Head-related Tranfer Function

[Wombat]

Paper: http://mafija.fmf.uni-lj.si/seminar...d-Related__Transfer_Function_Tilen_Potisk.pdf

 

[Cosmik]

He analyses loudspeaker crosstalk cancellation, but...

If the listener were to move his head slightly, this would change the elements of the matrix H. This would result in a loss in crosstalk cancellation control at certain frequencies (which form a harmonic series). When a crosstalk filter boosts some frequencies, we say that they bring spectral coloration i.e. affect the sound colour of the recording.

 

[Home Audio Fidelity]

Hi all,

For the ones interested, I have created a video with audio examples to better explain the concept of acoustic crosstalk in stereo reproduction:

With improved algorithms is it possible to drastically reduce the sensitivity to head position and the impact of tonal coloration. BACCH algorithm is one alternative (patented) and I came up with a different solution that is less sensitive to head position (to avoid head tracking for standard listening situation)
The audio examples are generated using using a simple binaural room simulation (image-source model) for the raw versions and by adding the crosstalk reduction filters for the processed versions.
I will be happy to provide support for the ones willing to test my algorithms.

Best,
Thierry

 

[Wombat]

Hi all,

For the ones interested, I have created a video with audio examples to better explain the concept of acoustic crosstalk in stereo reproduction:

With improved algorithms is it possible to drastically reduce the sensitivity to head position and the impact of tonal coloration. BACCH algorithm is one alternative (patented) and I came up with a different solution that is less sensitive to head position (to avoid head tracking for standard listening situation)
The audio examples are generated using using a simple binaural room simulation (image-source model) for the raw versions and by adding the crosstalk reduction filters for the processed versions.
I will be happy to provide support for the ones willing to test my algorithms.

Best,
Thierry

Purely binaural.

No bone conduction? No leakage? No neural cross-processing? Just asking.

 

[restorer-john]

No bone conduction? No leakage? No neural cross-processing? Just asking.

What's the canal separation? oops, channel.

 

[Wombat]

And ear-lobing.

 

[davidc]

This is exactly what Bob Carver did with his Sonic Holagraphy back in the 1980's:

https://patents.google.com/patent/US4218585

The effect was pretty wild, but was very sensitive to any head movements. He has said that he incorporated this concept into his current speaker design.

Polk Audio did the same thing but with an extra set of speakers in each cabinet with their SDA series also back in the 1980's. I was surprised to see that they still have a line of speakers with the technology.

https://www.polkaudio.com/discover/sda-technology

 

[phoenixdogfan]

This is exactly what Bob Carver did with his Sonic Holagraphy back in the 1980's:

https://patents.google.com/patent/US4218585

The effect was pretty wild, but was very sensitive to any head movements. He has said that he incorporated this concept into his current speaker design.

Polk Audio did the same thing but with an extra set of speakers in each cabinet with their SDA series also back in the 1980's. I was surprised to see that they still have a line of speakers with the technology.

https://www.polkaudio.com/discover/sda-technology

On the one hand we have purveyors of loudspeaker DSP like Bacch who purport to remove crosstalk from the soundfield for loudspeaker listening claiming it makes for a better listening experience, and on the other hand we have purveyors like Smyth who purport to introduce cross talk into the headphone listening soundfield claiming that makes for a better listening experience.

Obviously both positions can't be correct. And if the answer is some crosstalk is gooood but too much is baaad, then what support do any of these purveyors have to prove they are providing the Goldilocks amount? How would they measure that? How would a speaker so treated look if it ran through Amir's battery of tests, particularly those on the Klippel which measure directivity and in room response?

Moreover, since Bacch and Home Audio Fidelity have products which claim to do similar if not identical things (albeit at vastly differing price points) how do they compare to each other in accomplishing their performance objectives? And, please let no one make the claim that one product is necessarily superior to the other because of their respective prices. After all, how many of us would prefer a Wilson Alexandria to a Salon 2 in spite of the vast gulf in their retail prices? As we have come to find out, prices have precious little if anything to do with how performant a product is in audio.

 

[onion]

On the one hand we have purveyors of loudspeaker DSP like Bacch who purport to remove crosstalk from the soundfield for loudspeaker listening claiming it makes for a better listening experience, and on the other hand we have purveyors like Smyth who purport to introduce cross talk into the headphone listening soundfield claiming that makes for a better listening experience.

BACCH has a headphone filter (with head tracking) that emulates its filter for loudspeaker XTC (that creates a 3-d sound field external to the head). I suspect that is v similar to the Smyth Realiser.

Obviously both positions can't be correct. And if the answer is some crosstalk is gooood but too much is baaad, then what support do any of these purveyors have to prove they are providing the Goldilocks amount?

The limiting factor IMO is the set of decisions made during the recording production process. When I listen to a Louis Armstrong/ Ella Fitzgerald redbook album that was recorded presumably with a stereo mic that captured their positions relative to each other, the effect is spectacular. It's as close to a live performance as I've ever heard. When I listen to a friend's DnB album, sometimes, the 3d audio effects sound incredible and sometimes they sound weird. In all cases, they sound as 'artificial' as the genre itself (compared to acoustic/ live/ vocal performance)

I think the general idea behind XTC (replicating the ILD and ITD of live point-source emitters in a 3d sound-field at the ears) is a sound idea grounded in physics. It's the recording and production process associated with the music we consume that may make things messy or less effective. Having said that, over 90% of music I listen to sounds better with BACCH with a huge benefit being that it does not induce listening fatigue.

How would a speaker so treated look if it ran through Amir's battery of tests, particularly those on the Klippel which measure directivity and in room response?

Dunno. With BACCH, the impulse response, frequency response and magnitude of XTC at the ears is automatically measured with each calibration.