Brent or Newton Root finding in DLL

  • 16 December 2019
  • 1 reply
  • 5 views

Anyone know of an example user defined surface DLL, that uses Brent–Dekker or Newton root-finding for the ray-surface intercept?

1 reply

Userlevel 7
Badge +3
AFAIK, all the sample files use a generic Newton-Raphson method, but it's no problem to replace that with something better suited to the surface under development.



We use Newton-Raphson in the UDS examples because it is reliable over a very wide range of surface types. You can use methods that converge faster, but you also generally have to handle lots of special cases, and remember that during optimization you may have a specific set of parameter values that you would never enter by hand, so your code must be general enough to handle that.



I would personally stick with N-R unless I had a compelling reason to do otherwise. Is there a specific problem you're experiencing?

Reply