I think I have a slight. bias because in the early phases of his work, Dr. Chouieiri promoted it as if he had discovered binaural sound and crosstalk cancellation. There was a "gee whiz" quality to the promotions. That was clearly not the case. Since then, he and his aids have obviously gone further, and I have no problems with this. The concepts are valid, and with today's computing power, many things are possible - for a single listener willing to listen in a sweet spot or through headphones.
...
Not long after I joined Harman I negotiated a licensing arrangement to employ the Cooper-Bauck "transaural" crosstalk cancelling method into a product called VMAX - Virtual Multi-Axis..... Sadly it was not to be. Management of the period could not find a way to incorporate "software" as a product in our offerings - licensing being an obvious possibility. It drifted into nothingness.
....
The key difference between what I have described and what I think you are hearing through the BACCH system is that VMAX simulated loudspeakers at the correct stereo locations. Of course we listened to the "naked" crosstalk canceller, and what we heard parallels some of the descriptions I have read in this thread. When there are no phantom loudspeakers to provide directional anchors, almost anything is possible, and much of it is far from what the artists and recording engineers intended - the "soundstage" is very fluid, and envelopment can be profound. In this mode it is really a sound-effects generator and opinions of like or dislike will predictably vary. Please correct me if I am wrong.
Cheers, Floyd
I am glad to read the lively, well informed and insightful discussions on BACCH and crosstalk cancellation in this forum. Also, I would like to thank you, Dr. Toole, for sharing your memories of getting Harman products licensed to use the "Cooper-Bauck crosstalk cancelling method” back in the mid-nineties. Indeed crosstalk cancellation (XTC) in audio goes back to the seminal work of Cooper and Bauck in 1961, and binaural audio further back by decades, while the idea of crosstalk cancellation for visual 3D perception, which is a perfect analog to the audio XTC, goes back even further to the work of Wheatstone who invented the stereoscope in 1838.
It would be preposterous for anyone since Wheatstone (for photo/video) and Cooper and Bauck (for audio) to claim they invented crosstalk cancellation (or binaural audio). My research team and myself go to great pains describing this history in our XTC-related papers and presentations. For an excellent account of that history I recommend the first chapter of the recently published book “Immersive Sound: The Art and Science of Binaural and Multi-Channel Audio” from Focal Press (ISBN-13: 978-1138900004). That chapter was written by Prof. Braxton Boren of American University (who prior was a post-doctoral researcher at my lab for two years).
Even in interviews when I am not asked about the history of XTC I mention that XTC has existed for a long time and that our work focused on finding a solution of the central problem of the inherent coloration of XTC filers (even the permanent FAQ page on theoretica's website states "the technique of crosstalk cancellation (XTC) has been known for some time") . This tonal coloration problem is described in detail in Chapter 5 of the same book referenced above (which is an extended and updated version of the paper Mr. Borduin gave a link to) and which incidentally has a discussion and bibliography of the many papers on XTC since Cooper and Bauck (Note: I get no proceeds or benefits from the sales of this book). Basically, as I detailed in that chapter, XTC filters can be shown analytically (and through audition) to cause coloration to the sound (i.e. tonal distortion) that is problematic by Hi-Fi standards. BACCH filters, are a particular case of optimized XTC filters that have no tonal distortion. In practice BACCH filters are produced from a 2-point HRTF measurement done in-situ using the individual two-speakers-one-listener set-up and involve a patented method for producing a non-causal finite impulse XTC filter that has no tonal distortion.
I am sincerely sorry that some well-intended people, who reported on the internet about hearing BACCH, found our contribution too technical (or perhaps not sensational enough) to either comprehend or explain, and implied through their “gee whiz” reporting that we had invented XTC (or binaural audio). Unfortunately, not only this gives the wrong impression to some that we are making claims of inventing a technique whose existence precedes the work of my team by decades, but also denigrates the true value of our contributions and those of other workers in the field. Echoing your wise dictum “meanwhile there are still live concerts” I can add in the above context "meanwhile there are still journals” (and serious publications to which the serious inquirer can turn to find out the truth on both the history and the technical contributions.)
The combination of these tonal-distortion-free optimized XTC filters, advanced and robust head tracking algorithms that extend the sweet spot over the wide view range of a webcam or infrared sensor, and powerful 64-bit audio processing with modern multi-core CPUs (among many other advanced algorithms for high-precision impulse response (IR) deconvolution, IR alignment and interpolation and other methods driven by research in AR/VR) have allowed XTC-based 3D audio to enter the realm of not only AR/VR (where the requirements are exacting) but also of high-quality HI-FI. My research team and I would be delighted to host you for a visit to our lab in Princeton NJ for a set of demos that would give you an update on the state of the art of a field to whose commercialization you and your co-workers at Harman significantly contributed 23 years ago. (I should add that XTC, which is very well developed now, is no longer a focus of research at my research lab, where most of the work is on 3D soundfield navigation using arrays of higher order ambisonic (HOA) mics; HRTF synthesis from head scans; and acoustic isolation between adjacent sound fields.)
Please allow me respectfully to correct your guess ("simulated loudspeakers at the correct stereo locations") at what people are reporting (on this forum and elsewhere) about what they are hearing with BACCH. I believe your guess is perhaps a bit biased by your recollections of the circa 1995 Harman demos using the Cooper-Bauck XTC method to use "binaural cues simulating loudspeakers at specific locations” whose focus was on “deliver[ing] phantom 5 channel home theater from two loudspeakers”.
While we have developed a commercial 3D mixer, that uses individualized HRTFs for binaural virtualization (through both a pair of speakers or headphones) of sources (including virtual surround systems) in 3D space and navigating 3D sound fields, the use of BACCH filters that is of most interest to audiophiles is not the virtualization of speakers (creation of phantom speakers) but rather the removal of the limitations, due to crosstalk, on the levels of ILD and ITD cues that exist in virtually all stereo recordings and that can be transmitted to the ears of the listener by a pair of loudspeakers. Properly executed XTC allows the listener to receive these cues and better perceive the 3D location of the sources associated with them instead of the locations of the speakers. For instance the frequency-averaged ILD of each of the two loudspeakers in a regular stereo triangle does not exceed about 5 dB, therefore (speaking first of acoustic recordings) any source during recording that presents more than 5 dB of ILD to the stereo microphones (which would be the case of any source to the extreme right or left of the stage at angles much larger than 30 degrees) would be necessarily perceived by the listener during playback as locked into the speaker if XTC is not applied. A BACCH filter can provide an XTC level as high as 20 dB which would allow the ILD of sources, even those with extremely high ILD (e.g someone whispering in your right ear), to be reproduced correctly during playback.
Therefore what people are reporting about when they listen to an acoustic recording played through a BACCH filter has nothing to do with phantom speakers or location of the actual speakers, but everything to do with the more correct reproduction of the sources whose spatial cues were captured in the recording. Whether the cues are only ITD (as would be the case with a recording done with spaced omni mics), or ILD (as would be the case with ORTF mics) or ILD, ITD and spectral cues (as would be the case with a dummy head mic) all acoustic stereo recordings are based on one or a combination of these cues, and all of these cues would be severely limited by the crosstalk — the BACCH filter simply raises the ceiling on the delivery of these cues to the listener during playback, and the result is not a phantom speaker but rather a literally disappearing speaker. In fact, the locations of the speakers become, to a first order (i.e. neglecting the effects of reflections) completely immaterial to the 3D sound image. One of the demos that we often use to illustrate this is to create a BACCH filter for a certain 2-speaker-1-listener geometry, listen to the 3D image through that filter, then radically change the speakers geometry (even a very asymmetric one and with different distances to the listener) create the BACCH filter for that geometry and show that the resulting 3D image has not changed a bit.
Another very important felicitous result of raising the ceiling on the levels of ILD and ITD cues that can be delivered thanks to high-level-XTC, is the much more convincing reproduction of late reflections and reverb captured in the recordings. Such reflections (if captured in the stereo recording) can now be reproduced at the ears of the listener, thanks to the high levels of XTC, with much more correct levels of de-correlation and immensely enhance the sense of spatial realism and envelopment, compared to when they are perceived to come from the speakers.
The above arguments should not be controversial in the context of acoustic recordings but admittedly they do not perfectly apply to a recording mixed from artificially panned (often mono) stems (as would be the case of most recordings of pop music). There the image is artificial in the first place, and through a BACCH filter it would be artificially projected from a limited space between the two speakers to a 3D region of space the extent of which depends on the artificial ILD and ITD cues introduced by the panning. A purist might well object to this 3D projection as not one intended by the mixing engineer (who had no way of hearing it in 3D since the mixing was not done through XTC). It has been our experience that the vast majority of such purists are purist only theoretically. When in practice they are asked to compare the artificial 3D image through BACCH to the artificial image without it, they almost invariably prefer the former artificial image to the latter. For that you can take the words of the many listeners who have reported their impressions after days of listening, or even better come visit and listen.
Apologies for the lengthy reply and, again, thank you for adding your thoughts to this discussion.
Edgar Choueiri
