On Tue, Feb 23, 2010 at 04:37:53PM +0300, alex stone wrote: Jconvolver can do this, there's a config file in the distribution. This is not the recommended way to use such a reverb. The outputs from jconvolver should just go into the AMB mix bus. If you use Ardour you can connect them directly to the master bus, or you could use a 'bus' strip to have a gain control at that point. The input of jconvolver should normally come from post-fader aux sends on the same strips as where you insert the AMB panners - the panner and the aux send get the same post-fader signal. This way you can set the level of reverb for each panned source separately. The relative level of the reverb is also preserved when you use either the channel or master faders, as the reverb sits in between those, as a side chain of the master bus. In fact you can do this becaus a convolution reverb is a linear process: reverb (a) + reverb (b) = reverb (a + b). To be used like this the reverb must *not* include the direct sound. Ideally the input to the panner should also have a variable delay - to put a source at a larger distance you would then increase the send to the reverb and increase the delay in the direct sound via the panner. But this is a refinement that is difficult to arrange in Ardour, it requires using two strips for each panned source. If the reverb has many inputs corresponding to different source positions you would have as many aux busses, plus one extra for the 'tail' of the reverb which is usually a separate input for jconvolver. Ciao, -- FA O tu, che porte, correndo si ? E guerra e morte !

hi alex, fons! On 02/23/2010 02:37 PM, alex stone wrote: as fons said, jconv does a nice job of it, and i'm using it exclusively now. i did my first experiments with farina's method, described here: http://pcfarina.eng.unipr.it/Aurora/B-Format_to_UHJ.htm it's essentially the same, only that farina provides one IR per input/output pair, where fons does the same more elegantly with a single hilbert transform (they do the 90° phase shift indicated by the "j" coefficient as you find them on farina's page), and a couple of delays. the huge delay is a problem for me, because i like to A/B between UHJ and first or second order ambi - but it's easily fixed by inserting a ladspa delay plugin in the other channels. the "simple delay with cubic interpolation" works fine for me (part of the swh set iirc). the nice thing about ardour is that a mono plugin will automagically be replicated n-times in an n-channel bus. then my explanation probably sucks :( to clarify: tetraproc, a stand-alone jack application, is first and foremost an A->B converter, i.e. only useful if you have to deal with the output of a tetrahedral microphone directly. in addition, it has a virtual microphone processor that can derive any coincident stereo pair from a b-format. if you want to do just the latter, and you already have a b-format, it's easier to use the vmic plugin which is part of fons' AMB plugins - it does the same job, but is simpler to use (no jack patching). if you have a b-format and want to produce stereo, these are the two options you have: 1. use UHJ to fold it down automatically (including extra surround information that allows the consumer to blow it up to horizontal surround again, with a few compromises). but since uhj basically moves the rear sound stage to the front, it can sound slightly over-ambient (not as extreme as blumlein stereo, though). 2. use a virtual stereo mike to pick out a stereo perspective. it allows for more control, is a lot easier to produce, and it's probably the safer option if you want your mix to be reliable on consumer gear... i'm experimenting with uhj, but i've found that the sound stage can vary wildly between different playback systems. it's superior to my ears when all goes well, but difficult to control. and "controlling" here means to tweak your surround mix and anticipate what it will do to the UHJ. weird way of working. if on the other hand you never listen in surround and only mix for UHJ, the b-format will likely not make much sense when played back ambisonically. :) same here. i may have two years of a head-start, but fons has endured extremely dumb questions on the way... so i'm happy to be of assistance to others now! there is no ladspa UHJ encoder yet. i would definitely go that way, to be able to easily provide a b-format output. since you are dealing with artificial sources anyways, why stick to first order? do your panning in higher order instead. the use of resources is minimal (although it will create an insane amount of jack ports, so use -P2048). it won't make any difference to your UHJ, since it only ever sees WXY, but it would enable you to produce really crisp n.1 or HOA decodes without any extra work. just one, totally unrelated, but smart(-ass) nonetheless: could you please edit your replies to quote only the part you are directly referring to? i'm usually very curious about things you have to say on the list, but my mouse wheel finger is getting RSI... best, jörn