On 03/14/2013 12:37 PM, Paul Davis wrote:


On Thu, Mar 14, 2013 at 6:42 AM, Jörn Nettingsmeier
<[email protected] <mailto:[email protected]>> wrote:

    On 03/12/2013 08:08 PM, Tim E. Real wrote:

        But having said that, yes I'm wondering about a true 'stereo
        pan' feature.
        How would such a feature work?


    there is no one true stereo pan.

    a pan law for intensity stereo (i.e. a panned image or an XY
    coincident microphone pair) would increase one channel and decrease
    another such that the total energy remains constant. a cosine/sine
    law is usually used, because

    cos^2 + sin^2 = 1

    ardour3 attempts to do this, by allowing you to reduce the width (by
    introducing crosstalk), and then letting you move the compressed
    image left or right. sort of works, but only for pan-potted stuff.

    a pan law for run-time stereo (i.e. spaced omnis) would have to use
    delays, leaving the original level intact.

    the ardour3 panner gets this type of signal horribly wrong, because
    you _never_ want to introduce crosstalk in spaced omnis - instant
    comb-filtering hell.

    for stereo techniques that incorporate both run-time and intensity,
    such as ORTF, NOS, EBS, you-name-it, you need different amounts of
    gain change _and_ delay.

    that's why nobody wants to use a ready-made stereo balance control -
    it is almost guaranteed to do the wrong thing for the source
    material at hand.


git add libs/panners/spaced_omni_panner
git commit
git push


point taken :)

the problem is, you usually have a mixture of the above. hence, no way to get the stereo panner right. unless the user knows exactly what s/he is doing, and then s/he doesn't really need a stereo panner :-D

btw, sorry for the pot shot at ardour specifically - in fact, most if not all DAWs get it wrong. iirc, some DAWs with a focus on classical (sequoia, pyramix) have some code to mogrify complex stereo, but i haven't used it.

my approach to complex stereo re-panning is: split the signal into two mono busses, add a delay, add eqs if you have to, move faders. and while we're at it, the same approach works for processing in the MS domain. it's painful enough to only use it when you really need it, and hard enough to discourage casual users :)

for those who read german, here is a discussion we had about this in the VDT forum: http://www.tonmeister.de/forum/viewtopic.php?f=6&t=238&p=817


best,


jörn



--
Jörn Nettingsmeier
Lortzingstr. 11, 45128 Essen, Tel. +49 177 7937487

Meister für Veranstaltungstechnik (Bühne/Studio)
Tonmeister VDT

http://stackingdwarves.net

_______________________________________________
Linux-audio-dev mailing list
[email protected]
http://lists.linuxaudio.org/listinfo/linux-audio-dev

Reply via email to