Star Wars 1977 70mm sound mix recreation [stereo and 5.1 versions now available] (Released) — Page 7

dark_jedi said:

I am just messin with you dude, I probably won't even listen to the stereo tracks that I am putting SO much time into, and I am pretty sure the 5.1 mix will be the default on all 3 because that is the version I will most likely listen to, I do have a very nice high end set up and DD 5.1 all the way for me.

 

No worries.  I always make long-winded replies to offhand jokes - it's kinda my thing.  ;-)

So any news on the "new" ESB mix or are we just going to use the previously made version? just curious because SW and ESB are pretty much ready to be authored now.

satanika said:

Re: upmixing... Really, it is a pretty complicated/technical discussion.. I am no expert by any means. Interested parties can research for themselves matrix/active matrix; dpl, dplii etc. ..

In the end what I did was try various programs/techniques and gave samples for people to check out & decided on the atsurround active matrix decoding.

It should be similar to official dplii decoding. Similar/different for better or worse? Your mileage may vary... But likely minute in any case..

satanika what do you think of this software?

http://www.dts.com/DTS_Audio_Formats/DTS_Neural_Surround/DTS_Neural_Surround_UpMix.aspx

In the course of comparing the ESB '93 and '97 mixes to determine how best to edit them together, I came to an interesting realisation--some of the edits actually work better in the film than I had previously thought.  I'm thinking in particular of the transitions from Dagobah to the Imperial fleet after the cave scene and the X-wing rise; if you listen to the music on its own, the edits would sound bad, but in the film they work well.  The Imperial march theme is already bordering on overuse in the film as it is, and by adding two more bombastic statements it becomes too much, especially since there isn't anything happening onscreen to merit such loud music at those moments.  The more subdued segments that were spliced in call less attention to themselves.

So . . . what that means is that I've changed my mind about using the SE mix to replace those two parts of the soundtrack.  I will still be replacing Boba Fett's departure from Bespin, though, because the editing in that part is really choppy and I don't like hearing it.  It will be arranged in such a way that the only apparent change is the unedited music as heard in the SE, but with none of the dialogue differences (ie no extra Chewie vocalisations when Lando is being strangled, and no extra yells from Luke when he is pwned by Vader in the following scene).

Fans of the original stereo mix will be pleased to know that the 5.1 will also contain the snowspeeder crash sound as heard in the 35mm version.

Great to hear h_h, so is it still a ways off then? if so I will just stop everything for a while and take a break from ALL.

Tech question for you h_h, how can you make all the different PCM files have the same audio level? because some are very low and you have to crank up your receiver to hear them, while others, like the '85 Stereo remix is just blasting away, I want to make these so it is not so jarring when jumping from audio to audio.

Thanks

The Dialogue Normalisation option of AC3 encoding can be used to reduce the playback volume.  The ideal way to determine how much reduction is needed would be to measure the RMS level of the audio files.  For example, a while back I compared the '93 PCM to the mono mix, and found the '93 (and consequently also the 70mm) to have an RMS value of -25.1 db, while the mono measured at -20.5 db.  This means that the mono file will sound 4.6 db louder with the receiver set at the same volume level.  Using the Dialnorm function to reduce the mono version by 4 db or so will compensate for the discrepancy, making them sound closer during quiet sections.  This will also cause the superior dynamics of the 70mm to be much more apparent.

The setting for no reduction is -31, so a 4 db reduction would be obtained by choosing a Dialnorm of -27.  I imagine a similar setting would probably work for the '85 version also.  Make sure, however, that dynamic range compression is not used for any of the encodes.

The Dialnorm function is something that can cause confusion among casual listeners, because the inclination of the ear is often to associate "louder" with being "better".  The AC3 audio on the GOUT is lowered by 4 db compared to the laserdisc PCM, and it is easy to mistake this difference for a reduction in dynamic range.

OK now I am more confused, how do you measure RMS? and what did you mean by this?

"Make sure, however, that dynamic range compression is not used for any of the encodes."

if I don't figure this out I guess I could just leave it as is and people can just raise and lower the volume on their receivers.

Dynamic range compression makes the loud parts quieter and the quiet parts louder.  You do not want to do this.

Changing the dialnorm setting simply changes the volume before it gets to the receiver/TV.  It's basically the same as changing the volume on the TV, but it does it before it reaches the TV.  (If I understand HH correctly.)

The idea is you need to find out the overall volume level of each track, and adjust the dialnorm setting for each track so that they all output at the same overall volume level.  I assume "RMS" is the same as overall volume.

I don't know how to measure it, though, HH should be able to be more help with that.

RMS stands for "root mean square".  Rather than a simple average, it involves finding the average of the squares of the values being measured, then taking the square root of this value.  I believe Dolby recommends using this method for Dialnorm because it more accurately reflects the characteristics of human hearing.

I used a program called Wave Editor to find the RMS values when comparing the '93 mix to the mono, though this is only available for Mac.  But no doubt there are pc programs that can perform this function as well.

However, it probably isn't necessary to actually go to the trouble of measuring them.  Using a Dialnorm setting of -27 for the mono mix and the '85 will give a 4 decibel reduction which should bring them closer to the level of the existing stereo and 70mm mixes dark_jedi already has.  It won't be exact, but they will be less obviously loud when switching audio tracks on the disc, which is the goal of using Dialnorm in this case.  As long as they're pretty close, it shouldn't be a problem.

I had it set to -27 already when testing, the '85 Stereo Remix is just louder\fuller for the lack of a better term then the rest of the mixes, especially in the dialogue, should I just drop it a little more or just leave it? but I am guessing with the way it is now, when jumping from audio to audio, you will definitely notice this baby.

also all my searches are coming up with nothing, what should I google to be able to compare?

satanika said:

If you have wavelab then it can measure rms for you I think, but it's probably faster to just try a few different values for dialnorm like e.g. 23 and 19 etc.
Or you could use whatever software that supports replaygain to compare the soundtracks..

..

I don't/didn't know that dts neural upmix proggie, but I'm not opposed to trying new upmixing methods if anyone feels it's warranted, are there people interested in doing a new listening test?

Is the new listening test for the new ESB DD 5.1 that hairy_hen is doing?

The spitting sound was something I edited out on a whim last year for a previous soundtrack attempt, and the edit seems to have found its way into this version also.  I'm surprised you noticed!

The spitting was obviously a mistake--they must not have noticed that it was a part of the dialogue track until after the mix was complete.  They had Mark Hamill redub that line for the mono version, and this error is not heard in that mix.  Why it was never removed from the SE mixes I don't know--guess they forgot and didn't notice again.  It had sort of bugged me after Darth Editous pointed it out a few years ago, so I deleted it by looping a section of background noise.  Not having it in the film is no great loss; I'm certainly not going to produce a new version just to put it back in, lol.  But yeah, I guess the oversight does make it slightly inauthentic for that brief moment, alas.

I've done some comparing of Dolby Prologic II and DTS Neo:6 as output by my receiver, and for the most part I'm hard pressed to hear any significant difference between them when not listening ultra-critically.  The DTS version does have a greater degree of channel separation, though it comes at the expense of the sound field being slightly less 'stable'.  But they both provide pleasing results with matrix-encoded Dolby Surround tracks, and it wouldn't be easy to tell which was being used in a blind test.  For non-matrixed stereo, the results vary by the nature of the source material.

lol, I guess that makes sense about the hair.

Were you able to get a new receiver yet?  I hope you'll be pleased with the 5.1 mix.  ;)

Is that the only edit you made to the track beyond conforming it to the 70mm mix, or were there others?  Just wondering.

No other differences, aside from the narrower imaging and flattened dynamics associated with the 35mm replacements.  It's also possible that some instances of LFE bass may not have been in the original, or were mixed in a different way, though it's impossible to say for sure, of course.  In any case they sound like they could have been there, or should have.

Actually, I wonder if it would have been better to use the '85 mix for the replacements instead of the '77.  It sounds like it is an exact transfer of the 35mm version (with C-3PO's extra line dubbed on top of it), but it is reported to have wider stereo imaging than the original.  I haven't actually heard it in a while (never bothered to obtain a pcm copy because I didn't think I needed it before).  Seems likely the wider image would have been obtained through some kind of post-processing, so it's probably not an exact match to the '93, but it might have been closer and made it a bit more seamless.  Oh well.

The imaging discrepancy with the 35mm segments is more noticeable in upmixing than in 2-channel, because sometimes you can clearly discern that more information is coming from the center and less from the sides and surrounds.  I suppose if your speakers are all perfectly voice-matched (which they really should be) it would stand out less.  Still, it's only really apparent for a few of the edits, and it's a small problem in the scheme of things.  Your mileage may vary, of course, but I tend to find it less objectionable than some of the extra junk foisted on the '93 mix.

Interesting.  I've seen RMS measurement suggested elsewhere as an indication for setting Dialnorm, but I don't really know that much about the specifics.  Commercial dvd's are all over the place with this sort of thing too--I think a lot of them just reduce everything by 4 db without measuring, which is especially irritating if the material is already encoded at a lower level to begin with.  The GOUT audio has certainly been subjected to this, which is the main reason it seems anemic compared to the laserdiscs.  Dialnorm can be a useful device, which Dolby had good intentions in producing, but unfortunately it is too often misunderstood and misused in the real world.  An example their documentation provides is reducing the volume of commercials so they don't sound blaringly loud compared to programs with lower averages, but I don't think it has ever actually been used for this purpose!

When combining segments of the '77 stereo mix with the '93, adjustment was required nearly every time to get them to blend.  I used Belbucus' level-matched version, which plays at a similar volume to the '93, but even that was not identical because of the differences in dynamic range.  Most of the time reducing the '77 mix by 1 db provided the best results, but sometimes even if the dialogue came into alignment the background sounds wouldn't be the same.  Louder sections were raised anywhere from 2 to 4 db to try to compensate for the dynamic shortcomings, but while the results are certainly listenable, the more powerful original can obviously not be recreated.  Strangely, one especially quiet moment had to be raised 6 db in order to fit.  This was done from looking at the waveforms, with further fine-tuning by ear as needed.

You'll never get an exact level match between the various mixes.  Dialnorm can certainly help to even them out, but it's only an approximation.  Still, as long as the averages are in the same general area, that's what matters.

I do think the 70mm mix probably had more dynamic range than its '93 downmix, owing to having multiple discrete channels, and evidenced by the occasional brief clipping of the '93.  16-bit pcm has a pretty good range, but sometimes it does fall a bit short of what is needed to convey sound without compromise.  Present day 5.1 tracks have dynamic potential more akin to a 20 bit range, I believe.