source, leaving unanswered how the older plugin version then managed to do it, but he's checking on that.

Question to all here: does anyone of you guys know about similar "raw" Cinema DTS files like that particular version of Jurassic Park, naming the tracks .R1, .R2, etc.?

- decoding results are slightly different between versions (also 64-bit vs. 32-bit of Foobar) and it's unclear which one is correct or at least closer to some hardware CDTS-decoder

- the older versions including 0.3.4 don't support 6.1, however sources such as Fight Club using "ES" seem to ask for it. In that sense, simply sticking to v0.3.4 doesn't seem to be the complete answer either as then one omits the back surround channel

- the most apparent thing with Cinema DTS I personally have and verification with the user "FoLLgoTT" confirms, is that the bass part tends to be weaker than those of official releases, including "Fight Club". So either the Cinema DTS tracks are tamed on purpose, which is hard to believe for cinematic playback, certain (UHD) BD releases are "sweetened up" for more impressive bass effects, or something goes wrong with the LFE generation / low frequency filtering in general.

So I wonder if you could give v0.3.13 a try and present your verdict compared to v0.3.4. As far as I understood the author, the 32-bit-version of Foobar is allegedly closer to the real deal.

[FoLLgoTT]

WARNING: Version 0.3.14 was released recently, but it totally screws up the channel order. Additionally, the 7th channel is gone. So, there must be more changes than described in the read me.

EDIT: channel order of "Jurassic Park" is correct. "Gladiator" and "Fight Club" not.

[stwd4nder2]

Question to all here: does anyone of you guys know about similar "raw" Cinema DTS files like that particular version of Jurassic Park, naming the tracks .R1, .R2, etc.?

JP was the very first DTS release. I'm not sure of the second, but I don't have any other discs from 1993.

WARNING: Version 0.3.14 was released recently, but it totally screws up the channel order. Additionally, the 7th channel is gone. So, there must be more changes than described in the read me.

EDIT: channel order of "Jurassic Park" is correct. "Gladiator" and "Fight Club" not.

Gladiator and Fight club are DTS-ES mixes. Maybe it's just messing up those?