Star Wars 1977 70mm sound mix recreation [stereo and 5.1 versions now available] (Released) — Page 4

I would say the standard 448 would be enough, we will get out HD audio in October 2011 when the Blu-ray set comes out.

Did anyone here get hairy_hens FLAC file decoded to wav to make a ac3 file? no matter what I do I can't get this file to work so I can't test it against the 2.0 file, I used FLAC frontend to decode but softencode will NOT encode this at all, keeps shutting down on me, I also tried wav to ac3 and it won't even start, so any pointers are very much welcomed at this point, but at least I tested the 5.1 file made from this and it is AWESOME!

The problem is on my end, somehow I can't do anything with ac3 files now, nothing works and I'll be damned what happened, I am really getting sick of this Windows 7 Pro 64 bit.

Molly said:

For me, *all* 5.1 sounds more faint than 2.0.

I really have no idea what this means . . .

Do you have a 5.1 setup, or are you listening to them downmixed to stereo?  If so, Dolby Digital decoders when downmixing automatically engage non-defeatable dynamic range compression in accordance with the DRC profile specified in the AC3 bitstream, the end result of which is a hideous reduction in the power of the soundtrack.  This is all well and good if you are listening on crappy tv speakers that can't handle high peak volumes (which Dolby seems to assume, rightly or wrongly, to apply to nearly all listeners), but is completely unnecessary if you have a capable stereo setup.  The fact that this DRC could not be turned off was the bane of my movie-watching existence for two years, until I was able to add a center and rear speakers to my system.  Anything less than all five speakers will result in DRC (the presence or lack of a subwoofer makes no difference).  When I was finally able to demonstrate to my lady friend what I'd been griping about all that time, she was quite startled to say the least--the difference is not subtle.

So if this is what you're refering to, then unfortunately the only solution is to get full 5.1, or to choose stereo or DTS tracks when available, since they do not have this issue.  If you meant something else . . . well, then I can't help you. ;)

I don't know why d_j is having problems getting a successful AC3 encode from the uploaded stereo track--is anyone else having similar issues?  Maybe it's some kind of Mac/PC incompatibility?  I can do my own AC3 2.0 track and upload that if need be.

I agree that 448 AC3 for the 5.1 would be best for most folks.  Some part of me wants to have DTS and higher bitrates, though my speakers may not be hifi enough to actually resolve such that kind of subtle difference; and I'm not sure how much benefit there really is, since there is occasionally a certain harshness to the sound, most likely because of the age of the material.  But in theory preserving as much fidelity as possible is always a good thing for those who can take advantage of it.

I really couldn't give less of a crappe about the films coming out on Bluray unless the original versions are included and restored.  And even if they did, it's not at all certain they would master the audio from the 70mm mix--no matter how good the picture was, I might well want to mux it with my own soundtrack, unless by some chance they actually did come through and do it properly, and how much chance is there of that, really?  I certainly am not remotely interested in having the rubbishy 2004 audio mix in glorious Dolby TrueHD . . .

The chances of the BluRay release being what we want are slim to none, so I say bring on a fan reconstruction (and yeah, not just a remix of the 2004 crap).

OK Thanks to Dr. M I have fixed my issue, I have it encoded and will do some testing in a couple of days and report back, but I believe that that both the 2.0 and 5.1 files will be just fine, sound wise and volume wise.

I mean liba52dec downmixing.

That will do it, definitely.  It is very rare to find an AC3 decoder that can be made to disregard the DRC profile when downmixing to anything less than the left, right, center, left surround, right surround layout.  In fact, I would be surprised if such a thing existed at all outside of pro equipment.  I spent a lot of time scouring through avsforum to find similar reports, and the explanation (and Dolby's own documentation) points solidly to the notion of forced dynamic flattening on the downmix.  Ostensibly to avoid overloading the digital to analog converters, but I really can't see how that would happen if 24-bit DACs are used, unless the DAC counts its volume scale down from maximum and not up from the noise floor, so that even 16-bit dynamics could be pushed into clipping--but I don't know enough about that to comment in an informed way.  Even so, I don't think I've ever experienced any problems with clipping when downmixing high volume DTS tracks into stereo, so it really seems like Dolby should have allowed for this 'helpful feature' to be turned off.  Oh well.

As for Dialogue Normalisation, this is not defeatable under any circumstances, and is often a source of confusion about the dynamic range of a soundtrack.  It is much easier to tell when, say, a loud LFE bass sound or other high volume peak is reduced than it is hear that spoken dialogue and other relatively quiet sounds average at a lower level--the ear becomes accustomed to it, and the perception of 'loud' or 'soft' varies accordingly.  Witness numerous reports of the dynamic superiority of the 1993 laserdisc mixes compared to the GOUT: while it is true that the GOUT version of Empire is inexplicably flattened (Belbecus commented on this when he first made the PCM available, and I can confirm it), the other two films are not--the only difference is a 4 db gain reduction most likely attributable to Dialnorm.  Adjust the playback volume to compensate, and I'll wager it would be much more difficult to tell the difference between the level-matched AC3 audio vs. uncompressed.

These factors are part of why I haven't wanted to make the 5.1 version generally available yet, because I haven't been able to test it and see exactly why the volume is reduced.  Satanika assures me that no DRC or Dialnorm was used in the AC3 encoding, so I'll have to see what I can find out.

What were the main changes you re-created with this mix?

I've given a fairly thorough accounting of the work involved throughout this thread, but in short, the main source was the 1993 laserdisc mix, from the GOUT-synched PCM provided by Belbecus, which is essentially a 2-channel downmix of the 70mm soundtrack with additional sound effects added in.  I removed these extraneous sounds (some of which are harmless while others prove distractingly unnecessary) using the 35mm stereo mix, also provided by Belbecus.  Since the 70mm featured surround sound, and undoubtedly contained more bass than the 1993 version could accomodate, I put together a custom LFE track for the 5.1 version, the majority of which was sourced from the 1997 special edition, with some portions also derived from the 2004 dvd if I felt they worked better, and on rare occasions even isolated and amplified from the 1993 mix itself, for places where neither special edition proved satisfactory.  The volume levels of these added bass effects were individually adjusted to provide an ideal blend with the older soundtrack, and to create as enjoyable an auditory experience as possible.  Satanika took care of the upmixing from stereo to surround, since I can't get that to work on my computer.

See here for the list of changes made to the 1993 mix: http://originaltrilogy.com/forum/topic.cfm/70mm-mix-to-93-mix-comparison/topic/6501/

I've finally been able to listen to the 5.1 version on my sound system.  The low volume level issues I was having earlier don't seem to apply when the file is played in its native format; I was about three-quarters convinced of that already, but having looked at the waveform and listened in 5.1 mode I'm now certain that there is no DRC as I had feared.  The average volume is the same too, so no dialnorm or anything like that going on.  A very small amount of clipping occasionally occurs in the center channel on the highest peak volumes, but like the clipping I documented earlier in the '93 source, it is very brief and quite inaudible.

For whatever reason, when playing the file in VLC via headphones (which was all I could do away from home), the volume was substantially reduced, but during proper 5.1 playback this does not occur.  Add that to the potential phase issues that may arise when downmixing something that has already been upmixed, and I will add the caveat that this 5.1 mix ought not to be played without a 5.1 system.  For ideal results the stereo version should be used if such a system is not available.

Though the problems I thought were present turned out to be a false alarm, in the course of listening to the 5.1 file I discovered another issue, one which I confess has me somewhat baffled.  I made only brief mention of it earlier in this thread, but on a few occasions I used the bass effects from the 1993 mix for the LFE channel, isolating them from the rest of the mix and adjusting to an appropriate volume, in short amplifying what was already present in the '93 mix.  On my 2.1 AC3 encode, they sound the way I wanted them to sound, but on the 5.1, they seem by varying degrees to be subdued, being only marginally louder than the '93 if at all.  I don't know what is causing this discrepancy.

I have heard that AC3 encoding, being lossy in nature, may alter the phase relationships between the channels, so that what should blend together becomes less coherent.  The likelihood that this would occur with the LFE channel is greater, because lossy encoding generally devotes more of its bandwidth to the midrange, where human hearing is the most sensitive, and less to the high and low frequencies.  Bass frequencies are also more likely to interact with each other and their acoustic environment in anomalous ways.  But I can't say specifically what is causing this problem in the mix--whether it is the lossy encode of duplicated bass, or whether the upmixing caused some cancellation, or even if my receiver is simply processing the two versions differently.  The reason I used a few '93 bass effects was because I felt that neither special edition worked as well for those sounds, but it looks like I'm going to have to make the SE bass work if I want it to sound right.  Should have realised something like this could happen, but it's an area that is really beyond my knowledge, since I am not a trained professional.  Blast . . .

I don't think anyone but me would have noticed the discrepancy, since I'm the one who put this thing together (and my subwoofer is of higher fidelity than what many people generally own, indeed it is probably the highest quality component in my system), but I want this to sound as good as I can make it.  Hopefully it won't take long to fix.

I haven't yet done that, but it is an excellent idea and I will certainly do it as soon as possible.  I suspect it is a phase issue, because when I looked at the waveform of the LFE channel, after decoding to six mono files, it appeared to be the same as the source.  I have my receiver set for a crossover of 80 hz, so that everything below that in the main channels is diverted to the subwoofer.  This filtered bass may be partially cancelling itself out when combined with the LFE channel in these instances.  In addition to just playing the subwoofer without the main speakers, I'm also going to switch the settings so that the mains play full range and the subwoofer only plays the LFE channel and see what happens then.  It's possible that it could be a quirk of the speaker distance settings in the receiver, and that the results would be different in another room and with different equipment.

Listened to it some more and I'm pretty certain my receiver has nothing to do with it--the results are the same no matter how the settings are changed.  It has to be in the encode somewhere, but I still can't tell exactly what is responsible.  I'd estimate it's at least a 3 db difference.  Perhaps if I put more of these bass effects in the LFE channel and increase the levels I can make them come out where they should be.  Hmm . . .

Well, I'm still not 100% sure of the cause, but I've managed to solve the problem of reduced bass.  Took a while for it to occur to me, but it turned out the key was to invert the phase of the duplicated bass in the LFE channel--once I did that and re-encoded to 5.1, it sounded much closer to the stereo + LFE file I've been using for comparison.  I'm presuming that somewhere in the upmix process, the phase of these bass effects became reversed, so that it was partially cancelling itself out.  Now it combines the way it should.  The entire principle of spreading a stereo source over more speakers is based on phasing tricks, adding and subtracting the channels together in various ways, so I suppose it shouldn't come as a surprise that this could happen.

So, now that it's taken care of, expect to see the final 5.1 version made available soon!  It took a bit longer than I expected, but you can be assured it will all sound exactly as it should.

I've double-checked once again just to be sure everything sounds the way it should, and I'm pleased to report that the 5.1 mix is now entirely finished and ready for general use.  Here are the download links from Satanika--an AC3 of 448 kbps split into two RAR files.  It should be all set to mux with any GOUT-derived video, whether you use the video as is or with your choice of AviSynth script processing.

Part 1: http://www.sendspace.com/file/qq11vb

Part 2: http://www.sendspace.com/file/jd98ys

If any of the links for this or the stereo version expire, just let me know and I'll be happy to re-upload them.  I've also updated the first post to include the download links for both so they'll be easy to find.

With two different versions of the soundtrack, there may be some potential confusion about differences between them and which one would best serve a particular type of audio setup.  I'll try to make that clear here.

The original 70mm mix featured discrete channels in what we would now probably call a 4.2 layout, with three front screen channels, a mono surround channel driving an array of speakers in the sides and back, and two channels exclusively reserved for low frequency effects, driving whatever speakers or subwoofers a particular theatre was equipped with.  The laserdisc mixes from which this recreation is derived are in a matrixed stereo format, with the four main channels folded down into two, encoded in such a way that it can be expanded or "upmixed" back into surround.  Note however that this upmixing cannot fully separate the channels from each other, and a certain amount of crosstalk between them will exist regardless of what algorithm is used.  The original discrete channel format cannot be fully recovered without an actual transfer from the source, only approximated.

Such upmixing is normally carried out by the receiver during playback, and the stereo version heard this way will deliver excellent results.  The bass content, however, will be limited to what is present in the laserdisc mix (which is not insignificant).  The 5.1 version was made in order to allow use of the LFE channel, since this was an important part of the 70mm theatrical experience.  The upmixing took place in software, with the ATSurround plugin of Foobar 2000 I believe (correct me if I misremembered that), and the LFE channel added in afterwards.  The surround experience is very similar to the results from Prologic II, and the added bass gives considerable impact to explosions and spaceship flybys, as well as some more subtle low end.  Surround effects emanate equally from both rear speakers, with additional rear stereo ambience derived from the mains.

Which one you should use depends what kind of sound system is being used.  I'll give some general guidelines:

If you are listening via 2-channel analog connection, be it the headphone jack of a computer or the stereo out of a dvd player, choose the stereo mix.  This applies regardless of whether you are using your tv speakers or large floorstanding tower speakers, and whether you have a surround setup or not.

If you have a 5.1 setup, and are bitstreaming the AC3 signal to a Dolby Digital decoder using a digital connection, or are using a multi-channel analog connection, you should definitely choose the 5.1 version.

If you use a digital connection but do not have a full 5.1 system, you should most likely choose the stereo version.  Use your choice of upmixing algorithms in your receiver to match the signal to your speaker layout if you wish.

If you have a subwoofer but less than five speakers, you may achieve acceptable results from using the 5.1 mix, but be aware that by downmixing a file that has already been upmixed from stereo, the quality of sound from the main channels will most likely be inferior to what you will hear from the stereo mix, due to phase artefacts and channel crosstalk.  Were the original discrete mix available, it could be downmixed without any significant problems, but that is not the case with this version.

(If you really want to, you could encode your own AC3 in 2.1 layout using the stereo and LFE tracks I've provided, but know going in that the playback of such a format is erratic, depending entirely on the equipment used.  My own Onkyo receiver discards the LFE channel when using Prologic II, retaining it only in stereo and DTS Neo:6 modes.  A high-end Denon receiver at a hifi home theatre store was completely incapable of using the LFE channel at all, regardless of what listening mode was used; a similarly priced Yamaha successfully used the LFE while automatically engaging THX Surround Decode, and worked in stereo as well.  I make no guarantees about going this route.)

I don't know what the surround effect of this mix is like on a 6.1 or 7.1 system, through Prologic IIx or similar algorithms.  Ideally the rear channel effects would be spread equally among all the back speakers, to simulate the monaural surround of the original, rather than being relegated to the center back speakers only, but I can't say one way or the other, having never heard it this way.

Anyway, I've gone on a bit, but hopefully these guidelines will be helpful in creating an ideal listening experience for the film.  I really hope everyone who obtains this mix will be pleased with it, and I'm very interested in hearing any opinions and feedback you might have.  Have to say I'm rather relieved at finally being finished with this--hopefully now I can just watch the film without worrying too much about the sound not being quite what I want it to be!  However close this comes to simulating the real 70mm mix, if I've improved the available sonic experience of the film for anyone, through the sound itself or from providing a look back into the film's history, then I've done what I set out to do.

Hairy_Hen--this is a really amazing piece of work. People tend to ignore audio or accept lesser of it because it seems less "tangible" than images, especially with the iPod generation that has accustomed people to low-quality audio while high-def picture is all the rage, but I'm glad there are a few audiophiles to match all the videophiles around here. This is the most interesting project since the mono mix restoration and I hope now that people will stop including GOUT and 2004 mixes on all future releases. I'm still downloading the files, but good work in principle! :p

I openly acknowledge that Belbecus' skill and knowledge in audio far exceeds mine, and the kind of work necessary to perform something like the mono mix restoration is beyond the tools and the ability available to me.  Consequently there are a few places where this 70mm recreation falls short of being completely seamless, mainly because while I can hear the parts that don't quite match up, I'm not able to make them sound closer.  Fortunately those parts are infrequent and not readily noticeable except under close scrutiny.  But I appreciate the compliments.  :)