[Sursound] DIY Ambisonic Mic

[Bob Cain]

Does anyone know the transfer function for compensation of the
tetrahedral array (Soundfield) microphone as a function of the diaphragm
diameter? I am thinking of constructing one from four of the
inexpensive, yet well regarded, 1" capsules that come in the Chinamics
from Bejeing. I'd like to record the A-format directly and then matrix
and compensate digitally to get to the B-format.

Perhaps 1" capsules are too large for this or unsuitable due to their
cardiod pattern. Opinions?


Thanks,

Bob
--
"Things should be described as simply as possible, but no simpler."

A. Einstein


////////////////////////////////////////\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

To contribute your unused processor cycles to the fight against cancer:

http://www.intel.com/cure

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\///////////////////////////////////////

[Angelo Farina]

The cardioid pattern is fine, but the 1" size is too big. This would cause
the mic spacing to be large (the faces will be on the surface of a sphere
having a radius of more than 25mm), and this will limit the high frequency
range of the velocity signals.
Regarding the derivation of the math function for compensating the signals,
I have to declare here that both the Gerzon's paper and the patent filed by
Gerzon and Craven are quite obscure. Perhaps Peter Craven could illuminate
us on this. I know only one man who was able to understand completely this,
and to develop even more refined processing functions: he is Mark Poletti,
and You should read his paper on the J.AES, december 2000.
In any case, I have a much more pragmatical solution: please, assembly Your
array, and measure the impulse response of the assembly (in A-format) in
free-field, along several directions.
Then do a first-order processing, which transforms the 4 signals coming from
the cartridges (FrontLeftUp, FrontRightDown, RearLeftDown, RearRightUp) in
the corresponding pre-B-format signals:
W' = FLU + FRD + RLD + RRU
X' = (FLU+FRD) - (RLD+RRU)
Y' = (FLU+RLD) - (FRD+RRU)
Z' = (FLU+RRU) - (FRD+RLD)
You need to measure at least 4 impulse response with the sound arriving from
different angles on the array (for example along X, along Y, along Z, and
along a diagonal direction - but more are the measurements, better will be
the results).
Then You set up a least-squares system, with the goal of finding 4
equalizing filters (FIR structures are the simpler to obtain), each of them
to be applied to the 4 pre-B-format signals, for producing the "True"
B-format:
W = Fw (x) W'
X = Fx (x) X'
Y = Fy (x) Y'
Z = Fz (x) Z'
The known term, in the linears system to set up, are the theoretical
responses of an ideal B-format microphone for a wavefront coming in the
direction correspondent to the measured A-format impulse response.
So if we measured 4 impulse responses:
- sound coming from X: W'x, X'x, Y'x, Z'x
- sound coming from Y: W'y, X'y, Y'y, Z'y
- sound coming from Z: W'z, X'z, Y'z, Z'z
- sound coming from 1st octant diagonal: W'd, X'd, Y'd, Z'd

Calling I the flat transfer function (or the Dirac's delta function, in time
domain), we obtain the following conditions:
I = Fw (x) W'x ; I = Fx (x) X'x ; 0 = Fy (x) Y'x; 0 = Fz (x) Z'x ;
I = Fw (x) W'y ; I = Fx (x) X'y ; 0 = Fy (x) Y'y; 0 = Fz (x) Z'y;
I = Fw (x) W'z ; I = Fzx(x) X'z ; 0 = Fy (x) Y'z; 0 = Fz (x) Z'z ;
I = Fw (x) W'd ; 0.577*I = Fx (x) X'd ; 0.577*I = Fy (x) Y'd; 0.577*I = Fz
(x) Z'd ;

Solving the least-squares linear system, we find the required equalizing
filters. A brute-force approach, as You see, but it should work. I never
attempted to use it for building a Soundfield-like microphone, but in the
past I used it for building a two-microphones sound intensity probe (before
I could afford to buy an original B&K probe), and it was effective. It
resulted to be much better than employing the theoretical equation, because
this way You are automatically compensating for the mismatch in amplitude
and phase responses between the capsules.
Bye!

Angelo Farina




----- Original Message -----
From: "Bob Cain" <***@znet.com>
To: "Sursound" <***@music.vt.edu>
Sent: Sunday, February 24, 2002 9:44 AM
Subject: [Sursound] DIY Ambisonic Mic

Post by Bob Cain
Does anyone know the transfer function for compensation of the
tetrahedral array (Soundfield) microphone as a function of the diaphragm
diameter? I am thinking of constructing one from four of the
inexpensive, yet well regarded, 1" capsules that come in the Chinamics
from Bejeing. I'd like to record the A-format directly and then matrix
and compensate digitally to get to the B-format.
Perhaps 1" capsules are too large for this or unsuitable due to their
cardiod pattern. Opinions?
Thanks,
Bob
--
"Things should be described as simply as possible, but no simpler."
A. Einstein

////////////////////////////////////////\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
\\\

Post by Bob Cain
http://www.intel.com/cure

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\////////////////////////////////////
///

Post by Bob Cain
_______________________________________________
Sursound mailing list
http://mail.music.vt.edu/mailman/listinfo/sursound

[Bob Cain]

Hi, Angelo. I've come back to think about this and have some questions below.

Post by Angelo Farina
In any case, I have a much more pragmatical solution: please, assembly Your
array, and measure the impulse response of the assembly (in A-format) in
free-field, along several directions.
Then do a first-order processing, which transforms the 4 signals coming from
the cartridges (FrontLeftUp, FrontRightDown, RearLeftDown, RearRightUp) in
W' = FLU + FRD + RLD + RRU
X' = (FLU+FRD) - (RLD+RRU)
Y' = (FLU+RLD) - (FRD+RRU)
Z' = (FLU+RRU) - (FRD+RLD)
You need to measure at least 4 impulse response with the sound arriving from
different angles on the array (for example along X, along Y, along Z, and
along a diagonal direction - but more are the measurements, better will be
the results).
Then You set up a least-squares system, with the goal of finding 4
equalizing filters (FIR structures are the simpler to obtain), each of them
to be applied to the 4 pre-B-format signals, for producing the "True"
W = Fw (x) W'
X = Fx (x) X'
Y = Fy (x) Y'
Z = Fz (x) Z'
The known term, in the linears system to set up, are the theoretical
responses of an ideal B-format microphone for a wavefront coming in the
direction correspondent to the measured A-format impulse response.
- sound coming from X: W'x, X'x, Y'x, Z'x
- sound coming from Y: W'y, X'y, Y'y, Z'y
- sound coming from Z: W'z, X'z, Y'z, Z'z
- sound coming from 1st octant diagonal: W'd, X'd, Y'd, Z'd
Calling I the flat transfer function (or the Dirac's delta function, in time
I = Fw (x) W'x ; I = Fx (x) X'x ; 0 = Fy (x) Y'x ; 0 = Fz (x) Z'x ;
I = Fw (x) W'y ; I = Fx (x) X'y ; 0 = Fy (x) Y'y ; 0 = Fz (x) Z'y ;
I = Fw (x) W'z ; I = Fx (x) X'z ; 0 = Fy (x) Y'z ; 0 = Fz (x) Z'z ;
I = Fw (x) W'd ; 0.577*I = Fx (x) X'd ; 0.577*I = Fy (x) Y'd ; 0.577*I = Fz (x) Z'd ;

Should this not be:

I = Fw (x) W'x ; I = Fx (x) X'x ; 0 = Fy (x) Y'x ; 0 = Fz (x) Z'x ;
I = Fw (x) W'y ; 0 = Fx (x) X'y ; I = Fy (x) Y'y ; 0 = Fz (x) Z'y ;
I = Fw (x) W'z ; 0 = Fx (x) X'z ; 0 = Fy (x) Y'z ; I = Fz (x) Z'z ;
I = Fw (x) W'd ; 0.577*I = Fx (x) X'd ; 0.577*I = Fy (x) Y'd ; 0.577*I = Fz (x) Z'd ;

Post by Angelo Farina
Solving the least-squares linear system, we find the required equalizing
filters.

Now for the real question. How do I do that? :-)

Using a Toeplitz solver I know how to find Fw given W'x from

I = Fw (x) W'x

but I don't have a clue how to find each F given the over constrained set
of four (or more) conditions on it.


Thanks,

Bob
--
"Things should be described as simply as possible, but no simpler."

A. Einstein


////////////////////////////////////////\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

To contribute your unused processor cycles to the fight against cancer:

http://www.intel.com/cure

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\///////////////////////////////////////

[Eric Benjamin]

Angelo

I'd like to explore this a bit further. Actually, I guess I'm asking you to explore it further! I should say that I like this approach very much, since, as you say, it should compensate for errors in the response functions of the individual capsules.

Measuring the Impulse Responses of the microphone along the three principle axes, and then using a least-squares method to solve for the filters that force each IR to be a delta function, attempts to make the frequency response to be flat along each of those axes.

Does this accomplish all that is desired? Numerical analysis of the soundfield array shows that, for instance, the polar pattern of the W output becomes clover-leaf shaped at high frequencies. Measurements confirm that the actual SoundField microphone does this too. In his original paper describing the SoundField microphone Gerzon showed two filters to compensate the response of the W, and the X, Y, and Z outputs, to have flat diffuse-field response. But this does not seem to be what was done in practice. The W output of the SoundField Mk IV is quite flat, on axis, up to about 18 kHz or so. So this implies that the on-axis free-field response has been made flat, not the diffuse field response. Won't using the least-squares solution distribute the errors equally across frequency? Perhaps, instead, a solution should be sought that corrects the response up to only 10 kHz or so, and then lets the HF response do whatever.

Does the least-squares approach cause the polar patterns to be a close as possible to the desired zeroth and first order spherical harmonics?

What is the effect of an angular error in the apparatus when measuring the experimental IRs?

Perhaps most importantly, what is the effect of capsule polar patterns which change with frequency.

Comments anyone?

Subject: Re: [Sursound] DIY Ambisonic Mic
Date: Sun, 24 Feb 2002 14:42:21 +0100
The cardioid pattern is fine, but the 1" size is too big. This would cause
the mic spacing to be large (the faces will be on the surface of a sphere
having a radius of more than 25mm), and this will limit the high frequency
range of the velocity signals.
Regarding the derivation of the math function for compensating the signals,
I have to declare here that both the Gerzon's paper and the patent filed by
Gerzon and Craven are quite obscure. Perhaps Peter Craven could illuminate
us on this. I know only one man who was able to understand completely this,
and to develop even more refined processing functions: he is Mark Poletti,
and You should read his paper on the J.AES, december 2000.
In any case, I have a much more pragmatical solution: please, assembly Your
array, and measure the impulse response of the assembly (in A-format) in
free-field, along several directions.
Then do a first-order processing, which transforms the 4 signals coming from
the cartridges (FrontLeftUp, FrontRightDown, RearLeftDown, RearRightUp) in
W' = FLU + FRD + RLD + RRU
X' = (FLU+FRD) - (RLD+RRU)
Y' = (FLU+RLD) - (FRD+RRU)
Z' = (FLU+RRU) - (FRD+RLD)
You need to measure at least 4 impulse response with the sound arriving from
different angles on the array (for example along X, along Y, along Z, and
along a diagonal direction - but more are the measurements, better will be
the results).
Then You set up a least-squares system, with the goal of finding 4
equalizing filters (FIR structures are the simpler to obtain), each of them
to be applied to the 4 pre-B-format signals, for producing the "True"
W = Fw (x) W'
X = Fx (x) X'
Y = Fy (x) Y'
Z = Fz (x) Z'
The known term, in the linears system to set up, are the theoretical
responses of an ideal B-format microphone for a wavefront coming in the
direction correspondent to the measured A-format impulse response.
- sound coming from X: W'x, X'x, Y'x, Z'x
- sound coming from Y: W'y, X'y, Y'y, Z'y
- sound coming from Z: W'z, X'z, Y'z, Z'z
- sound coming from 1st octant diagonal: W'd, X'd, Y'd, Z'd
Calling I the flat transfer function (or the Dirac's delta function, in time
I = Fw (x) W'x ; I = Fx (x) X'x ; 0 = Fy (x) Y'x; 0 = Fz (x) Z'x ;
I = Fw (x) W'y ; I = Fx (x) X'y ; 0 = Fy (x) Y'y; 0 = Fz (x) Z'y;
I = Fw (x) W'z ; I = Fzx(x) X'z ; 0 = Fy (x) Y'z; 0 = Fz (x) Z'z ;
I = Fw (x) W'd ; 0.577*I = Fx (x) X'd ; 0.577*I = Fy (x) Y'd; 0.577*I = Fz
(x) Z'd ;
Solving the least-squares linear system, we find the required equalizing
filters. A brute-force approach, as You see, but it should work. I never
attempted to use it for building a Soundfield-like microphone, but in the
past I used it for building a two-microphones sound intensity probe (before
I could afford to buy an original B&K probe), and it was effective. It
resulted to be much better than employing the theoretical equation, because
this way You are automatically compensating for the mismatch in amplitude
and phase responses between the capsules.
Bye!
Angelo Farina

[Angelo Farina]

First of all, You definitely need to read the Mark Poletti's paper, which
discuss all this in great details. Following Poletti, a certain degree of
dorectivity from the capsules is useful, so a subcardioid or cardioid
response is not a problem.
Second: when I referred to the least squares method, I was referring to the
Kirkeby's formulation in frequency domain, which includes a regularization
parameter. This reg.parameter can be made frequency-dependent, and so You
can have accurate correction up to a certain limit frequency, and then
release the constraints...
Of course the angular errors during the measurement will affect the results:
the best would be to place the microphones over a rotating table, and repeat
the measurement at 5° angular steps. This way, You will have 72
measurements, and the least-squares method will provide an equalization
which is, on average, correct for any angle.
Bye!

Angelo Farina

----- Original Message -----
From: "Eric Benjamin" <***@pacbell.net>
To: <***@music.vt.edu>
Sent: Tuesday, February 26, 2002 5:03 PM
Subject: [Sursound] DIY Ambisonic Mic

Post by Angelo Farina
Angelo
I'd like to explore this a bit further. Actually, I guess I'm asking you

to explore it further! I should say that I like this approach very much,
since, as you say, it should compensate for errors in the response functions
of the individual capsules.

Post by Angelo Farina
Measuring the Impulse Responses of the microphone along the three

principle axes, and then using a least-squares method to solve for the
filters that force each IR to be a delta function, attempts to make the
frequency response to be flat along each of those axes.

Post by Angelo Farina
Does this accomplish all that is desired? Numerical analysis of the

soundfield array shows that, for instance, the polar pattern of the W output
becomes clover-leaf shaped at high frequencies. Measurements confirm that
the actual SoundField microphone does this too. In his original paper
describing the SoundField microphone Gerzon showed two filters to compensate
the response of the W, and the X, Y, and Z outputs, to have flat
diffuse-field response. But this does not seem to be what was done in
practice. The W output of the SoundField Mk IV is quite flat, on axis, up
to about 18 kHz or so. So this implies that the on-axis free-field response
has been made flat, not the diffuse field response. Won't using the
least-squares solution distribute the errors equally across frequency?
Perhaps, instead, a solution should be sought that corrects the response up
to only 10 kHz or so, and then lets the HF response do whatever.

Post by Angelo Farina
Does the least-squares approach cause the polar patterns to be a close as

possible to the desired zeroth and first order spherical harmonics?

Post by Angelo Farina
What is the effect of an angular error in the apparatus when measuring the experimental IRs?
Perhaps most importantly, what is the effect of capsule polar patterns

which change with frequency.

Post by Angelo Farina
Comments anyone?

Subject: Re: [Sursound] DIY Ambisonic Mic
Date: Sun, 24 Feb 2002 14:42:21 +0100
The cardioid pattern is fine, but the 1" size is too big. This would cause
the mic spacing to be large (the faces will be on the surface of a sphere
having a radius of more than 25mm), and this will limit the high frequency
range of the velocity signals.
Regarding the derivation of the math function for compensating the signals,
I have to declare here that both the Gerzon's paper and the patent filed by
Gerzon and Craven are quite obscure. Perhaps Peter Craven could illuminate
us on this. I know only one man who was able to understand completely this,
and to develop even more refined processing functions: he is Mark Poletti,
and You should read his paper on the J.AES, december 2000.
In any case, I have a much more pragmatical solution: please, assembly Your
array, and measure the impulse response of the assembly (in A-format) in
free-field, along several directions.
Then do a first-order processing, which transforms the 4 signals coming from
the cartridges (FrontLeftUp, FrontRightDown, RearLeftDown, RearRightUp) in
W' = FLU + FRD + RLD + RRU
X' = (FLU+FRD) - (RLD+RRU)
Y' = (FLU+RLD) - (FRD+RRU)
Z' = (FLU+RRU) - (FRD+RLD)
You need to measure at least 4 impulse response with the sound arriving from
different angles on the array (for example along X, along Y, along Z, and
along a diagonal direction - but more are the measurements, better will be
the results).
Then You set up a least-squares system, with the goal of finding 4
equalizing filters (FIR structures are the simpler to obtain), each of them
to be applied to the 4 pre-B-format signals, for producing the "True"
W = Fw (x) W'
X = Fx (x) X'
Y = Fy (x) Y'
Z = Fz (x) Z'
The known term, in the linears system to set up, are the theoretical
responses of an ideal B-format microphone for a wavefront coming in the
direction correspondent to the measured A-format impulse response.
- sound coming from X: W'x, X'x, Y'x, Z'x
- sound coming from Y: W'y, X'y, Y'y, Z'y
- sound coming from Z: W'z, X'z, Y'z, Z'z
- sound coming from 1st octant diagonal: W'd, X'd, Y'd, Z'd
Calling I the flat transfer function (or the Dirac's delta function, in time
I = Fw (x) W'x ; I = Fx (x) X'x ; 0 = Fy (x) Y'x; 0 = Fz (x) Z'x ;
I = Fw (x) W'y ; I = Fx (x) X'y ; 0 = Fy (x) Y'y; 0 = Fz (x) Z'y;
I = Fw (x) W'z ; I = Fzx(x) X'z ; 0 = Fy (x) Y'z; 0 = Fz (x) Z'z ;
I = Fw (x) W'd ; 0.577*I = Fx (x) X'd ; 0.577*I = Fy (x) Y'd; 0.577*I = Fz
(x) Z'd ;
Solving the least-squares linear system, we find the required equalizing
filters. A brute-force approach, as You see, but it should work. I never
attempted to use it for building a Soundfield-like microphone, but in the
past I used it for building a two-microphones sound intensity probe (before
I could afford to buy an original B&K probe), and it was effective. It
resulted to be much better than employing the theoretical equation, because
this way You are automatically compensating for the mismatch in amplitude
and phase responses between the capsules.
Bye!
Angelo Farina

[Bob Cain]

Post by Angelo Farina
First of all, You definitely need to read the Mark Poletti's paper, which
discuss all this in great details.

Is that paper available somewhere without paying dues to an
organization?

Post by Angelo Farina
Second: when I referred to the least squares method, I was referring to the
Kirkeby's formulation in frequency domain, which includes a regularization
parameter. This reg.parameter can be made frequency-dependent, and so You
can have accurate correction up to a certain limit frequency, and then
release the constraints...

Same question. Hopefully papers that go beyond a statement of
accomplishment (the usual paper these days) to a complete description of
how to repeat the work.


Thanks,

Bob
--
"Things should be described as simply as possible, but no simpler."

A. Einstein


////////////////////////////////////////\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

To contribute your unused processor cycles to the fight against cancer:

http://www.intel.com/cure

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\///////////////////////////////////////

[Angelo Farina]

It is on the AES Journal, december 2000. However, if You do not find it
easily, I can send You a copy.
Regarding the Kirkeby method, I admit that some of the papers published on
the argument by Nelson and Kirkeby are quite obscure. On the other hand,
also I have published something on this topic (together with Nelson and
Kirkeby), and I hope that my formulation is clearly understandable. You can
download all my papers from my web site (HTTP://pcangelo.eng.unipr.it).
But the trick is so simple that I can explain it here.
Let's consider the simple case in which You have to design just a single
inverse filter (a filter which, inserted on your signal path, makes the
overall trasnfer function to become a perfect DIrac's delta). We call h the
impulse response of the system, and f the impulse response of the inverse
filter. The method is as follows:
1) Take the FFT of h: H=FFT(h)
2) Take the complex reciprocal of each spectral line, adding the regulariz.
paramater:
F=Conjg(H)/(Conjg(H)*H+epsilon) where epsilon is the reg.parameters, which
can be made to vary with frequency.
3) Take the IFFT of the inverted spectrum: f=IFFT(F)
The method can be easily generalized to the simultaneous computation of N
inverse filters instead of just one.
Bye!

Angelo Farina
----- Original Message -----

Post by Bob Cain

Post by Angelo Farina
First of all, You definitely need to read the Mark Poletti's paper, which
discuss all this in great details.

Is that paper available somewhere without paying dues to an
organization?

Post by Angelo Farina
Second: when I referred to the least squares method, I was referring to the
Kirkeby's formulation in frequency domain, which includes a

regularization

Post by Bob Cain

Post by Angelo Farina
parameter. This reg.parameter can be made frequency-dependent, and so You
can have accurate correction up to a certain limit frequency, and then
release the constraints...

Same question. Hopefully papers that go beyond a statement of
accomplishment (the usual paper these days) to a complete description of
how to repeat the work.
Thanks,
Bob
--
"Things should be described as simply as possible, but no simpler."
A. Einstein

////////////////////////////////////////\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
\\\

Post by Bob Cain
http://www.intel.com/cure

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\////////////////////////////////////
///

Post by Bob Cain
_______________________________________________
Sursound mailing list
http://mail.music.vt.edu/mailman/listinfo/sursound

[Ronald C.F. Antony]

Post by Bob Cain
Hopefully papers that go beyond a statement of
accomplishment (the usual paper these days) to a complete description of
how to repeat the work.

Quite my sentiment. People should have to decide if they want
academic credit (i.e. honor in exchange for contributing something
to humanity), or if they want monetary rewards (i.e. satisfy their
greed).
Our science was built on people who did research and shared it.
Today, almost any school you go to, all you learn is what was
done, but not how. They sell the IP to the highest bidder and
get an academic title thrown in.
The rest remain in the dark, unless they are so close to duplicating
the research that they can guess reading between the lines as to
what was done and how.
I left business school because all I learned was that the professors
have consulting companies that would solve particular problems if
I ever encounter them in my career (rather than teaching me how to
solve them). Then I studied computer science, just to get similar
results: anything truly interesting was either treated very
abstractly because it was some DARPA funded project or it
was cosponsored by some big company that hogged the resulting
intellectual property. (sort of like the FM synthesis patents
that Yamaha sits on for decades...)

Ronald
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~~~
"The reasonable man adapts himself to the world; the unreasonable one
persists
in trying to adapt the world to himself. Therefore all progress depends
on the
unreasonable man." G.B. Shaw | rcfa @ cubiculum . com | NeXT-mail
welcome

[Richard Dobson]

Well, why does it have to be 'greed'? I happen to operate on an
extremely low income as a part-time music teacher. I would like to be
able to devote all my time to software development and research (an
Ambisonic DAW anyone?), but I am not fortunate enough to be a member of
an academic department, or company, that will pay me to do it. CDP is a
non-profit-making co-operative, so no fortunes to be made there either.
So if I can develop some modest product that I can sell, that may just
subsidise me enough to continue to give so much of my time and code to
public projects such as Csound.

Greed? Simple subsistence needs in my case. If I can't get an income
from programming, I simply have to find something else to do. My bank
does not give me money! It is perfectly possible to be both a commercial
programmer ~and~ an Open-Source one. Two-dimensional opinions about
greed really don't help! If that's the perception, why should I bother?

</rant>

Richard Dobson

Post by Ronald C.F. Antony
Quite my sentiment. People should have to decide if they want
academic credit (i.e. honor in exchange for contributing something
to humanity), or if they want monetary rewards (i.e. satisfy their
greed).

[DG Malham]

Fortunately, there are business models which get round the problems you
describe - we just have to get people to subscribe to them...

a) Driving instructor model - there are countless books, videos and
computer games describing how to drive a car, all the fundamentals are in
the public domain, but we all still need driving instructors and driving
schools - this model is exploited by the Open Source community.

b) The (original) patent model - all your knowledge (about the patent
subject) MUST be openly available but the King/President/ Grand High
Panjandrum grants you the exclusive rights to exploit it for long enough
for you to recuperate your costs and make a bit on top. Moreover,
individuals or research organisations are at liberty to copy what you have
done, so long as they don't try to sell it.

c) A hybrid scheme, such as we are using in patenting the new form of O
format (as published at the AES Surround conference in 2001), where I
negotiated with the University that it would be freely available for
non-commercial usage (including release in free software for
non-commercial purposes)) but commercial use would require royalty
payments.


I can understand where these academics are coming from, though I don't
approve. It can be very galling when, after being open with students,
discussing new ideas and approaches that you have thought of (but not
published) only to have them go out into industry and make a fortune from
them, often without crediting you at all, let alone thinking about buying
you a beer :-). The temptation to say "The hell with it, I want to make
some money, too" is very strong. So far I have resisted this (though I do
do commercial designs, consultancy, recordings, etc.) having always
published and talked about what I do. But, on the other hand, when I find
that my income as an academic is barely more than the guy who comes and
collects my refuse each week - not that I begrudge him what he gets, I
wouldn't do that for twice the money - and for a far shorter working
week....

Dave
PS - to all those independents out there struggling to make ends meet and
grumbling about "@#!!&! academics with their job security and control over
what they do, they don't know when they are well off" actually, I do,
that's why I'm still here,

Post by Ronald C.F. Antony

Post by Bob Cain
Hopefully papers that go beyond a statement of
accomplishment (the usual paper these days) to a complete description of
how to repeat the work.

Quite my sentiment. People should have to decide if they want
academic credit (i.e. honor in exchange for contributing something
to humanity), or if they want monetary rewards (i.e. satisfy their
greed).
Our science was built on people who did research and shared it.
Today, almost any school you go to, all you learn is what was
done, but not how. They sell the IP to the highest bidder and
get an academic title thrown in.
The rest remain in the dark, unless they are so close to duplicating
the research that they can guess reading between the lines as to
what was done and how.
I left business school because all I learned was that the professors
have consulting companies that would solve particular problems if
I ever encounter them in my career (rather than teaching me how to
solve them). Then I studied computer science, just to get similar
results: anything truly interesting was either treated very
abstractly because it was some DARPA funded project or it
was cosponsored by some big company that hogged the resulting
intellectual property. (sort of like the FM synthesis patents
that Yamaha sits on for decades...)
Ronald
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~~~
"The reasonable man adapts himself to the world; the unreasonable one
persists
in trying to adapt the world to himself. Therefore all progress depends
on the
welcome
_______________________________________________
Sursound mailing list
http://mail.music.vt.edu/mailman/listinfo/sursound

[Ronald C.F. Antony]

Post by Richard Dobson
Well, why does it have to be 'greed'? I happen to operate on an
extremely low income as a part-time music teacher.

That's not what I was talking about. I'm talking about tenured
professors who betray the principles the universities they
teach at were built upon. People who in public decry the
ideas of people like Humbolt, and advocate that Universities
should act as purveyors of talent to private industry, rather
than as independent sites of learning, research and the
dissemination of knowledge in an effort to widen the horizons
of the enterprise we call "the human experience"

So there's nothing wrong with commercial research that's
done for profit. But that's what companies should have
their own research labs for. Instead they make a few lousy
donations (ten thousand here, or there, or some free software
and hardware that costs them close to nothing, might even
be last years model or some returned merchandise), and
in exchange they get access to intellectual property that
in commercial terms is worth orders of magnitude more.

These are the same people who lobby politicians to not
fund basic research, and to press universities to focus
on "practical" matters.

What I argued against is the murder of the academic spirit
in academic institutions. We have cases where students can't
take a class before signing a non-disclosure agreement,
or where students refuse to answer an exam question, because
they work in a start up company that happens to work in
the same field, and they fear that answering the exam question
might violate their employment contract.
You get the idea. This is a far cry from the humanist idea
of what universities should be.

Greetings,

Ronald
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~~~
"The reasonable man adapts himself to the world; the unreasonable one
persists
in trying to adapt the world to himself. Therefore all progress depends
on the
unreasonable man." G.B. Shaw | rcfa @ cubiculum . com | NeXT-mail
welcome

[DG Malham]

Can't find a single thing to disagree with in your latest emails - more
power to your elbow!

Dave

[Chris Travis]

Message from Peter Craven forwarded to SurSound by Chris Travis..

Subject: Re: Fwd: [Sursound] DIY Ambisonic Mic
Date: Wed, 27 Feb 2002 15:35:42

...

My understanding is that Gerzon derived the equalisation in the
soundfield mic patent by considering the effect of the finite capsule
spacing, assuming them to be perfect and acoustically
transparent. This calculation is quite easy, but I haven't checked
whether the formula in the patent is a good approximation to the result.
In practice the capsules are not acoustically transparent, i.e. they
present acoustic obstruction, and one may worry about whether the air
trapped between them will cause some sort of cavity resonance. The
advantage of a least squares technique is that any such effects can be
measured and corrected for.
1) Frequency-indepenent matrixing, followed by individual
equalisation of each B-format signal. With perfect capsules,
W gets one equalisation and X, Y, X get another.
2) Individual equalisation of each capsule followed by (1) above
3) Frequency-dependent matrixing, i.e. a 4x4 matrix each of
whose entries is a filter.
Angelo presents equations that start off as if he is adopting approach
(1), but as one reads on I am not so sure. Approach (1) should
correct for the problems of acoustic obstruction and other geometrical
effects having tetrahedral symmetry, but not for capsule variations.
It is indeed not obvious what the 'ideal' equalisation would be. A
least-squares approach, with measurements providing approximately
uniform coverage over the sphere, will equalise the zeroth and first
order shperical harmonic components of the microphone response. That
is not the same as equalising the response in any particular
direction, and nor will the diffuse field energy response necessarily
be equalised, though it shouldn't be far off until we get to high
frequencies where the second order spherical harmonic is making a
significant contribution to the energy.
The measurement directions should preferably be chosen to minimise the
effect of spatial aliassing. The next order of symmetry up from
tetrahedral is the cube/octahedron. This would suggest making six
measurements along the X, -X, Y, -Y, Z and -Z directions. After that
one could consider a dodecahedron. A single diagonal measurement
would not advance the cause significantly (in fact it would beak the
symmetry).
All that we can do with equalisation and matrixing is to ensure that
the zero-order and first-order spherical harmonic information produced
by the four capsules is correctly routed to W, X, Y and Z. If there
is contamination from second and higher-order components (which in
practice there will be at high frequencies), this cannot be corrected
by the equalisation and matrixing. Polar diagrams will still be
screwy, and all we can do is perhaps to taper-off the HF response so
that the diffuse-field energy response, taking into account the
higher-order spherical harmonics, does not rise too much.
I hope this is helpful.
Best wishes to all,
Peter Craven

[Angelo Farina]

It's very useful!
Thanks Peter!
Bye!

Angelo
----- Original Message -----
From: "Chris Travis" <***@sonopsis.ltd.uk>
To: "Surround Sound discussion group" <***@music.vt.edu>
Sent: Wednesday, February 27, 2002 10:33 PM
Subject: Re: [Sursound] DIY Ambisonic Mic

Post by Chris Travis
Message from Peter Craven forwarded to SurSound by Chris Travis..

Subject: Re: Fwd: [Sursound] DIY Ambisonic Mic
Date: Wed, 27 Feb 2002 15:35:42

...

My understanding is that Gerzon derived the equalisation in the
soundfield mic patent by considering the effect of the finite capsule
spacing, assuming them to be perfect and acoustically
transparent. This calculation is quite easy, but I haven't checked
whether the formula in the patent is a good approximation to the result.
In practice the capsules are not acoustically transparent, i.e. they
present acoustic obstruction, and one may worry about whether the air
trapped between them will cause some sort of cavity resonance. The
advantage of a least squares technique is that any such effects can be
measured and corrected for.
1) Frequency-indepenent matrixing, followed by individual
equalisation of each B-format signal. With perfect capsules,
W gets one equalisation and X, Y, X get another.
2) Individual equalisation of each capsule followed by (1) above
3) Frequency-dependent matrixing, i.e. a 4x4 matrix each of
whose entries is a filter.
Angelo presents equations that start off as if he is adopting approach
(1), but as one reads on I am not so sure. Approach (1) should
correct for the problems of acoustic obstruction and other geometrical
effects having tetrahedral symmetry, but not for capsule variations.
It is indeed not obvious what the 'ideal' equalisation would be. A
least-squares approach, with measurements providing approximately
uniform coverage over the sphere, will equalise the zeroth and first
order shperical harmonic components of the microphone response. That
is not the same as equalising the response in any particular
direction, and nor will the diffuse field energy response necessarily
be equalised, though it shouldn't be far off until we get to high
frequencies where the second order spherical harmonic is making a
significant contribution to the energy.
The measurement directions should preferably be chosen to minimise the
effect of spatial aliassing. The next order of symmetry up from
tetrahedral is the cube/octahedron. This would suggest making six
measurements along the X, -X, Y, -Y, Z and -Z directions. After that
one could consider a dodecahedron. A single diagonal measurement
would not advance the cause significantly (in fact it would beak the
symmetry).
All that we can do with equalisation and matrixing is to ensure that
the zero-order and first-order spherical harmonic information produced
by the four capsules is correctly routed to W, X, Y and Z. If there
is contamination from second and higher-order components (which in
practice there will be at high frequencies), this cannot be corrected
by the equalisation and matrixing. Polar diagrams will still be
screwy, and all we can do is perhaps to taper-off the HF response so
that the diffuse-field energy response, taking into account the
higher-order spherical harmonics, does not rise too much.
I hope this is helpful.
Best wishes to all,
Peter Craven

_______________________________________________
Sursound mailing list
http://mail.music.vt.edu/mailman/listinfo/sursound

[Michael Dunn]

In practice the capsules are not acoustically transparent, i.e.
they present acoustic obstruction, and one may worry about whether
the air trapped between them will cause some sort of cavity
resonance.

I wonder if stuffing the space inside the tetrahedron with foam or
something would improve its performance, or make it worse.....


-----------------------------------------------------------------
Michael Dunn | Surround Sound Decoder & Stereo Enhancer
Cantares | Self-Amplified Speakers, Test Equipment
74 George St. | Ambisonic Surround Sound CDs and Recording
Waterloo, Ont. | [phone] (fax: [phone])
N2J 1K7 | ***@cantares.on.ca
Canada | http://www.cantares.on.ca/
-----------------------------------------------------------------

[DG Malham]

John Vanderkooy once suggested to me that the best way to control this was
not to have space inside the area at all, but to mount the capsules flush
with the surface of a sphere (and use quite a lot of capsules). Although
this sounds like the opposite of what you want, since it effectively makes
the capsules totally non-transparent, because it makes the situation fully
controlled - and because we understand (well, some people do, not entirely
sure that includes me, except on a good day :-) the 3-d wave equations for
sound diffusing around a sphere. it makes it much easier to design the
appropriate compensation.

Thoughts?
Dave

Post by Michael Dunn

In practice the capsules are not acoustically transparent, i.e.
they present acoustic obstruction, and one may worry about whether
the air trapped between them will cause some sort of cavity
resonance.

I wonder if stuffing the space inside the tetrahedron with foam or
something would improve its performance, or make it worse.....
-----------------------------------------------------------------
Michael Dunn | Surround Sound Decoder & Stereo Enhancer
Cantares | Self-Amplified Speakers, Test Equipment
74 George St. | Ambisonic Surround Sound CDs and Recording
Waterloo, Ont. | [phone] (fax: [phone])
Canada | http://www.cantares.on.ca/
-----------------------------------------------------------------
_______________________________________________
Sursound mailing list
http://mail.music.vt.edu/mailman/listinfo/sursound

[chris woolf]

If they are flush on a sphere how do you make them directional? Or are
you trying to work with a ball of omnis?

Chris Woolf

Post by DG Malham
John Vanderkooy once suggested to me that the best way to control this was
not to have space inside the area at all, but to mount the capsules flush
with the surface of a sphere (and use quite a lot of capsules). Although
this sounds like the opposite of what you want, since it effectively makes
the capsules totally non-transparent, because it makes the situation fully
controlled - and because we understand (well, some people do, not entirely
sure that includes me, except on a good day :-) the 3-d wave equations for
sound diffusing around a sphere. it makes it much easier to design the
appropriate compensation.
Thoughts?
Dave

[DG Malham]

Hmmm - I hope people actually can understand this post of mine, that's one
helluva convoluted sentence, even for me. I shall be laying on an evening
class in understanding my emails shortly.

Dave

Post by DG Malham
John Vanderkooy once suggested to me that the best way to control this was
not to have space inside the area at all, but to mount the capsules flush
with the surface of a sphere (and use quite a lot of capsules). Although
this sounds like the opposite of what you want, since it effectively makes
the capsules totally non-transparent, because it makes the situation fully
controlled - and because we understand (well, some people do, not entirely
sure that includes me, except on a good day :-) the 3-d wave equations for
sound diffusing around a sphere. it makes it much easier to design the
appropriate compensation.
Thoughts?
Dave

Post by Michael Dunn

In practice the capsules are not acoustically transparent, i.e.
they present acoustic obstruction, and one may worry about whether
the air trapped between them will cause some sort of cavity
resonance.

I wonder if stuffing the space inside the tetrahedron with foam or
something would improve its performance, or make it worse.....
-----------------------------------------------------------------
Michael Dunn | Surround Sound Decoder & Stereo Enhancer
Cantares | Self-Amplified Speakers, Test Equipment
74 George St. | Ambisonic Surround Sound CDs and Recording
Waterloo, Ont. | [phone] (fax: [phone])
Canada | http://www.cantares.on.ca/
-----------------------------------------------------------------
_______________________________________________
Sursound mailing list
http://mail.music.vt.edu/mailman/listinfo/sursound

http://groups.yahoo.com/group/SoundfieldMic
Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

[rolv-k-r]

What you essentially say, is that if you make it BAD, it is easier to correct
afterwards. ;-)
(Please don't get hung up in "BAD", instead look at the smiley, boyz.....)
-RK

===== Original Message

[DG Malham]

Omnis - my understanding of John's idea was that you measured the pressure
distribution over the sphere and could calculate what you needed from
that. I have no idea how well it would work, tho' it's worth noting that
as far as I am aware, the Soundfield can be made with omnis, it's just
that it is more efficient (meaning much less noisy) if you use capsules
with some directionality built in.

Dave

Post by chris woolf
If they are flush on a sphere how do you make them directional? Or are
you trying to work with a ball of omnis?
Chris Woolf

Post by DG Malham
John Vanderkooy once suggested to me that the best way to control this

was

Post by DG Malham
not to have space inside the area at all, but to mount the capsules

flush

Post by DG Malham
with the surface of a sphere (and use quite a lot of capsules).

Although

Post by DG Malham
this sounds like the opposite of what you want, since it effectively

makes

Post by DG Malham
the capsules totally non-transparent, because it makes the situation

fully

Post by DG Malham
controlled - and because we understand (well, some people do, not

entirely

Post by DG Malham
sure that includes me, except on a good day :-) the 3-d wave equations

for

Post by DG Malham
sound diffusing around a sphere. it makes it much easier to design the
appropriate compensation.
Thoughts?
Dave

_______________________________________________
Sursound mailing list
http://mail.music.vt.edu/mailman/listinfo/sursound