I am delighted to see a lively discussion of crosstalk cancellation
(XTC). My name is Edgar Choueiri and I am an academic researcher
working on XTC for the past few years.
I would like to add a few comments that may be relevant to the discussion.
XTC has come a long way since the pioneering work of Atal, Schroeder
and Damask and Bob Craver's ingenious C-9 Sonic Hologram. The list
of selected references given by Mr. Politis in a previous post gives
a good sampling of some of the important work in that field. A
number of other relevant publications are listed in the reference
list of a paper I wrote on the subject, which can be downloaded from:
http://www.princeton.edu/3D3A/Publications/BACCHPaperV4d.pdf
The Introduction section of that paper gives an overview (with
references) of the research efforts to solve some of the outstanding
problems in XTC over the past decade or so. One of the biggest
problems, and in my opinion the one that has kept XTC from being
widely accepted, is the severe audible spectral coloration (and
accompanying loss of dynamic range, typically in excess of 30 dB)
that results from applying XTC filters.
This problem was solved elegantly by the folks at the University of
Southampton's ISVR (Takeuchi and Nelson) who developed a method,
called "Optimal Source Distribution (OSD)", which requires many
(ideally an infinite number of) loudspeakers to azimuthally span the
front of the listener. A commercial implementation using a total of 6
loudspeakers was commercialized by Marantz (e.g. Marantz ES7001) with
the main function (alas, but understandably) being 5.1 surround. The
system can yield unprecedented XTC levels with virtually no
coloration (I have no relation to either ISVR or Marantz).
My laboratory has concentrated on achieving (and hopefully exceeding)
the same result but with only two loudspeakers. Recently, average XTC
levels exceeding 20 dB with no spectral coloration (i.e. the XTC
filter having ruler-flat frequency response) have been achieved in a
typical listening room with no sound treatment.
The basic approach is detailed in the above linked paper (in the
simplified context of two ideal point sources and no HRTF). In
practice the method is applied to design optimized HRTF-based XTC
filters (called BACCH filters) from actual impulse measurements
recorded with a dummy head or, for even more accuracy, with
microphones in the listener's ear canals.
Much on the method and recent results is documented at the lab's website:
http://www.princeton.edu/3D3A/
A simple but striking demo of the present capabilities of
state-of-the-art XTC technology can be heard by playing the following
XTC-filtered tracks through a Jambox (a compact bluetooth-connected
speakers unit):
http://www.princeton.edu/3D3A/Jambox+BACCHDemos.html
Workers/researchers in the field of 3D audio who happen to be (or
visiting) the Central NJ/ NY metropolitan are are welcome for more
extensive demos at realistic SPLs.
Regards,
Edgar Choueiri
p.s. Regarding the original question in this thread by Mr. Junfeng, I
should point out that BACCH filters (like other XTC IR filters) are
applied to the audio through VST or AU plugins that can do 2x2 (also
called "True Stereo") IR convolutions. Typically these IR convolution
plugins were developed for IR-based reverb. The plugins that I know
can do the required 2x2 convolution are: Angelo Farina's X-Volver
(free), Waves IR and SIR2.
Finally, I should point out that the latest generation of optimized
XTC filters, due their optimal nature (they are designed by
minimizing a cost function for a given speaker configuration) work
equally well for any speaker span, including the stereo dipole
configuration and the standard stereo triangle (+/- 30 deg span)
configuration.
--
_______________________________________________
Sursound mailing list
Sursound@music.vt.edu
https://mail.music.vt.edu/mailman/listinfo/sursound
