Yeah, probably. But:
1. I didn't care very much.
2. Math is hard.
3. Those channels aren't really used for very much in the vast majority of content.
4. I'll never watch something "important" with those speakers enabled.
So, if you can figure out what I should be using, to prevent clipping, but handle the proper "JRSS" to those two rear channels, I'd love to hear about it. Otherwise, this works well enough for me.
Two perfectly correlated signals of equal volume sum +6dB, two perfectly uncorrelated signals of equal volume sum +3dB. Real-world signals are rarely perfectly one way or the other, but tend towards lack of correlation. As Matt pointed out in the past, any real-world mixing will be a compromise between volume and clipping (i.e. you can downmix in a way that guarantees no clipping, but your volume will be very low). See, e.g. http://yabb.jriver.com/interact/index.php?topic=79282.msg538644#msg538644.
My recollection is that JRSS assumes channels are incoherent (which is mostly true, there is some overlap), which risks clipping in some "perfect storm" scenarios. In the case of moving the center to the L and R, it's a better idea to assume 6dB of gain from that duplication as the signal is (necessarily) perfectly correlated. I'm not sure that the whole JRSS matrix is public info, but we know that the mono downmix is -1.76 dB for LFE and -11.76 dB for other channels. But we can reconstruct a usable matrix using the math of audio summing .
For your current setup (if you really don't care about the surround or LFE info on those speakers), you should probably adjust the L and R info down by 1dB during the copy to RL and RR, and then add the center at -7dB. I think that will be more energy neutral; with your current settings, the center is -3dB from your L and R, even after the 6dB of gain from duplication. Those settings will be relatively safe, but won't guarantee no clipping. If you want to guarantee no clipping, use -3.6 dB on the L and R when copying and -9.6dB on the center.
If you want to add in the surrounds, there's a few different theories about how to do that (do you add both channels at different volumes/polarities to both mains, or just the left to the left and right to the right?). I don't know how JRSS works it out, but you won't go far wrong by adding the SL to the RL and the SR to the RR. In that scenario, you'd want to copy the L and R data to the RL and RR at -3.6dB, and add the center to both at -9.6dB, and the SL and SR at -3.6dB as well, etc. If you want to guarantee no clipping, you'll need -8dB on L,R, SL, and SR and -14dB on the center.
If you want to integrate the LFE, you need 4dB of separation between it and the other channels (when it sums +6dB, that will make up the 10dB differential between LFE and other channels), so then you're talking about (theoretically) adding the L,R,SL, and SR channels at -6.8dB, adding the LFE at -2.8dB and the center at -12.8dB. Based on the mono downmix target above, I think that's probably similar to the amount of headroom that JRSS creates so should be as safe as JRSS in regard to clipping. If you want to guarantee no clipping, you'll need -8.2 on the LFE, -12.2dB on L,R, SL, and SR and -18.8dB on the center [edited for math error]
If anyone has more direct knowledge of JRSS and is willing to chime in, I'd be interested in more info about what JRSS actually does. this is just what the math suggests.
Check out these calculators if you want to investigate/replicate on your own:
http://www.sengpielaudio.com/calculator-spl.htm http://www.sengpielaudio.com/calculator-coherentsources.htmhttp://yabb.jriver.com/interact/index.php?topic=87945.msg606061#msg606061
EDIT: Whoops ninja'd, I knew someone had posted something recently. Looks like that downmix weights the surrounds differently than the other channels, and comes to some different conclusions. 6233638, I'd be interested to hear any thoughts you have about why we came to such different conclusions (or where I went wrong
)
