0
$\begingroup$

First post, and this isn't my field of study, but I'm giving this a go anyway:

My latest endeavor is to use a bit of DSP magic to match up various pairs of headphones, at least in terms of their impulse response measured at the entrance to the ear canal. So far I've taken 64 measurements on 3 sets of headphones, in this way:

1) Put on binaural mics

2) Put on headphones

3) Take IR measurements

These are in a nested loop with 4 iterations each, so I end up with 64 measurements. The IRs I get using a program called Aliki, free software in Linux that uses sine sweeps.

My goal is to generate two filters for each headphone pairing (1<->2, 1<->3, 2<->3). Currently I'm eyeballing a "best" IR for each headphone using spectrograms, matching up the position of peaks, padding, then doing FFT division to get a transfer function (which I have read can be a really bad way of doing things). Results are actually not bad sounding, but I wonder if I can do better with a bit more legwork on the data.

So my question is: What's the best way to combine these IRs to get (possibly) better filters?

Thank you.

$\endgroup$
2
  • $\begingroup$ Not really related to what you're asking, but since I've been looking for a pair of binaural mics: What model/brand are you using and are you happy with them? $\endgroup$
    – Jazzmaniac
    Aug 3, 2015 at 8:19
  • $\begingroup$ The "master series" binaural mic from Sound Professionals. So far no issues, though I've only used them via phantom power from my E-MU 0404. Haven't tried them on-the-road yet. $\endgroup$
    – RRod
    Aug 3, 2015 at 12:47

1 Answer 1

1
$\begingroup$

Why not to simply match the frequency responses and sensitivities? That's what labs are actually doing.

Since you have the IR's for each device, simply calculate the Frequency response of each repetition, average and compare between devices. In Python that is very simple task.

$\endgroup$
1
  • $\begingroup$ Perhaps I am just overthinking. I guess my wonder is how much phase measurements matter at all to this sort of thing. $\endgroup$
    – RRod
    Aug 3, 2015 at 12:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.