You can use w64 for files greater than 4gb.

BTW Tom the printmaster can be mag or digital. Is it any particular channel you are seeing with clipping? And more importantly can you hear it

I can not render out of premiere to a w64, but I will do more research into it.
Thanks for your input on all this guys.

@zoidberg I see. The most of it tends to be in the center channel. I'm not sure if I can hear it since I have nothing to compare it to, but it's not like anything stands out to me, no.

@CSchmidlapp Now that I think of it, yeah it might make sense to reduce gain during the resampling process due resampling potentially introducing more clipping since the curve can be higher in between samples, which is where the resampled signal might have a sample point. But as long as it's only done for the resampling it's fine I guess. Theoretically in order to have no loss at all it would probably be best to set to 24 bit first, then do the reduction, then resample with good resampler, then either keep 24 bit or dither back down to 16 bit .... but I suppose I'm being a bit too obsessive maybe, heh. Just checked that SRC comparison chart and EAC3TO with the SSRC algorithm actually seems to be a pretty good resampler too, so I probably shouldn't worry.

I can also recommend w64, but compatibility can be a problem sometimes.

It could well be that some sort of limiter has been applied to avoid a hard clip. It's a shame there is no more magnetic striping (except for small gauge film), there's something about that saturated/overmodulated sound

(2020-05-09, 07:47 PM)zoidberg Wrote: Just out of curiosity when resampling/slowing down to 23.976 48kHz are you processing each individual reel or a stitched together file?

I myself don't as I couldn't care less about sticking to any BD video specifications when my network player supports 24p in conjunction with 44.1 kHz FLAC just fine, but if, I'd process the whole thing after having it stitched together.

(2020-05-09, 07:47 PM)zoidberg Wrote: I only ask as when I did my first Cinema DTS sync, processing the files in eac3to resulted in clipping being detected and corrected which usually meant a small amount of -ve gain.

This should be due to intersample clipping only as this is the only case where it would make sense for eac3to to lower the gain on a given PCM source. Unlike lossy codecs which might overflow coming from the frequency domain, being scaled down to 16 or 24 bit PCM, any real clipping in a LPCM (non-floating point) would already be part of it to begin with.

(2020-05-09, 08:31 PM)CSchmidlapp Wrote: Eac3to did not detect any when I ran it through, but it may be already there.

Neither does eac3to detect any clipping on TrueHD or DTS-HD MA tracks as any clipping "baked in" is the way it is, anyway.

(2020-05-10, 12:01 AM)TomArrow Wrote: @CSchmidlapp Now that I think of it, yeah it might make sense to reduce gain during the resampling process due resampling potentially introducing more clipping since the curve can be higher in between samples, which is where the resampled signal might have a sample point.

Yep, the infamous (but probably overrated) "intersample clipping" during reconstruction. If eac3to does the resampling, it takes this into account by applying a negative gain when needed in the 2nd pass. Not sure about the other SRC, but when in doubt, I'd say that one could just reduce the gain by a few dB. Given proper dither, one virtually doesn't lose anything, neither at 16 bit.

In general, I find it totally idiotic to stress out the "near 0dBFS zone" anyway - even if the whole flick would only peak at about -6dBFS (god forbidden these days, I know), one would theoretically lose one single bit of performance an has to turn up the "volume" control on the AVR by the same amount (how frightening  ). That whole limiting and brickwalling is just one thing: retarded and completely unnecessary.

(2020-05-10, 11:48 PM)little-endian Wrote:

(2020-05-09, 07:47 PM)zoidberg Wrote: Just out of curiosity when resampling/slowing down to 23.976 48kHz are you processing each individual reel or a stitched together file?

I myself don't as I couldn't care less about sticking to any BD video specifications when my network player supports 24p in conjunction with 44.1 kHz FLAC just fine, but if, I'd process the whole thing after having it stitched together.

Ive done exactly this.
I have a 24p synced 16bit 44.1 kHz.wav master.
I'm planning on creating BD discs at some point, so the BD specs are necessary in this case.
Also the resulting master would only sync to my project due to changing the opening logos not present on the original blu-ray.

I do plan on uploading a .mkv version to start with (in BD spec so people can burn there own discs with TSMuxer ect with little effort), so I could supply the CDTS and LD untouched as FLAC files separately, for those who want a more 'pure' version.

(2020-05-10, 11:48 PM)little-endian Wrote:

(2020-05-09, 07:47 PM)zoidberg Wrote: I only ask as when I did my first Cinema DTS sync, processing the files in eac3to resulted in clipping being detected and corrected which usually meant a small amount of -ve gain.

This should be due to intersample clipping only as this is the only case where it would make sense for eac3to to lower the gain on a given PCM source. Unlike lossy codecs which might overflow coming from the frequency domain, being scaled down to 16 or 24 bit PCM, any real clipping in a LPCM (non-floating point) would already be part of it to begin with.

(2020-05-09, 08:31 PM)CSchmidlapp Wrote: Eac3to did not detect any when I ran it through, but it may be already there.

Neither does eac3to detect any clipping on TrueHD or DTS-HD MA tracks as any clipping "baked in" is the way it is, anyway.

(2020-05-10, 12:01 AM)TomArrow Wrote: @CSchmidlapp Now that I think of it, yeah it might make sense to reduce gain during the resampling process due resampling potentially introducing more clipping since the curve can be higher in between samples, which is where the resampled signal might have a sample point.

Yep, the infamous (but probably overrated) "intersample clipping" during reconstruction. If eac3to does the resampling, it takes this into account by applying a negative gain when needed in the 2nd pass. Not sure about the other SRC, but when in doubt, I'd say that one could just reduce the gain by a few dB. Given proper dither, one virtually doesn't lose anything, neither at 16 bit.

In general, I find it totally idiotic to stress out the "near 0dBFS zone" anyway - even if the whole flick would only peak at about -6dBFS (god forbidden these days, I know), one would theoretically lose one single bit of performance an has to turn up the "volume" control on the AVR by the same amount (how frightening  ). That whole limiting and brickwalling is just one thing: retarded and completely unnecessary.

Being a bit of a novice in the world of audio beyond the basics, all this goes alittle over my head to be honest.
I'm kind of following along but it's all 'back to school' time for me lol.

Would turning down the volume of the clips / reels before 'converting' out of Foobar2000 be more desirable?
Perhaps the creator of the plug-in could implement a -gain function or something, as people here have mentioned it could be that causing the clipping?

Just to add I can not tell the difference by ear between the original reels, my synced master and the slowed and resampled version.

I suspect the clipping is introduced during slowdown/resampling although I can't remember if keeping the file 24bit made any difference. Obviously if you know the negative gain applied you can increase playback level by the same, for those who like their levels consistent. I don't follow your comments about brickwalling, are you referring to home mixes?

I don't have any experience of streaming with media players, I take it the FLAC is being decoded and sent as multi-channel PCM to your receiver? Are you sure the LFE channel is getting +10db boosted as would normally happen when decoding AC-3 or DTS

@CSchmidlapp

No worries, don't hesitate to ask for details if certain points are unclear.

That being said, there are still some things which are unclear to myself even after quite some years fiddling with the theory of digital audio (and trying to separate the voodoo stuff from what really matters).

One for instance is the gain control for data reduced sources. As far as I understand it, lossy codecs like MP1/2/3, AC3, AAC, Vorbis, DTS (the coherent acoustics variant for home) all operate in the frequency domain using floating point calculations and thus don't have any "word length" like LCPM or DSD (1 bit, heavily dithered and noise shaped, virtually only used on the SACD) but only some average equivalent SNR-wise.

Considering this, the advertised "96/24 DTS discs" or any "24 bit" statements for the core such as DTS, is already misleading as it is up to the decoder to which bit depth it decodes it to. I think technically, it's not even possible to know for sure what "bitdepth" a given PCM master originally had before encoding. It could only halfway be estimated by the quantization noise.

Now with ADPCM it seems to be a rather static mapping, in case of the 4:1 variant, which also Cinema DTS uses (maybe some here played around with Microsoft  and IMA ADPCM back in the days) 4 bit per sample, so depending on the not so crucial acoustic parts, the SNR will be lower if I skim read the document correctly. This also might be the reason why the decoder outputs 16 bits and not any higher wordlength which most AC3 and DTS decoders do. I guess the author Maxim could shed some light into this.

1990s theatrical DTS based used the original Apt-X 100 coding scheme which was a lossy fixed 4:1 compression ratio. It is NOT lossless. So transcoding these tracks to lossless Master Audio doesn't make sense. It's akin to encoding MPEG Audio to FLAC.

Infact the consumer Coherent Acoustics variant is far superior than the original Apt-X DTS and at full rate 1509kbp/s it is far superior in fidelity to 90s theatrical DTS and uses more sampling bands as well.

DTS went lossless around '04-'05 theatrically in favor of a lossless scheme based on Coherent Acoustics that eventually became DTS-HD MA yet retained the DTS bar code on the edge of the film strip so they were still able to offer discrete digital on special 70mm prints for example which Dolby could not do with the SR-D large barcode prints, only magnetic 6-Track Dolby Stereo.

So for all intents and purposes the closest consumer format transfer/container for original Apt-X theatrical DTS would be Coherent Acoustics 5.1 with a bitrate of either 960kbp/s or 1152kbp/s. But since original theatrical DTS and Coherent Acoustics are both lossy going full rate 1509kbp/s to prevent audible generation loss is ideal.

(2020-10-03, 03:02 AM)WiLDCAT Wrote: So for all intents and purposes the closest consumer format transfer/container for original Apt-X theatrical DTS would be Coherent Acoustics 5.1 with a bitrate of either 960kbp/s or 1152kbp/s. But since original theatrical DTS and Coherent Acoustics are both lossy going full rate 1509kbp/s to prevent audible generation loss is ideal.

... Or, even better, use a lossless compression format instead to completely eradicate even the theoretical possibility of perceptual degradation, as he's done. Not the most storage efficient, but preferable from a preservation perspective. I dunno any other reason you'd want to deliberately do lossy -> lossy transcoding apart from file size and the difference is arguably not worth the cost.

Nonetheless, thank you for the fairly tidy summary of the compression methods used by DTS formats, it's nice seeing that put fairly succinctly in one place.