AleXH

Hi!
Everybody know how formula downmix 5.1->2.0? For downmix to 2.0 used amplitude 5.1 channel only or their phases too? IMHO for simple case (user put in center square) formula must be fR=fR+0.5center+0.5lfe-sL, fL=fL+0.5center+0.5lfe-sR. I'm right?

Imho now (3550) surrounder+ downmix not enough lfe in result 2.0 signal.

Max M.

you'll find the info here.

>Imho now (3550) surrounder+ downmix not enough lfe in result 2.0 signal.

Do you mean this has been changed with 3550? - if it has then it is some kind of bug. The LFE mixing coefficient for downmixing always was / should be 0.5.
(I don't have 3550 installed - so i can't check it out)

__________________

Russ

It is still 0.5.

The formula (with default settings of VCenterA=1, VRearA=1, VRearW=0.5) is something like:
fL = sL - (sR * 0.5) + (C * 0.5) + fL - (LFE * 0.5)
fR = sR - (sL * 0.5) + (C * 0.5) + fR - (LFE * 0.5)

Max M.

Thanks Russ.
(i just noticed the LFE is inverted for some reason - which is confusing (i can't recall any special reason for this - it's probably just an overlook) - but considering the definition of LFE signals - this should not be really an issue)).
So i guess AlexXH will provide more details on "now" and "how".

__________________

AleXH

I'm sorry for my very bad english, I'm speak russian.
2Max M.

IMHO This right, Russ decrease probability overload register. Phase for LFE doesn`t really matter, so as length wave more larger than length between inside ears.

2Russ

fL = sL - (sR * 0.5) + (C * 0.5) + fL - (LFE * 0.5)
fR = sR - (sL * 0.5) + (C * 0.5) + fR - (LFE * 0.5)

But why???
IMHO:

Take phases signal in senter square:

fL______CS______fR


________U________


sL______________sR

CS - center & subwoofer, U - user

sL rotated to fL at PI/2, to fR at PI;
sR rotated to fR at PI/2, to fL at PI -> have:

fL=fL+cos(PI/2)*sL+cos(PI)*sR=fL+0*sL+(-1)*sR=fL-sR
fR=fR+cos(PI/2)*sR+cos(PI)*sL=fR+0*sR+(-1)*sL=fR-sL

About mixing coefficient 0.5 for center and subwoofer:

when source signal moved from position center to fL or fR then distance extended at sqrt(2)=1.4142, sound volume inverse ratio to distance and dimish to 0.7071. Consequently coefficient must be 0.71:

As a result volume for user position=coeff*(volume inverse distance to fL) + coeff*(volume inverse distance to fR)=0.71*0.71+0.71*0.71=1 (Original volume)

, and formula for downmixing 5.1->2.0 for user in center square must be:

fR=fR+0.71*center+0.71*lfe-sL
fL=fL+0.71*center+0.71*lfe-sR

Max M.

well, first of all - the coefficients of center and rear channels mixing are fully adjustable (see the formulae in the mentioned thread).
secondly (though it's quite irrelevant because of above) the coefficients are more a question of a personal taste raither than of common sense math - so the "must be" is not something really applicable here, well here're few points:
* actually, the prevailing 5.1 speaker placement scheme is not the square.
* there's simply no (and can be no) "perfect" setup (even one tuned to personal taste is still highly depended on a source material - it may sound nice for one source and artificially awful for another)
* (this context only, i.e. "the coefficients") whatever bad subjective listening tests outweght whatever good common sense args.
* this thread is quite illustrative for the total subjectivity of the subject: just wonder - you noticed the LFE "issue" (while it has the "right" coefficient) and did not hear any "center" bug (while it mixes in "wrong" way) until you actually saw the numbers

Regardless of above, you know you can always come up with your own downmixing setup and advertise it as the "must be"
(or even scratch your own plugin - for example you may start with editing of cnv51to20 (it is the most basic downmixing "da only" plugin (i.e. you don't need any C++ stuff to modify it) that performs exactly the same thing as the surrounder in the discussed mode))

__________________

Max M.

The angle in geometric space is not really equal to signal phase for mixing - i.e. simple inverting of a channel won't place it behind you.

__________________

Russ

Hehe, not me, I just posted the info...