I'd like to talk a bit about one problem (amongst many) I find when mixing films in surround sound:
Bass summing.
It's not a new problem, but as speaker counts have gone up, it's definitely been exacerbated. While the issue is kinda multidimensional, discrepancies in bass level at the point of replay exist partially because of the interaction between:
- Bass management.
- Lack of bass management.
- Speaker coupling / summing in the acoustic domain.
The amount of bass extension (how low the system can go) obviously has an effect too, but I find that less disruptive to the experience than multiple decibel level differences in the 40 to 100Hz range which come about through bass management or lack thereof.
In cinema (theatrical) mixing we don't have to worry about bass management much. The screen channels only go down to around 40Hz, but ya know, so be it. We can dump to the LFE if we need to. In general, the translation room-to-room is decent in regard to "bass summing" as everything stays discreet. Each channel we had in the mix exists as a physical speaker location except for surrounds in Atmos. Sure, sometimes the screen channel speakers couple to each other differently in different rooms, but it doesn't affect things too much.
However, when we move out of that world and into Home Entertainment, things get difficult. The big unknowns are (a) Will they have bass management, and what frequency will they crossover? (b) How many speakers will they have?
The reason these questions matter is that the way low frequency content sums in bass management is different from how it sums in the room. This is because, as most people on this forum will know, in general bass management involves making a mono signal path and slamming together all the "speaker" signals below a certain frequency. In room, the summing is dependent on a few things, but generally the screen channels have more coupling to each other than they do to the surrounds. In all cases though, summing in the signal processor of fully phase coherent material from multiple speakers will result in a level push of [20log(number of speakers)] so 6dB for each doubling of sources. The acoustic summing will be less than this.
At the other end of the scale, if I ran a 40Hz signal 180 degrees out of phase into the left/right pair, someone who's totally in to their bass management region at that frequency would get literally SILENCE as a result of that signal. Without bass management, you'd get a wide stereo bass effect.
Because low frequency content is generally quite coherent across channels (unless the mixer has done something deliberately that isn't, or it's music recorded on a large stage with mics a great distance apart) using bass management therefore in general gives more bass push to a mix as a whole than allowing the main speakers to run full range.
Assuming the consumer is using "speaker based" bass management, the more speakers they have, the bigger the LF push compared to the unmanaged version UNLESS there's only ever "bass" in one speaker at a time in the mix. Now, this is possible. I could brutally brickwall chop all speakers off at 160Hz and put everything below that into the centre channel. (side note: I can't put that content in the ".1" as in most default stereo downmixes the decoder mutes that off, so anyone listening via a 2.0 decode would have no bass at all.) I'm not seriously suggesting this as a solution but it's about the only thing that would be an absolute fix for the issue rather than the fudges we end up doing in the real world.
What I'm talking about here is of course just the basic "maths" at the decoder/bass management stage, so it's all before you even get the signal as a far as "preference" curves on the consumer's line-up, which in some cases compound the error by having more bass tilt-up than studios. Despite personal preference for spectral balance between Re-recording Mixers being relatively similar in my experience, this bass management issue means that mix-to-mix, a consumer will probably find some mixes have loads of LF lumping up and others don't. If you don't use bass management, things might be more consistent from film to film, EXCEPT the mixers have to take in to account that most, but not all, people are bass managing and have to compromise the mix accordingly. And, there's no standards for this yet.
Now, I'm not against bass management. It probably does more good than it does harm. It's just that it's implementation is so unpredictable right now all I can do is try my Home Entertainment mixes without bass management, and bass managed at a random frequency on an random number of speakers (probably 7.1.4 since that's what the clients usually wants it mixed on) and then make compromises to the mix such that neither experience sounds terrible. Generally this means if the bass managed replay gets too LF prominent anywhere, my options are (a) finding some way of de-correlating the offending signal, (b) re-distributing the LF content to a smaller number of speakers or (c) reducing the LF content to bring it back in to the realm of acceptability. The last option (c) being either as a whole sound element if it's an object in Atmos, or by dipping the LF content in some of the speakers if it's a channel-based surround recording.
Help is at hand in the form of "OBJECT BASS MANAGEMENT" in Atmos. This removes the "number of speakers" variable so it's a step in the right direction. So far as I'm aware, not many people are using this in the consumer field yet? Please correct me if I'm wrong.
For your amusement, here's what happened in my mid-field studio with and without bass management, when playing phase coherent filtered pink noise, measuring with a single mic at the mix position. (This was a couple of years ago and I've lost the details of the pink noise but it wasn't heavily sub-sonic. Maybe rolled off 12dB/Oct @ 40 & 180Hz or something and the bass management would have been around 80 or 100Hz).
ACOUSTIC SUMMING (dB) / BASS MANAGED (dB)
L = 74.0 / 74.0
L+R = 77.9 / 80.0
L+C = 79.4 / 80.1
L+C+R = 82.2 / 83.5
LS+RS = 77.2 / untested
L+C+R+LS+RS = 85.1 / 88.1
As you can see, this test only spanned up to 5.0, but as the speaker count increases, so does the discrepancy. By the time you're in 7.1.4 or 9.1.6 the difference is really quite large. Maybe I'll run a real world test of that if I get time...
Other things of note on those numbers, you can see the Left & Centre couple more in this frequency range than the Left & Right, but that's a whole world of room calibration pain that few people venture in to
I don't have those monitors any more or remember much about the calibration, but you can see that the single left speaker playing on it's own was bass managed accurately in level at least. I suspect the whole room was time aligned with some reasonable degree of accuracy to the centre channel. The FR of each speaker probably adjusted individually rather than summed.
/Rant