13/06/11 02:55, « Fons Adriaensen » <[email protected]> :
> > There is no reason why an XTC system should remove center > bass signals, and as far as I know none of them do this. > > I suspect you are confusing 'out-of-phase' and 'difference'. [[Profanity warning: the following contains crude and probably inexact mathematics.]] In fact, I was still referring to the BACCH paper: the "out-of-phase" response seems to be defined as the magnitude of the difference between what is (commonly) called H1(w) and H2(w), h1 being the diagonal term for the symetrical XTC matrix and h2 the contradiagonal term. And the "in-phase" response is simply magnitude( H1(w)+H2(w) ). Still in the same paper, you see that the overall envelope spectrum is defined as the maximum between "in-phase" response and "out-of-phase" response. The trick is to minimize that envelope spectrum excursion so that it stays below a certain value, hence achieving the low XTC coloration Marc could hear in his A-B-C listening test. Anyway, I must admit I had no opportunity to hear that particular filter, but still: the "in-phase" response, in the low frequencies, is way below other levels (see figure 2 for the stereo-dipole cas: -10dB for unregularized filter). The well knows inversion techniques (Farina, etc) state that: H1 = Hi * (Hi^2 - Hc^2)^-1 H2 = - Hc * (Hi^2 - Hc^2)^-1 where Hi is the ipsilateral response (eg. left speaker -> left ear) and Hc is the contralateral response (eg. left speaker -> right ear) > > For such a signal, the difference L-R is very small. > Reproducing it correctly using normally spaced stereo > speakers (which requires amplitude differences instead, > this recreates the original signals at the ears) means > you need high gain for this L-R signal. To reproduce > it (and create an off-center bass) using closely spaced > speakers and crosstalk cancellation requires even more. > But none of this means that a mono (center) bass would > be removed. If we have a mono bass sound panned to the center, we'll eventually have Hi ~= Hc so the inversion leads to infinite values for H1 and H2 in the low frequencies, unless one uses regularization. This is my local (personal!) explanation for the XTC bass problem, but I may be wrong. But back to that Massive Attack bass line: if we use something like E=1 (full regularization, which means no XTC) in the low frequencies, when you pan a mono bass sound to the center, both loudspeakers will emit the same signal = (h1 + h2) (*) bass = (hi - hc) (*) bass = ..... = 0 (*) bass. Hence: no more bass. Yet most XTC filters often sound more bassy than they should. So, I'm wrong, but I don't know why (yet). Thanks, Best, FM _______________________________________________ Sursound mailing list [email protected] https://mail.music.vt.edu/mailman/listinfo/sursound
