How much sub 200Hz content is really found on the surround channels?

[3ll3d00d]

Does anyone have any objective data on this?

 

The reason I am is because I am building some new surrounds and I am trying to judge how much excursion I really need from them.

[Infrasonic]

I can't think of any objective data on the subject, unfortunately. Would be nice if there was an automated method for breaking down the percentage of usage in multichannel movie content.

 

There is quite a bit of full bandwidth and/or full SPL content in the surrounds. I would build surround to take whatever bandwidth you require of them and all the SPL capability you will use. Pretty easy to figure out.

[3ll3d00d]

I would build surround to take whatever bandwidth you require of them and all the SPL capability you will use. Pretty easy to figure out.

yeah so I have done that analysis and my current implementation is arguably just enough but only arguably so If many films have substantial 60-160Hz content in the surrounds then I could (should?) lean towards another approach but if such content is unusual then it's probably not worth it (as the more substantial model is not such a nice fit in the room and costs more). 

[Infrasonic]

Depends on what kind of movies you watch.

 

It's tough for someone else to tell you whether you should or not. Many people think you should have identical speakers all around if possible. It's obvious that isn't exactly necessary. Most people have large mains and small center and surrounds. I'd advise movie watchers to treat their CC exactly the same as their mains. Then move on from there. There are some pretty potent surround speakers around that aren't large. Check out the Volt series DIYSG speakers. They can handle reference at those frequencies.

 

Also, you're being very reasonable with the 60-160hz listing. There is PLENTY of content in surrounds in that bandwidth. Even height speaker content has high level (albeit seldom) stuff to take good advantage of.

[maxmercy]

Depends on the film.  Many of the better bass films contain significant material down to <10Hz in the surrounds.  Iron Man, Transformers 1, Batman Begins, The Dark Knight come to  mind.  I know this because I have graphed/analyzed the surround channels as well as LCR and LFE when I determine BEQ correction for films.

 

JSS

[SME]

I tend to agree that anything goes in the surrounds, even if they tend to get less bass "on average".  If in doubt, I recall that I have that 4 channel (2 front + 2 surround) Space Shuttle Liftoff recording, which has crazy amounts of ULF in all four channels and is meant to be played at +5 above (85 dBC) reference.  I'm guessing that the 80 Hz HPF part of the track is lower than WCS for reference level, but I'm not certain.  Some of the transient crackles may well have a lot of energy above there.

 

Another question is whether reference level playback for surrounds is even enough.  I have only 2 surrounds, so any 7.1 track could potentially hit me with peaks that are 6 dB higher.  For Atmos down-mixes (including "Atmos for home" spatially-encoded) that are done well and preserve as much dynamic range of the original Atmos track as possible without using any kind of compression/limiting, playback at levels higher than reference may be required to achieve the "right" loudness.  I'm not sure how much this will happen (if ever) in the real world, but I do believe there are BDs out there, in the live music genre if nowhere else, that just beg to be played above reference.

[3ll3d00d]

Thanks for the info. It seems that no short cuts are on offer here after all

 

The design I have (a sealed radian 5208) runs out of xmax (4.5mm) at reference at ~80Hz, my design goal is -5 so that cuts excursion by ~2mm and it is wall mounted so that gives me a bit more output too (haven't calculated this accurately yet). If I port it then this cuts excursion in half for the same SPL hence it is just a question of how hard you can run a coax. I suspect what i have is sufficient and I am overthinking it.

 

@SME I don't follow your point about peaks being 6dB higher going from 7.1 to 5.1. I know there are different approaches to downmixing from 7.1 to 5.1, ie different ways to redistribute the extra channels, but 0dBFS is still 0dBFS. Ultimately there is no ideal solution to that problem as far as I am aware, some channel specific content will be too quiet but just turning it up will make other sounds, that were distributed across channels, too loud. Don't you have room for 7.1 anyway? Your room sounds like it is fairly large.

[SME]

In your assessment of excursion, are you considering the effects of the bass-management HPF?  The required excursion at 80 Hz with a 4th order filter should be half as much as without.  Wall mounting won't really reduce excursion unless the models you are using include some kind of baffle step compensation.  Models usually assume that you are flush mounting to an infinite baffle already.  However, if you are mounting it close to the ceiling or floor, you will gain that extra benefit.  A ported design can definitely reduce low-end excursion by *a lot*, but only if the design works with the drivers you chose within your space constraints.  I don't know that you are over-thinking anything here.  I think people routinely compromise on their surround performance, due to both budget and space constraints.  That's not an unreasonable decision to make, but I believe it should still be acknowledged as less than ideal rather than making shaky arguments about how "surrounds don't have to deal with as much as the mains".  This is only really assured with theatrical mixes where FS for the surround channels is -3 dB that of the fronts.  (Although, I think this may have also changed with theatrical Atmos.)

 

If I play back a 7.1 track on my 5.1 system, the surround channels get will mixed down in my AVR before playback.  The rear channels in the mix probably go 100% to the surrounds.  The side channels get split between the surrounds and front left/right, but most signal probably still gets sent to the surrounds.  So the increase in WCS peak demand is probably not quite 6 dB, but close.  And 0 dBFS in the soundtrack is not 0 dBFS in the AVR, and so long as the AVR has sufficient headroom, there's no issue down-mixing 7.1 to 5.1 as long as the speakers can play it.

 

My room is large as far as bass is concerned, but the layout is unusual for a HT/listening room.  Much of the sofa butts up against a wall about 9 feet wide, so the only area I could install rear speakers is on the ceiling directly above it.  But that would be pretty pointless as these "rear" speakers would still be in direct line with the "side surrounds".  This location would make more sense for overhead Atmos speakers, but I don't plan to go there until I have a truly dedicated room.  (I want to install bass traps where those overheads would go, too.)

[3ll3d00d]

Yes I've included the HPF in that. 

 

I still don't see your point about a 6dB increase in WCS signal, seems to indicate worst case of coherent content on both channels is folded into 1 channel without any attenuation. I would be surprised if it didn't attenuate during this process and I'd be surprised if there was coherent output on each channel anyway (would indicate that the channel was simply copied). However I've been surprised before so who knows 

 

FWIW Google suggests the dolby plus method is 

 

Ls = [(Lss 0dB) + (Lsr -1.2dB) + (Rsr -6dB)]

Rs = [(Rss 0dB) + (Rsr -1.2dB) + (Lsr -6dB)]

 

which makes for a +3dB bump worst case

 

I suppose it depends on exactly what your AVR is doing (which may or may not be the same from one AVR to the next)

[SME]

You are right that my analysis was oversimplified, but my calculations suggest I underestimated rather than overestimated:

 

    LS_level = RS_level = 20*log10( 10^(0/20) + 10^(-1.2/20) + 10^(-6/20) )

                   = 20*log10( 1 + 0.871 + 0.501 )

                   = +7.5 dB

 

I am very confident in my math, but please let me know if I'm missing something.  I have an easier time believing the WCS is closer to +3 dB than +7.5 dB.  I believe Dolby's approach has always been to ensure *sound power* is approximately equal in the mix-down.  For example, when mixing a discrete center channel to front left and front right, Dolby (IIRC) recommends -3 dB attenuation before mixing to each.  To the extent that the fronts combine in completely phase, some sound may be presented up to 3 dB hotter than in the discrete track.  However, if the fronts are assumed to be uncorrelated (aka decorrelated), then then -3 dB to each channel cancels and the overall level change is 0 dB, at least on average.  It's on average because uncorrelation is a statistical quality in which phase differences are assumed to always be completely random.)  It is a reasonable first-pass assumption that the sound arriving from multiple speakers is uncorrellated in typical listening conditions, even if the signal being sent to them is correlated.  (This assumption breaks down in a few interesting ways, as I'll get to later.)  In general, to determine how uncorrelated sources combine, we sum their *sound power* output instead of their *sound pressure* output.

 

Sound power, when expressed in dB, doubles every 3 dB instead of every 6 dB as sound pressure level does.  It's just like amp power.  Just remember that, all else the same, power always doubles with +3 dB.  If you play sound at the same level through two uncorrelated speakers instead of one, the sound power doubles, a gain of +3 dB.  If you apply a gain of -3 dB before mixing to the fronts, you reduce the sound power output by from each speaker by half, and with both speakers playing, the result is a wash.  You might call this the principle of sound power conservation, and my observation is that Dolby follows this principle in its matrixing/down-mixing recommendations.

 

So what does the WCS look like for each FL and FR with a mixed-down discrete center all at WCS?  Here it goes:

 

    FR_power = FL_power = 20*log20( 10^(FL_dB/20) + 10^(FC_dB/20) )

                                             = 20*log20( 10^(0/20) + 10^(-3/20) )

                                             = 20*log10( 1 + 0.708 )

                                             = +4.65 dB

 

So even though the average sound power in the mix remains constant and the demands are spread out across two speakers, the WCS peak is actually quite a bit more than +3 dB.  Now let's look at the more complicated case of the surrounds.  First, let us verify that sound power is "conserved" in the down-mix with the coefficients you gave:

 

    RS_power = LS_power = 10*log10( 10^(LSS_dB/10) + 10^(LRS_dB/10) + 10^(RRS_dB/10) )

                                             = 10*log10( 10^(0/10) + 10^(-1.2/10) + 10^(-6/10) )

                                             = 10*log10( 1 + 0.759 + 0.1)

                                             = +3.03 dB

 

What's going on here?  Well, we lost a rear speaker on each side, and the each "side" surround (abuse of terminology) now has to pick up the slack.  In order to conserve sound power for the down-mix, the sound power output for each surround increases by +3 dB.  Yup, the power balance is there, but that surround is taxed with not +3 dB, not +6 dB, but +7.5 dB of peak SPL capability?  Wow!

 

I'll just point out that the same problem arises when trying to mix theatrical Atmos tracks to a home format.  The problem is there whether or not the track is intended for Atmos in-the-home playback.  Either way, the soundtrack is no more than 7.1 on the disc.  The Atmos objects are spatially encoded into those 7.1 tracks just like center and surround track were mixed down into Dolby Stereo and Dolby Pro Logic tracks back in the day.  The same factors basically come into play there but on a much larger scale because of the enormous number of objects that Atmos can support.  The actual peak SPL required for an down-mixed Atmos track that's otherwise-unaltered is likely rather incredible for those busier mixes that featured upwards of 100 or more objects in the theatrical mix, and this is true even though the loudness (a property more closely related to sound power than SPL) on the track would, in theory, remain the same.  Even an Atmos home mix done at a proper 85 dBC might require considerable dynamic range reduction to avoid clipping.

 

So I guess if you have enough amp power to bottom your surrounds, you'll probably want a limiter.    Oh, and as I said earlier, the assumption that even correlated signals sent to multiple speakers become uncorrelated before they arrive at the listener breaks down pretty badly.  On a system with decent bass response (and no serious acoustic problems), correlated bass signals are much more likely to still be correlated after they arrive at the listener.  And of course if you throw in bass management, the correlated signals get summed back in the processor instead of at the listening position, and so the result remains completely correlated.  This leads to an even more bizarre conclusion: The bass demands for playing a 7.1 track on a 5.1 system are greater than for playing a 7.1 track on a 7.1 system.  The WCS is higher.  I think there may be exceptions, but I believe bass management is almost always done *after* the down-mixing.  So what does the 7.1 on 5.1 WCS look like?  Let's see:

 

WCS = 20*log10( 3*10^(105/20) + 2*10^(112.5/20) + 10^(115/20) )

          = 125.8 dB

 

The increase in WCS isn't as big as I thought it would be (compare to 125.1 dB for 7.1 on 7.1), but still, talk about ironic!

[3ll3d00d]

Thanks for the clarification, I was just thinking in sound power terms when I said 3dB.

 

The question remains how this translates into a real world outcome. As we've seen with bass management, devices have finite headroom both digitally and in the analogue stage so how it treats such a signal will be implementation specific. Does it just let that increased signal through thus exposing the end user to a theoretical +7.5dB peak signal? Does it run out of headroom and clip? Does it attenuate such that the peak remains at 0 thus reducing the level of the other channels accordingly? Does it apply some sort of compression to actively avoid clipping? I have no idea of the answers there.

 

In my case such downmixing happens in jriver (jrss) and I don't know how this is implemented there either.

[3ll3d00d]

In my case such downmixing happens in jriver (jrss) and I don't know how this is implemented there either.

I asked over on their forum and they confirmed it simply sums the two channels and then attenuates all channels to avoid clipping. This is probably one reason why I have found myself cranking the MV up on some films.

[SME]

I believe most AVRs and processors must go through some kind of Dolby certification in order to feature their technologies.  What I don't know is what requirements are imposed.  I'd take a guess that AVRs with anything Dolby will down-mix 7.1 to 5.1 according to Dolby's recommendations, but I don't know if that's certain.  Obviously, Jriver doesn't license any proprietary Dolby technology, so it would not be subject to such requirements.

 

I think you are right that every implementation is different with respect to levels and clipping behavior.  I know my Denon 3313CI has exactly the same limits on its mains output as on its sub outputs.  They all top out at around 4V RMS.  For full 7.1 sub WCS headroom at MV "0", you pretty much need the sub trim to be set to -12.  The WCS for mains is normally about 20 dB less than for the sub, but in the case of 7.1 to 5.1 mix-down using the Dolby coefficients, the WCS is closer to -12.5.  So on my Denon, I would need to set my surround trims no higher than about "0" to output WCS cleanly.

 

And then there's the wider question of whether WCS ever happens in a 7.1 mix.  I don't doubt that surrounds play at a lower level than the mains *most of the time*.  But is *most of the time* good enough?  Will we some day encounter a big 80 Hz hit that pegs all four 7.1 surround channels and bottoms our woofers as a consequence?

 

The info from JRiver is good to have.  Let me ask you, purely out of curiosity, would you prefer JRiver do it according to the Dolby recommendations?  In my custom processor (if I ever realize it), I can take 7.1 from the AVR and do whatever I want with it.  Should I mix it in a way that can create +7.5 dB peaks?  I mean this rather rhetorically, because it's not clear to me that there's any "right" way to do it, unless by some chance all the 7.1 mixes were getting monitored in 5.1 with Dolby's down-mixing in the chain.

[3ll3d00d]

The info from JRiver is good to have.  Let me ask you, purely out of curiosity, would you prefer JRiver do it according to the Dolby recommendations?  In my custom processor (if I ever realize it), I can take 7.1 from the AVR and do whatever I want with it.  Should I mix it in a way that can create +7.5 dB peaks?  I mean this rather rhetorically, because it's not clear to me that there's any "right" way to do it, unless by some chance all the 7.1 mixes were getting monitored in 5.1 with Dolby's down-mixing in the chain.

I think you hit the nail on the head in that there's no "right" way to do it as you don't know where the content came from to make those channels. 

 

mojave (aka desertdome) makes the argument for simply discarding the additional channels in http://yabb.jriver.com/interact/index.php?topic=96385.msg676397#msg676397 

 

I don't know whether the jriver way is better or worse than any other approach. Personally I don't think I like the sound of the dolby plus approach (of spreading it across the surrounds) so i'm inclined to go with either discard or sum as the valid options.

[SME]

mojave (aka desertdome) makes the argument for simply discarding the additional channels in http://yabb.jriver.com/interact/index.php?topic=96385.msg676397#msg676397

 

FWIW, I have listened to many 7.1 tracks on my current 5.1 system with my Denon AVR.  I haven't confirmed the signal levels that it outputs for signals sent to each 7.1 input channel, but I'm fairly certain it's going by Dolby's recommendations.  I can definitely investigate this further at some point in the future.  In any case, the phantom rear imaging works remarkably well on my system if sitting in the sweet spot.  I remember an especially cool effect in the movie "Malificent" (toward the end) in which a sequence of sounds were made to originate from a sequence of points along the semi-circle from one side across the rear to the other side, in a way that was completely consistent with what was depicted on-screen.  I'm fairly certain that effect would have been spoiled by either discarding the rears or mixing them like Jriver does.  Still, I can imagine hypothetical cases in which the JRiver approach might be preferred.