On Tue, Mar 18, 2014 at 03:32:55PM -0300, Iain Mott wrote: > Thanks a lot Fons. When I pan pink noise with at W at -20dBFS RMS, the > individual X and Y channels peak at about 3dB higher. Is that why you > said to meter at 86dB and not 83?
No. In a stereo system, with the levels as 0dB on the K-20 meter, each speaker produces 83 dB SPL. Assuming the signals are mostly decorrelated, the total level will be 86 dB. So the 'reference SPL' is 86 dB. > I wrote to a forum on Bob Katz's site and he responded saying no, all > should be set to 83dB SPL..."Don't worry about how it adds up. This > increases the headroom for the mix engineer, who doesn't have to push > each channel as hot to get the same loudness.... and therefore the > enjoyment of multiple channels. It also partly explains why stereo is so > crippled a medium." That's less than helpfull. As long as most of the signal is coming from a front pair of speakers and the others won't add much to the total level, Katz is right. But if you mix real surround with the sound coming from all directions instead of 'front with surround ambience', this doesn't work anymore. Using his advice, if you'd have 100 speakers, to get the standard level you'd have to mix to -20 dB (relative to 0 dB on the K-20 meter, hence -40 dB RMS absolute level). This is just not realistic. > I guess also with cinematic sound - it's more rare to have all channels > on full all at once (or maybe that's the films i watch!). > > For my current purposes, I'd like to reproduce as best as possible, > ambiental b-format recordings over an array of speakers - and preferably > try to match SPL measurements taken at each recording location. Do you > think the formula above would be correct to match levels in this way? > ie. if I make a recording at a site where the SPL is 70dB, during > playback I meter this material (the W channel) at -13dB RMS on a k-20 > meter, and in the case of a 14 channel system, calibrate each speaker > channel at 71.5dB SPL (x = 83 - 10log14). Your only chance to get this right is to calibrate *via the decoder*. If you follow the procedure I explained, then 0 dB on the K-20 meter for W will corresponds to 86 dB SPL, no matter how the sound is distributed over the speakers. That's assuming you don't pan two or more strongly correlated signals to different directions (if you do that the result is no longer really Ambisonic). Let's define the 'gain' of the decoder as follows: * Send a single panned signal at reference level in W to the decoder. * Measure the sum (of levels for LF, of powers for HF) of all outputs of the decoder. * Decoder gain = this sum divided by reference level. Now assume that the decoder gain measured that way gain is adjusted to 0 dB. Then the gain from the W input to any output will be well below 0 dB, by a factor that depends on the number of speakers, order, and frequency range (for a dual band decoder). In other words, the -10 log (N) you refer to above *is included by the decoder*, but more accurately, taking into account all factors. Ciao, -- FA A world of exhaustive, reliable metadata would be an utopia. It's also a pipe-dream, founded on self-delusion, nerd hubris and hysterically inflated market opportunities. (Cory Doctorow) _______________________________________________ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound