microwave office optimization problem

hi there,
I am trying to optimize a multi-section power divider,
after some iterations, there is an error which means the optimization can not be done by set conditions, e.g. some line widths must be smaller than zero ,...
now, I want to know how I can yield the best results?
if needed I can upload the file in this post.
thanks

You can set at beginning smaller limits of the tuned components. Also be careful setting the limits, to be reasonable numbers. The simulator allow initially to input any numbers.

I never met the message you mention: "some line widths MUST be smaller than zero" !

oh, sure this is not the exact error term but the cause! the error is for example
STEE.SL17 : (T/W2) must be < 1.00
where 1 is the lower limit.

this does not mean than any variable is to be smaller than zero! It just means that W2>T.
If however you do get one or more line width variables that the optimizer wants to converge to zero, it probably means the problem is not properly set up. Although optimizers can save a lot of time and hard work in RF design, the cannot remove the requirement for knowledgable problem set up. A solution must always exist, and often a unique one, plus boundary conditions can be important. If the problem is overdetermined, sensible initial conditions may help with converging to the required minima.

Hi

Couldn't agree more with ucah! There is a substantial underlying theory associated with multisection Willkinson power divider design. The theory is very similar to filter synthesis although the results are different. A good book to start with is:

J. A. G. Malherbe ? Microwave Transmission Line Couplers, Artech House, 1988

As it is not recommanded to design a filter without any close initial guess, the brutal optimization approach would not work with multisection Willkinson as well. With proper initial solution for line impedances and resistor values, optimizer should work well and give meaninigful results.

flyhigh

yeah, with proper initial guess, you may give a pretty good result ,
there is a paper that I used for initial choices:
"A class of broadband three port TEM-mode hybrids"
, but the error is still over there!