The site now passes all NIST fitting tests

[zunzun.com]

The site now passes all NIST fitting tests from

 

http://www.itl.nist.gov/div898/strd/nls/nls_main.shtml

 

     James

[Antoine Nguyen]

Hi James,

Thank you for your site, zunzun.com is wonderful!

Could you add a new function in NIST Gauss 2D, just the 3rd gaussian term :
y = a*exp(-b*x) + c*exp(-(x-d)² / f²) + g*exp(-(x-h)² / i² + j*exp(-(x-k)²/l²)

Antoine

[zunzun.com]

No problem.  I will post here when it is ready for testing.

James

[zunzun.com]

If you only have one or two data sets to fit, try the "User Defined Function" and generate whatever custom equation you wish.  Let me know of you still need the custom equation added.

James

On Monday, May 13, 2013 4:52:07 AM UTC-5, Antoine Nguyen wrote:

[Antoine Nguyen]

Thank you James,

It could be nice if it's added for people working with electron beam lithography to fit the point spread function.

I use "User Defined Function" but i haven't seen any différence with the NIST Gauss 2D. My function is there:
a * exp(-b * X) + c * exp(-((X-d) ** 2) / (f ** 2)) + g * exp(-((X-h) ** 2) / (i ** 2)) + j * exp(-((X-k) ** 2) / (l ** 2))
the last gaussian term  has no effect.

Here are parameters from zunzun:

    a =  4.2959285202346500E+06
    b =  2.2539555503209049E-01
    c =  2.9153139412909677E+04
    d =  1.1566695154995712E+03
    f =  1.4495453382919127E+03
    g =  9.6035937758306427E+03
    h =  4.3148590099164913E+01
    i =  1.9212373967295566E+01
    j =  6.8222179301733349E+05
    k = -2.0375826220249195E+01
    l =  3.2330163121693836E+01

and parameters adjusted by me, the last gaussian has effect :
    a = 4.2959285202346500E+06;
    b = 1.5539555503209049E-01;
    c = 2.4153139412909677E+04;
    d = 1.1566695154995712E+03;
    f = 1.5495453382919127E+03;
    g = 9.6035937758306427E+03;
    h=0;
    i = 2.9212373967295566E+01;
    j = 2.8222179301733349E+03;
    k=0;
    l = 3.2330163121693836E+03;

Best regards,

Antoine

[zunzun.com]

OK.  I will add the following:

y = g*exp(-(x-h)² / i² + j*exp(-(x-k)²/l²)

is that correct?

James

[Antoine Nguyen]

Hi James,

Juste the last term : j*exp(-(x-k)²/l²) since g*exp(-(x-h)² / i² is already in NIST Gauss 2D.

Thanks,

Antoine

[zunzun.com]

OK.

James

[zunzun.com]

Please try it:

http://zunzun.com/Equation/2/Engineering/Electron%20Beam%20Lithography%20Point%20Spread/

James

[Antoine Nguyen]

Hi James,

I'm sorry, i apologize for explaining badly! I meant adding j*exp(-(x-k)²/l²) to the NIST Gauss 2D, the final equation to add to zunzun.com is :

y = a*exp(-b*x) + c*exp(-(x-d)² / f²) + g*exp(-(x-h)² / i² + j*exp(-(x-k)²/l²)

Thanks,

Antoine

[zunzun.com]

OK.  Try now.

James

[Antoine Nguyen]

Thank You James,

It works well!

Best regards,

Antoine

[zunzun.com]

Excellent.  Rock on, Antoine!

James