Issue creating cross-talk cancellation filter

Hi,

Firstly, thanks to Angelo Farina for all the work and dedication on these plugins and in this research area!

I am experimenting with creating a cross-talk cancellation filter, but the waveform isn't cancelling out when I run the crosstalk IR against the original HRTF (and I don't hear the widening I would expect)

Step 1. I created a 2x2 binaural IR/HRTF

I tried two ways of creating this:

A. myself, with binaural microphones in my ears, via a sine sweep
The resulting 2x2 WAV is at: e1.pcloud.link/publink/show?code=XZkpDbZH1Fxy8cpTVF5DVlljj5mkphmtg07



Note that you can see the ITD between the left and right channels.

B. I also tried with the impulses from a TV Studio BRIR which I downloaded from github.com/ShanonPearce/ASH-Listening-Set/tree/main/BRIRs/R19
Obviously this won't work aurally, but is useful as a test. The 2x2 formatted WAV is at: e1.pcloud.link/publink/show?code=XZ0pDbZWm6ojjxkmwVX9lD00LiHtBm6InS7



Likewise, the ITD is visible.

Step 2. Run Invert Kirkeby

Here are the results on the TV Studio IR (I also ran on my home-recorded IR, the results on both look almost the same):





Inverted IR WAV: e1.pcloud.link/publink/show?code=XZCLDbZ4prPfAYCBApAnyU8Ytmz77W2g7O7

(this also seems to be the same whether 'Cross-talk cancel only' is turned on or off?)

Step 3. Convolve inverted filter against original HRTF



This is the output:


As you can see, it hasn't cancelled out as compared to the instructions at pcfarina.eng.unipr.it/Aurora_XP/Kirkeby's%20Inverse%20Filter%20v4_0.htm

Edited by mark9white - 2/13/2023, 02:52 PM

The error is selecting "cross-talk cancel only". This removes the calculation of inverse filters, performing just the cross-talk matrixing.
This option cannot work with real IRs measured, it can only work for Dirac-like pulses (synthetic IRs created with simulation software, or manually).
Just deselect thst option, and a proper set of cross-talk cancelling filters is generated.

Thank you!