SME

Thanks for the link to the paper. It is a very good read. I do question their Table 10.2 specifying the SPL to calibrate to as a function of room size alone. I believe it doesn't depend on room size at all but rather on other factors that tend to roughly scale with room size. Other considerations include room treatment, speaker distance, and speaker dispersion. I'm guessing that a near-field monitoring setup, even in a very large room, should still be treated as a system in a small room and calibrated to, e.g. 76-78 dB (versus 85 dB for large rooms). So if the mixers followed the guidelines in this table, it's certainly possible that a film could be mastered as much as 9 dB below theatrical reference. Now about so-call near-field mixes. I don't object to their creation, but I'm not convinced that all the changes made in these mixes improve the final result. I also fear they are frequently performed on middling equipment under the assumption that it's "closer to what consumers are using". Let me review the changes made during the near-field mixing process, according to what I've gathered from other forums: Corrections are made to account for the use of discrete surround speakers in most home systems as opposed to the diffuse arrays of surrounds used in theaters and dubbing stages. I fully support this work, and with Atmos coming, dipole surrounds having diffuse sound are quickly becoming obsolete. The dialog level is often boosted. I understand that this is done to improve dialog intelligibility on many real world playback systems, many of which may be badly mis-configured or have severe acoustical problems. Apart from these cases, I think the reason a lot of people complain about dialog intelligibility is that the macro-dynamic range of the track is too high for their system or playback environment. Some people ride their volume controls between quiet and loud parts of the movie while others just resign themselves to not being able to hear what's said on screen. My hunch is that family movies are likely to get the most dialog boost, so the kids can follow along even when they are playing softly. See below for my comments on dynamic range. Overall levels are reduced. Admittedly, it's not 100% clear to me what this statement means. I usually take it to mean that the playback level used during the remixing process was reduced. Another interpretation would be that the level of the digital signal, relative to full-scale, was reduced. More to say on this later. Dynamic range is reduced. This I don't agree with at all. Why does dynamic range get reduced? Nothing in that ATSC article says anything about reducing the dynamic range of a signal when reproducing it in a smaller room. If someone wants to argue that consumer systems can't handle the full theatrical dynamic range, or that consumers will complain, that's fine, but the solution(s) to this problem are dynamic range compression in the playback device and/or a separate "made for TV speakers" mix with substantially reduced dynamics all around. Did I miss anything important? Okay. Let me point out that if dynamic range is intentionally reduced in the "made for home" mix, for whatever reason, the sound quality will be damaged, unless it had poor dynamics to begin with. If the dynamic range reduction is done poorly, clipping can result. If it is done skillfully and the reduction isn't too much, the quality loss may only be subtle on the best of systems. I've heard plenty of soundtracks that I thought sounded compressed but nevertheless sounded very clean and good. What if dynamic range is not intentionally reduced. Can clipping happen in the remixing process? I think the answer is a strong yes. Here's how I can see it happening: The "near-field" mix room and system are calibrated for a reduced playback level, say 79 dB. This level may have been chosen based on any number of factors such as recommendations from Table 10.2 in the article you sent, similar guidelines published elsewhere, or even just hearsay recommendations from others in the industry. The point is if this level was chosen to be too low, the mixers will be inclined to push the faders up to bring things in line with the theatrical presentation, potentially causing things to clip if they don't have softer limiting somewhere in the chain. That's not all. Remember what I said about the frequency response curve? Suppose 79 dB is the right calibration level in this studio for pink noise between 500-2000 Hz, but the bass is too weak at this level without some EQ to boost the bass. So again, the mixers push faders up to bring the bass into proper balance. Unfortunately, the bass is usually the strongest signal present in the soundtrack, so clipping is especially likely when this is done. This can be partly mitigated using bass management tricks like re-directing the bass to different channels. I imagine that this process is often very time consuming, and given the reality of tight schedules and deadlines, compromises are inevitable. Options might include high pass filtering (to free up headroom), accepting some clipping, or just leaving the bass levels reduced. One more thing to mention is that some "mode for home" mixes may be rejected and re-done simply because they don't sound loud enough at "volume setting XX" on someone's AVR. As much as I wish that the madness of the Loudness War did not infect the world of film, it seems unavoidable. ============ As for what kind of weight to use for integrated loudness measurements, I think the K-weighting discussed in the ATSC paper is probably a good place to start. I think it would be interesting to look at loudness and loudness dynamics over a few different time scales. I haven't followed up on the reference for evaluating LKFS yet, so I don't know how readily available that paper is, but the technique appears to be fairly simple while taking into account the bulk of the relevant psychoacoustic phenomena involved in loudness perception.

Thanks for your observations. Yes, almost everything I stated above is speculative and/or subjective. I am also interested to see if I can support or reject any of these statements with good data. I am a programmer and am capable of writing the necessary data analysis tools myself. The main barrier to me doing this is decoding the audio bitstream to PCM from the discs, which may not be possible on my Linux systems. Anything I have to use Windows to do makes me move a lot slower. As for systematically detecting clipping, that's a harder proposition. Harder is proving (or disproving) my hypothesis that a lot of clipping gets introduced in the "made for home" remix process. It would be very interesting to see data that compares the "made for home" and theatrical mixes on those Blu-rays that offer a choice, that is if I could find any such Blu-rays. Edit: I previously misstated that "Oz the Great and Powerful" and "The Fifth Element" have copies of the theatrical mix on the disc. Upon closer review, that does not appear to be the case for either title.

Giving it thought, you are most certainly right that clipping can happen just about anywhere in the production process. If movies didn't have deadlines, perhaps the sound teams could scrub every last instance of it, but that's not reality. That said, I generally notice much less clipping in the mixes I choose (based on subjective loudness) to play at closer to 0 dB. Those that I listen to at around -6 dB may be clean or have clipping only in one or two scenes (e.g., HTTYD). Many of those I listen to at -10 dB are riddled with it (like STID). There are also "skillfully compressed" tracks that are very clean but still get played at closer to -10 dB. (I'm thinking Looper here.) Did anyone else here find "Gravity" to have a particularly loud mix? I was thoroughly disappointed by that soundtrack (and the whole movie for that matter). I played back my Blu-ray copy at -10 dB. At the beginning when things first start to fall apart, the astronauts raise their voices as they panic. I heard what sounded like extensive hard digital clipping in that dialog, which was very distracting. I did not see the theatrical version, so I don't know if the clipping was audible there too. Perhaps the clipping was put there intentionally by the sound team to "simulate" overloading of the radios, but I think the effect would have convinced me a lot better if they'd added some analog qualities to the clipping instead of just digitally over-driving it. Apart from this, all the action scenes sounded very loud, harsh, and compressed to me with no interesting micro-dynamics. Is that what an Academy Award winning soundtrack sounds like? Or did the original soundtrack get the life squished out of it in the remixing process? To pick one more example: the "Prince of Persia: The Sands of Time" Blu-ray. I played this one back at close to "0", but I had to fix a quirk. The center channel level was about 3 dB too high. Presumably this 3 dB boost was done to raise the dialog level relative to the rest of the mix (for the "home environment"), and a remix from stems wasn't possible. In any case, this Blu-ray has a phenomenal soundtrack, clean with excellent macro and micro-dynamics throughout, as is the case with most Disney releases. I did hear one flaw though: Right at the climax of the film, the most passionately spoken syllable of all the dialog in the film clips. With every other utterance sounding crystal clear to me, I imagine the dialog team went through great pains to ensure the dialog track was nigh-well perfect. So it's amusing to speculate that the Blu-ray remixers thwarted their efforts on that one syllable by boosting the center 3 dB without using a soft limiter. Oh well. I admit I may be overly sensitive to clipping. I dabbled a lot in computer music back in the day and I learned the sound of digital clipping very well so I could avoid causing any of it. I can't believe I've seen literature proclaim that signals can be boosted up to 20% into clipping without audible consequences. This is the sort of reasoning that leads to egregious examples like STID. Even compression bothers me when it's over-used. If a loud movie scene gets compressed so much that it loses most of its micro-dynamics, then I'd much rather that scene be quieter but retain the dynamics than be loud and flat.

To me, these are concerns of the playback system and not mastering process. I believe Audyssey's room correction technology even attempts to measure and compensate for room size effects so that "0" sounds correct for a theatrical track. From what I can see, Audyssey calibrates my system closer to the 81 dB (or rather 71 dB for a -30 dBFS rms reference signal) number. There's no need for the studios to try to correct for this in the mix. Most people set the playback level by ear anyway. I disagree that a Blu-ray release with > 30 dB dynamics and 4 or 5 star level is necessarily a theatrical mix. I am sure there are many cases to the contrary. What would happen if you took Elysium and remixed it for playback at -10 dB from reference? I don't know what the overall rms level of the film was, but subjectively speaking, I thought the movie was relatively quiet, even when played back at theatrical reference level. I assume then, that one could remix Elysium for playback at -10 dB from reference while preserving almost all 28 dB of dynamics. Such a mix would rate at 5 stars all-around in contrast to the 4.5 stars the film actually received. Despite these points, I appreciate the simplicity of the measurements and the data presented here. We can all agree that the system isn't perfect, and many titles effectively cheat the system in one way or another, whether by being mixed for lower playback level or containing loads of DC noise. Considering how frequently one encounters these cheating, perhaps it would be helpful to give the subjective "execution" category more weight. One additional number that I'd like to see that need not factor into the rating is the overall RMS level of the track. Better yet, an overall equal-loudness-weighted RMS level would be even more useful. The purpose of this metric would be to give us context for interpreting the bass numbers. I'm not going to get excited about a 4+ star bass movie that also has a 90 dB RMS playback level, knowing that I'll likely play it back at least -10 dB from theatrical reference. Obviously, this wouldn't exactly tell us what playback level the artist intended us to use, but some of us may not want a track at 90 dB RMS whether or not the artist intended it that way. ("The Dark Knight Rises" comes to mind as having an infamously loud soundtrack.) I realize that obtaining weighted RMS levels may slow the workflow too much. Some day when I get around to it, I may try my own hand at analyzing these soundtracks. Right now I lack a Blu-ray drive and appropriate software to get a clean digital rip. If I get around to it, maybe I can add a thread for movie loudness measurements or something.

I don't believe so. This is the only nspace Shuttle Launch recording I've found that isn't either fake (e.g. IMAX Hubble movie) or clipped to death. To my ears on my system, it is totally clean. However, note that if your system cannot adequately resolve the high frequency transients from the solid rocket booster crackle, then it may sound clipped to you.

By the way, now that I'm thinking about it, I have media to submit for measurement. I think this should have no trouble claiming 5 stars all the way around: http://www.digido.com/articles-and-demos12/13-bob-katz/13-we-have-lift-off.html This is a live recording of a NASA Space Shuttle launch at 3 miles away. Be sure to download the 5.1 channel 24-bit 96kHz FLAC version. Note that the correct playback level is +7 relative to theatrical reference! The recording is 4 channel (front-left, front-right, surround-left, surround-right). The 5.1 channel FLAC contains silence for the LFE and center channels. According to Bob Katz, this thing has 119 dB at 25 Hz and 116 dB at 16 Hz and below. On my system with multiple 16 Hz ported Hsu subs (capable of producing clean and audible bass to 12 Hz in-room), I can typically hear and appreciate the difference between 3 and 4 stars extension. While this recording sounds incredible on the system, I have no doubt my system doesn't even begin to do it justice. Those of you with walls of woofers or rotaries will likely be very pleased.

I think what's needed is to weight the response using equal loudness contours (ELCs) to assess how subjectively loud it sounds. Dynamics could be determined using a ratio of the max ELC-weighted response over short-times vs. ELC-weighted long-time response. There's still the question of how long of a window to use for the short-time average. It might be helpful to use a few window sizes with different filters (i.e., low-pass with lower cut-off for longer window) to give fair weighting to the ULF. The tricky part is knowing the proper reference volume for each track. In my mind, that's a more fundamental issue since I'm fairly certain many highly-rated (and some not so highly-rated) bass Blu-rays were mixed with monitoring at lower levels, so that theatrical reference is too loud. For example, Star Trek sounds to me like it was mixed at or very near theatrical reference; whereas, STID sounds like it got mixed closer to -10 (and clipped to death in the process of giving up 10 dB headroom). If this difference in reference volumes were taken into account, STID would end up with an even lower rating! Sad to say, we may never know the correct reference levels for a variety of Blu-ray releases. The best we could try to do is guess by doing some kind of analysis against the dialog. From what I've read out there, both -3 and -6 are common monitor level choices for DVD and Blu-ray mixes. From what I hear (subjective interpretation) listening to media some other recent releases were likely mixed at or near -10 dB. I think Looper may also have been mixed closer to -10. This is a big deal, because if Looper were played back at -10, the deep bass wouldn't be anywhere near as loud subjectively, and it sounds more like a movie with 3 stars extension than one with 5. I'd sure appreciate it if the studio would print the reference level somewhere on the box or at least populate the appropriate metadata like Dialnorm properly. Anyone remember the VHS days when they added the disclaimer at the start of the film: "this film has been modified to fit your TV screen"? I think they should be obligated to add a similar notice whenever they do a separate mix for Blu-ray. It's rather disingenuous in my view to claim that the a lossless "DTS-HD" or "Dolby TrueHD" track is "bit-by-bit" identical with the theatrical master as some of the early Blu-ray advertising implied. Really, I wish they'd just ship the theatrical mixes in the lossless formats and do the "made for the home" mixes with reduced dynamics in lossy Dolby Digital or DTS. My experience has been that the differences between a good lossy codec and the lossless original are most apparent with tracks having very high dynamic headroom. Also, I'd hazard a guess that it is in the process of creating the "made for the home" mixes where a lot of clipping as well as high-pass/high-shelf filtering tends to happen. In many instances the filtering we frequently observe may have been added in that mix process to reduce or prevent clipping.