Yeah... all this is totally not worth splitting hairs over. I care about big differences, usually the result of upstream mastering choices. I consider even the differences between lossless vs. lossy 192 kbps AC-3 compression menial in comparison.
And after all, the whole point of dithering is that at 16 bit you get into the range where quantization errors become audible. The rule wouldn't make much sense otherwise
Or alternatively, rendering out the main audio parts without dithering (muting the patches etc or having them on a muted track) and then exporting the patches etc with dithering, and then just summing both files. though that seems a bit overly cumbersome, heh
Well, rules should be evaluated based on the context. I think dithering does make sense, as the dithering noise sounds better than quantization errors, however when you have a bit-perfect LD PCM track, then dithering is already applied to that, so you end up getting maybe double the amount of dithering and thus also more dithering noise in total. would make more sense accepting some quantization errors in the small patches
Probably both - I don't think I've ever A/Bed a truncated vs. dithered output. I suppose I do it because it's audio 101 that you should always dither when going to 16. It's not a rule I feel especially inclined to break when all this amounts to placebo-level differences.
I would assume - perhaps wrongly, I'm not an expert - that quantization errors shouldn't ever end up being louder than dithering noise, so if dithering noise is inaudible, so should quantization errors from rounding be.
Hmmm fair enough. Though one could argue about whether small patches justify dithering the entire audio file. To clarify, are you saying that you can't hear dithering noise or that you can't distinguish a dithered audio export from an undithered one at all? If the latter, why even bother dithering?
I use headphones for everything too. And when syncing, I always end up patching in bits from another source and that inevitably involves volume-matching - so a dither is necessary.
In any case I think dithering an LD PCM doesn't make any sense. Even if it gets internally processed at 32 bit, a mere syncing will not change the values, so they should just end up getting rounded to the original raw values when outputing at 16 bit again.
since under normal listening conditions even the background noise of your room will likely be louder than the dithering noise, depending on how high the volume is
I used to use MBIT+ religiously for everything until I realised it makes no audible difference to me. Now I just use a simple TPDF dither. Go with what sounds best to you.
I thought it was very noticable when I did some experiments, though I did set the dithering strength to high (using Izotope MBIT+ dither). Different noise shaping resulted in a different sound as well.
Unless you have super hearing, you're unlikely to ever notice the difference. I certainly can't and as I've probably shown I'm rather pedantic about these things.
@Moshrom Curious, why do you suggest dithering to 16bit instead of 24bit? At 16 bit, dithering audibly changes the sound, at least in quiet places, and depending on the dithering settings it ends up sounding different. It would make more sense for preservation work to dither to 24bit, since it will be practically inaudible. Or am I making a mistake?
Yes, I understood. By "output", I meant final output, i.e., a straight conversion to WAV with no syncing/processing intended. And my editor is AviSynth, so even if I were syncing, I could stay in 16 or 24 bit unless I specifically needed to do certain kinds of processing.
I'd like to read that thread. There was even an attempt made to get the dev of MediaInfo to remove that field from its output (for lossy tracks) to stop people asking about it over and over again.
This message that I read questioned all of my beliefs about EAC3to I also thought that we should always decode lossy formats in 24 bits, but maybe the person was wrong, and that libav / FFmpeg is already updated on EAC3to
Thank you, but I just read a recent thread (here or on videohelp, or OT, I don't remember) where the person advised to update (libav I think) so that EAC3to decodes in 24 bits only if necessary, because it tends to decode in 24 bits even when it is not necessary.
There's nothing to update: lossy tracks don't have bit-depth. because they are encoded in the frequency domain. This is why eac3to decodes them to 24bit. You might have better luck with technical questions if you post a thread, rather than using the shoutbox.
@Moshrom Yeah, it's the pre-SE MGM R1 DVD, the one with the very wide AR (2.69:1 according to dvdcompare.net). It sounds all right, from what I remember, but it's been a while since I listened to it. I'll fish it out at the weekend.
To be fair, the only Kubrick on DVD that is master supervised are the original R1 box set LD ports (complete with original mixes), and the R2 EWS Snapcase open matte uncensored. The rest are just revisions on a master work.
Is that the R1 DVD from 1998 - how does it sound? Stephonovich will probably be capturing the Criterion LD once he sorts out his new bit-perfect setup.
Almost certainly. It **might** end up being the "Barry Lyndon" situation, in which the Criterion mono, although not brilliant, is the best of a bad bunch. I'm still hanging on to the original R1 DVD, partly for the AR.
I think that shot had crew and equipment visable in 4:3 but not in the widescreen version. Warner probably wanted to make that shot tighter in the new transfer so that it wouldn't reveal any crew or equipment blunders in 4:3 or 1.78.
Looks like Warner zoomed in and DNR'd that shot in particular when they made the new master. Why DNR and cropping is anyone's guess, considering they're not well known for that as are Universal.
Had Fight Club a new mix for home theatrical releases? On the German BD with the DTS-ES 6.1 Audio it is explicitly mentioned that it is the original theatrical mix.
I'm also looking to find a DVD of Breathless, the remake with Richard Gere, where the open matte original 4/3 DVD version is said to feature extra nudity But those don't seem to be available in used stores.
I'm pretty sure the credits+opening title sequences on the Brosnan films were shot Super 35, so they could be shown in 4:3 without cutting off names or vertically squeezing. Die Another Day allegedly had some sequences shot in Super 35 such as the helicopter shots, and apparently the CG FX shots were rendered in 1.78.