Hi Richard. Good to hear from you. (More below.)

On Feb 28, 2011, at 3:39 PM, Richard Lee wrote:

> I had a page on ambisonia.com with a lot of what has been discussed and was 
> trying to find it before Jerry, who provided much of the info, noticed our 
> discourse.
> 
> Unfortunately ambisonia.com seems to be unobtainium.  I know this has been 
> asked before but is ambisonia.com archived anywhere and will it be available 
> again?
> 
> I'd like to thank Jerry for posting even more details of the history of this 
> fascinating subject.  Transaural (ie Jerry) are definitely the originators of 
> the 2 most important concepts; 
> 
> speakers close together 
> and 
> having the bass units further apart.
> 
> These turn xtalk cancellation from a fancy technical exercise to something 
> which promises better performance.
> 
> Robert Greene was the first to suggest a physical barrier but I don't think 
> having your nose stuck to a vertical mattress is as elegant as the Greene/Lee 
> neckbrace.
> 
> One of the points I made on my webpage is that x'talk cancelled "binaural" 
> over speakers is a SURROUND technology needing only 2 'speakers' but the 
> encoding is complex.  Answers to the encoding question are invited.
> 
> I'll second his reminder of Dr. Duanne Cooper's role in the birth of 
> Ambisonics.  His name appears on the patents and Michael always spoke highly 
> of him and his contribution.
> 
> Jerry, Eric Benjamin mentioned that he'd found a commonly used set of HRTFs 
> (KEMAR I think) flawed.  I think this is why he prefers spherical head 
> models.  Have you a recommendation for a publicly available HRTF set?

Unfortunately I can't make a recommendation from the various publicly available 
HRTF sets because I'm not familiar with them. We were lucky enough to have 
access to a rather sparse but sufficient set of HRTFs from a private source 
that I trust. I mentioned in the original post that there were a few obvious 
glitches that could be sort of "puttied over" but otherwise I guess the proof 
of their suitability was really in the results.

I compared the results using our HRTF set to those of a sphere model. (FWIW, 
Fortran code for the sphere model was published in an AES conference paper by 
Cooper and I in around 1979 or 1980 [thank you, Lord Raleigh, for doing the 
hard work]. There is a sign error in it, however, and the stopping criterion 
for the series summation is suspect. Later, Dick Duda approached me at a 
conference and I subsequently gave him corrected Pascal code which he then used 
in his own research. Also, he and Bill Mertens published a very nice piece on a 
spheriodal head model, probably in JASA.) My recollection is that the sphere 
model matches the real-model heads pretty well up to about 2 KHz for a 
crosstalk canceller, and up to about twice that for a speaker spreader. I think 
some of the differences between the two models tend to cancel out in a speaker 
spreader due to similar features appearing in the numerator and denominator 
parts of the defining equations, especially if the angle differenc
 e between the real speakers and the virtual speakers isn't too extreme.

I used to know Steve Orfield in Minneapolis who used to (maybe still does) run 
a kind of sound quality testing business. Steve was kind enough to lend me his 
facilities and a technician for a day so that I could measure HRTFs on his B&K 
head. The main thing that I learned is that this takes more than a day. We 
spent a great deal of time trying to get the wobble out of the turntable as 
measured by a plumb bob hanging from a very tall step ladder. What is the time 
constant of a plumb bob on a long string? Plus, the turntable was not a fine 
B&K unit but a consumer (home use) unit with a piece of plywood attached with 
degree tics in one degree increments. The turntable had a very small amount of 
wobble but it was enough to make trying to get say < 1/2 wavelength at 20 KHz a 
real challenge. I never trusted those results and never used them.

Some years later I gained access to a KEMAR. I can't remember which model of 
ear(s) it had but there was a visibly thin area in the conch area that would 
almost certainly cause accuracy problems. I contacted the Knowles people about 
this and they sent another ear, but it was exactly the same. I mean, you could 
see light shining through it, it was so thin. I applied a bit of rubber cement 
to the back side to thicken it a little. I don't know if Knowles' mold had worn 
over the years and I wonder how widespread this is. I kind of thought at the 
time that it could be the source of some odd things that I noticed in the upper 
frequencies of some KEMAR HRTFs. This is sort of important because the 
nonminimum phase nature of some HRTFs is due, I believe, to the focused 
reflection of sounds arriving from generally frontal directions back into the 
ear canal. In a "normal" near, I suspect that this reflection is stronger than 
the direct sound which would cause a nonminimum  phase respons
 e, and the frequency range for the quarter-wavelength one-way path difference 
required to cause a notch is just right for the notch that is seen at around 
8-10 KHz.

Jerry
_______________________________________________
Sursound mailing list
Sursound@music.vt.edu
https://mail.music.vt.edu/mailman/listinfo/sursound

Reply via email to