🚀 go-pugleaf

It is possible to take whatever function you are using to fit your response
curves and determine best-fit parameters using pretty much any minimization 
type program. If there is difficulty in writing the derivative analyitically,
you can still use the Simplex algorithm. This procedure has been used to fit
complex impedance data to model circuits. The trick is to modify the object
function to account for logarthmic behavior. For example, normally one uses
F(x,y,z)-F(x.,y.,z.)  and minimize this
(where x, y, and z are the calculated and x., y., and z. are the
data values.. The most simplistic way of accounting for large value changes
over a wide range of data is to simply divide the difference by the data value.
Of course, one can become much more creative.... The simplex algorithm works
with any function that you can write.. no need to differentiate.
See Marco Caceci, Wiliam Chacheris in an article "FItting Curves to Data"
in Byte, Mqy 1984, P 340-62 (best) or
William Press et. al in Numerical Recipes

In article <1992Sep8.083909.1@zodiac.dsto.oz.au>, st_waldman@zodiac.dsto.oz.au writes:
> 
> Hi,
> 
> I am interested in curve-fitting sound pressure and impedance
> response functions in the frequency domain using a rational
> fraction polynomial formulation. Does anyone have any source
> code to such a routine in Fortran, Pascal, or C?.
> 
> I have seen a paper titled "A Frequency Domain Curve Fitting
> Algorithm With Improved Accuracy", by Jim Adcock (Hewlett-
> Packard) and Ron Potter (Consultant) that was presented at a
> modal analysis convention a few years ago.
> 
> The above algorithm can fit to data that is log-spaced in the
> frequency domain, which is the common form for audio/acoustic
> measurements as published in loudspeaker manufacturers data
> sheets.
> 
> My application is related to modelling of loudspeaker systems so
> that it will be possible to design crossover networks that take
> into account each driver's natural response characteristics.
> 
> Any suggestions or pointers where I could obtain program source
> code that implements this algorithm will be gratefully received.
> 
> Maybe even someone from Hewlett-Packard can help me?
> 
> Thank you.
> 
> -----------------------------------------------------------------------
>  Witold Waldman
>  Aeronautical Research Laboratory   Internet: witold@hotblk.dsto.gov.au
>  Melbourne                          Fax     : +61 3 646 7771
>  Australia                          Tel     : +61 3 647 7488
> -----------------------------------------------------------------------
> 
>