cool experiment

Hi all,

I just thought I'd post a neat litte idea that some of you may be interested in. I checked the forum to see if it has popped up before and couldn't find it.

Ever want to experiment with binaural filtering (like that used in wavearts panorama) but don't want to spend the cash? Well, back in the day the MIT media lab put out a detailed series of impulse responses using a dummy with microphones on its ears. You can download the impulse responses yourself and if you have some sort of convolution software (like Perfect Space) you can convolve them with your audio to pan things around in a circle or up and down (-40 degrees to +90 degrees). I took a crack at it this morning and it works well, except that the end result needs some serious eq to sound decent again. Also, you MUST use headphones to get the effect. It probably won't work on speakers. Here's what you do, in a nutshell:

Download the wave files from their site:

http://sound.media.mit.edu/KEMAR.html

I chose the windows zip file of all of the files. They're each only, like, 512 samples long so it was a quick download (cable modem).

Next, decide the elevation and azimuth (left to right, behind to in-front) - I suggest elevation zero (level with your ears) and 90 degrees (directly to the right of you - zero degrees would be right in front of you). Now, this is the only trick so make sure you remember it:

In order for this to work, you have to combine the two complimentary files L0e090a.wav and L0e270a.wav into a stereo interleaved file. I did this by importing them into separate tracks in SONAR (don't forget to pan them hard left and hard right, respectively) and exporting them as an interleaved steroe wave file. Once you're done, open the interleaved file as an impulse in Perfect Space and turn off the dry sound, you only want the wet sound. Make sure it's panned to the center, as is the instrumet track. The convolution itself will pan your instrument to the right for you. I suggest doing this first on a basic instrument track with no effects (I used a kick drum). You should hear the instrument coming from the right. Remember to use headphones.

A note about the file names:

Each directory name has the format ``elevEE'', where EE is the elevation angle. Within each directory each filename has the format ``XEEeAAAa.dat'' where X is either ``L'' or ``R'' for left and right ear response, respectively, EE is the elevation angle of the source in degrees, from
-40 to 90, and AAA is the azimuth of the source in degrees, from 0 to 355.

Also, keep in mind that you have to use complimentary files from the same ear mic to get this to work. You can't use 90 degrees from the left and 90 degrees from the right. You have to use 90 degrees from the left and 270 degrees from the left. Or, for X degrees, it's X degrees from the left and 360 - X degrees from the left. You can read more about this in the help doc in the section called "The Data".

http://sound.media.mit.edu/KEMAR/hrtfdoc.txt

Overall, the sound quality suffers from the convolution. It sounds thin and crummy, but some of that can be fixed with EQ. I think a better set of impulses taken by an audiophile would be much better. There may be others freely available on the web, or maybe someone with the equipment, talent and desire will post some impulses at some point. For what it's worth, the help doc explains the process of collecting the data. You could do it with two mics and your friend's head, instead of a dummy.

Anyway, that's about it. Hopefully someone will have fun with this little experiment. You could, of course, download the free demo of Panorama, but I thought this was more fun.

ORIGINAL: Paul Russell

alternatively, Spin Audio (www.spinaudio.com) offer a whole suite of binaural plugs which work quite well. But the binaural technique is subtle and works bets on headphones, much the like the Wavearts Version
Their M2 reverb ain't so bad either.

Thanks for posting that. I think the best part (for me) is it provides names of other research institutions that have created HRTFs, even some that clame to have a flatter frequency response. Hopefully I can dig some of those up, especially the ones from CIAIR, Nagoya University. As far as buying anyone's software is concerned, the engineer in me winces at the thought of paying for a program that essentially just implements the convolution operator when I already have one (Perfect Space). I just want some decent HRTF's.

I do admit, though, that panning would be a pain unless Perfect Space allows you to morph between impulses in real time. I haven't really read up on it yet. I'm considering just biting the bullet and learning how to code up VSTs. It shouldn't be that hard to crank out a little C++ app that convolves your input with an HRTF, lets you choose a speed for rotation in the XY and ZX and ZY planes, lets you pick some basic lfo shape for the panning, etc. It's just a matter of finding the time...

ORIGINAL: Paul Russell

It shouldn't be that hard to crank out a little C++ app that convolves your input with an HRTF, lets you choose a speed for rotation in the XY and ZX and ZY planes, lets you pick some basic lfo shape for the panning, etc. It's just a matter of finding the time...

the Spin Audio stuff does a lot more than that. It's pretty much the best solution for HRTF available as a VST. It depends how you value your time, but IMHO the Spin 3D suite is pretty good value for money in that respect.

Cool. Well, I definitely appreciate the link. What I'm really hankering for is to replicate a mac program I used back in my pre-pc days called Soundhack. It was free and you could use binaural filtering to pan things around. Really, all I care about is being able to statically pan things, but it can't be too hard to implement a simple lfo for it. Since I've already go the realtime convolver, though, I think it would be cool to squeeze a little extra functionality from it. I'll let y'all know if I find any HRTFs that yeild a better sounding output.