Star Wars 1977 70mm sound mix recreation [stereo and 5.1 versions now available] (Released) — Page 24

I second the last post!

H_H, can I ask how hard it would be to do Dolby Digital lossless versions? I ask because Apple TV won't pass through DTS. Or is there a way for me to make them from the source files? I'd guess I'd need to do something to make the volume levels equal, too?

Dolby Digital is lossy.  Dolby TrueHD is lossless, but there is no free encoder available, and the commercial ones are very expensive.   Price rather than difficulty would be the stumbling block.

Gotcha. I'll just stick with Handbrake's AC3 conversion.

I'm able to encode TrueHD, but for whatever reason this format isn't commonly used and may cause compatibility problems of its own.

For creating AC3, by far the easiest way to get access to a Dolby-certified AC3 encoder is with the Compressor application from Apple.  You'll get the superior sound quality the official algorithm offers as well as a host of video options, all quite affordably.  I've compared its output to that from Dolby Media Encoder and they are quite indistinguishable at a given bitrate as far as I can tell.

One of the things I like about using Compressor for AC3 is that it can encode straight from Apple Lossless files without having to go to WAV or AIFF first, which is useful for me because I often archive things in Apple Lossless to save hard drive space.  Compressor does, however, have the drawback of defaulting to a DialNorm value of -24, which equates to a 7 dB reduction in volume, so you have to remember to set it back to -31 if you want your levels to remain unaltered.  By contrast, Dolby's own encoder now starts at -31 instead of -27 the way it apparently did in the past. 

Both programs have a setting for -3 dB surround channel attenuation, which is used for lowering the rear channel levels from theatrical to home calibration standard and is necessary for most 5.1 film mixes when transferring to home video.  This is not needed for any mix that I have created, and so should be unchecked—otherwise the rear channels would come out too low.  Similarly, the setting for 90-degree surround channel phase shift should be unchecked for my stuff, since I already accounted for such things when putting the 70mm version together.

I guess I never did actually mention it here, having only talked about it sporadically in other threads, but there is indeed a new version of the 70mm mix out there.  The changes from the previous revision are minimal, with the main improvement coming from having used the Dolby Media Decoder application to upmix the laserdisc audio into five channels with Prologic II, which results in greater channel separation and consequently a more believable sound field.  (The better the acoustics of your room and the more accurate the sound system, the more likely it is that you'll notice the improvement.)  I also changed the level of one bass effect in the LFE channel so that it would blend better with the rest of the mix, but everything else is exactly the same as before.  The other two films will eventually receive the same treatment.

I'll be posting in more detail about this and about optimizing the other audio tracks later on, so watch for that if you're interested.  Links to all the new files I've made will be provided.

Great information. Thank you.

Oh, incidentally Apple's website for Final Cut Pro X says that a new version is coming this year. That could likely mean at next week's event where the new Mac Pro will presumably be announced. It might just be a point release to add optimizations for the new machine, though.

May not be the place for it, but when you post the level-matched FLACs for the new SWDE, will you start a new thread or repost in this one?

hairy_hen said:I'm able to encode TrueHD, but for whatever reason this format isn't commonly used and may cause compatibility problems of its own.

Wow!  I didn't realize that you had Dolby Media Encoder.  That must have set you back a bit.

Dolby TrueHD is only optional for BD, so this is the probable explanation for the compatibility problems – not that I've ever experienced any.  While it's not used as much as DTS-HD MA these days, it's common enough on BDs.

I believe the the LFE track comes from the 1997 SE mix (sourced from the laserdisc). The 2004 mix is too disgusting to be very useful .

Of course it isn't authentic.  But then, neither is the lack of true discrete channels from the actual 70mm version.  This isn't a real 5.1 mix, after all—upmixing of any kind results in crosstalk between channels, because it isn't possible to fully separate a stereo downmix back into its multitrack components.  Prologic II gives a more convincing representation of this than other processors, but it in the end it's still a compromise based on limited source material.

It's true that the original Prologic gives a mono surround channel while its successor adds a greater number, but do not assume from this that it would sound closer to the discrete channel version.  It does not.  Prologic II is capable of panning surround effects between rear channel speakers if the stereo source has been encoded that way, but when playing a track like this, where the surround effects from the original mix are all in mono anyway, they simply get sent equally to both rears.  The stereo content in the back is only whatever duplication from the front that isn't fully separated out, which mostly turns out to be the music.  All content sent to the rear speakers is delayed in arrival time so that any such correlated signals will be perceived as coming from the front, with the rear version simply coming across as ambience.

The fact that there is any music coming from the surrounds at all is inauthentic to the original, because back then these kinds of films were mixed so that the surround channel contained only sound effects, and the music was entirely confined to the front soundstage.  When you listen to the 1997 mix, which mostly comes from the four-track master, you can hear that it is a true discrete version, and this comes across somewhat differently than any upmix from stereo ever could.  It actually sounds a bit strange to me because I'm so used to hearing it this way . . . but in that respect it is unquestionably more like how the real 70mm version would have sounded.  I did briefly consider trying to use it as a source when putting my mix together, but I quickly realized it was virtually unusable (aside from the LFE) because it had been dynamically compressed to a great degree compared to the 1993 mix, which has virtually no limiting whatsoever.  The great number of additional sound effects would have made working with it more trouble than it was really worth, also (and of course there isn't a lossless version of it, either).  So I decided that the compromise of sacrificing discrete channels in favor of vastly extended dynamic range was a worthwhile tradeoff.

In any event, what really matters is that it sounds as good as it possibly can given the source materials available, and if it succeeds in that regard then I'm very pleased.

edit: duplicate post.

Oh yeah, there was a question about links for the new version: I'll probably post them in a new thread since it's for several tracks together, but I'll link to them here.

Any chance you could upload the new stereo version of your 70mm mix h_h?

New stereo version?  I haven't made a new stereo version.

The 'new' 5.1 mix is simply a better quality upmix of the same stereo track I'd used before.  Improved channel separation and an edit to the level of one bass effect are the only things that were changed—the stereo source hasn't been altered in any way.  If you have the TeamBlu SW77 that just came out, it can be found on audio track #2.  But since not everyone seems to have it, I can upload it somewhere, sure.

While we're on the subject, I'd like to take a moment to reiterate something I've mentioned before, but bears repeating: if you're listening to my 70mm recreation on a stereo sound system, and not a 5.1 system, significantly better sound quality will be attained by choosing the stereo track instead of the surround.  This is because there is a certain amount of crosstalk between the front and rear channels which even the best upmix can't fully separate, and with a delay being applied to the rears it is inevitable that there will be phase distortion when they are recombined in a downmix.  The technical term for this distortion is 'comb filtering', and it manifests as a strange sort of warbling instability and hollowness in the sound due to the unpredictable ways in which certain frequencies cancel out and others become boosted.  I did my best to minimize the effects of the comb filtering by manually editing the rear channels after upmixing was performed—by shifting the surround delay to 10 milliseconds rather than Prologic II's minimum of 15, and by inverting the polarity of the right surround, I was able to make it sound significantly less distorted when downmixed than it otherwise would have been.  (The inversion of the right rear also had the effect of decorrelating the mono surround effects from each other rather than allowing them to phantom image behind the listener, giving them a more diffuse quality.)  If I hadn't performed these two steps, downmixing the 5.1 track would sound far, far worse than it does now.

So while the downmix I've described does sound acceptable enough for me to not continually fuss over whether the stereo version is or is not included on any given project, it really is much better just to use the stereo track when listening on a 2-channel playback system.  The sound is purer since there is no comb filtering whatsoever, and the lack of an LFE channel is a moot point since this is almost always dropped from downmixed 5.1 anyway.

If I could be absolutely sure that no one was ever going to downmix the upmix, I would have kept it closer to the Prologic II output and not bothered to try to make stereo playback of the 5.1 work.  In such a case I would have insisted upon the stereo version also being included, preferably as the default track on a disc, with the user having to manually select the 5.1 if they had a surround system available.  But since I don't have that level of control over how other people listen to it, some flexibility is in order.

A few months ago there was some discussion of the merits of DTS-HD MA vs Dolby TrueHD for authoring custom Blurays.  Since I'm able to use the Dolby Media Encoder application, TrueHD would have been my first choice, since the format allows for any given surround mix to also include a separate stereo version within the same MLP audio stream, which the decoder would automatically select when the system was set to 2-channel playback.  This would ensure that the proper mix was heard regardless of what kind of sound system was being used, without the end user having to select a separate audio track.  But since TrueHD's implementation on the Bluray format requires that it also contain an interleaved AC3 track for backwards compatibility, and since only a few absurdly expensive authoring programs have the capability to interleave MLP and AC3 in the required fashion, it doesn't seem to be an option.  Alas . . .

Anyway, I'll get to work on uploading the stereo 70 track soon.

Ah, I figured that the improvements may have just been to the upmix but wasn't sure. I do remember you saying before that it is better to have the stereo track than the 5.1 track on a 2.0 system, but I didn't manage to get it. Thanks for the extra info.

There aren't yet, but I'm actually starting work on a new version of ESB right now.

As with SW, the main benefit will be the improved surround decoding.  Further changes from the previous version will be minimal.  I think I'll have another shot at integrating the missing snowspeeder crash sound into the '93 mix to see if I can get it to work more seamlessly.  As yet I'm undecided as to whether the original stereo mix or the '97 SE would be a better source for this (it sounds somewhat different in each).

I'm thinking of reworking the LFE channel somewhat, but I'm not sure how extensively I'd want to change things around, since it already sounds rather good in the previous version.  Unfortunately, the in-theater 70mm recording of ESB is too low quality for me to make out the low end of the mix with much clarity—even when boosting the bass by 18 dB I can still only just barely hear it.  With a hifi subwoofer or headphones I can pick out enough nuances in bass tones that I think I know what the original is supposed to sound like; or at least, I have a very rough idea of what bass was there and what wasn't.  It actually sounds like it's pretty similar to what's already in the '93 mix, which isn't really all that surprising. But even if the content is starting to come into focus a little more, the levels remain a complete mystery, and anything I do is still only a guess at how they might have done it back then.  So in the end I just have to go with what sounds good, as before.