1.77781% Error
First off lets start with a comic that caught my attention:
Ah hits right at home.
Ok so whats up with that awesome percent error calculation? Well it is the error between my analytic Green Function solution on the axis of the rectangular waveguide compared to Ansoft solutions. I am so confident in my Green’s Function that I am running a second order ansoft interpretation solution to maybe clean up some error.
(orange is the green’s functions)
I expect the main error comes in at the ends where the discrete convolution would muck up the solution.
On to cylindrical waveguide, then multiple slots. I might make the holy grail yet!
You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.
March 14th, 2008 at 1:42 am
Nice! That’s excellent. Is this with that multiplier term or did you find a different error?
Remember, each one is using different approximations — HFSS pretends the structure is discrete, your Green’s function assumes equivalent principle and infinite thin aperture, and so on…there will likely be some finite difference between the two. No pun intended. within 1% on analytical vs. computational is fantastic.
Second order? As in second order basis function? When did you get v11? The iterative solver is money.
March 14th, 2008 at 10:10 am
Got it a month back or so. They upgraded our license when I payed them more money.
What is the iterative solver?
Second order taxes my system hardcore! 17GB of RAM. I need to wait till this weekend to let it run.
March 14th, 2008 at 10:55 am
in the setup menu, you can choose an iterative matrix solver option, which speeds up the solution.
The second order basis functions aren’t always beneficial. Sometimes it eats up too much memory, although at the same time it allows the mesh to be more coarse than with a first order* basis function. It’s a trade off that you have to try and see whether the extra order on the basis function will help.
*by the way, ansoft calls their basis functions a little different than normal — their 0th order is normally considered 1st order, since it is basically pulse functions. Their 1st order is actually a second order basis function, and their second order is actually a third order basis function…I don’t know why they label it the way they do.
March 14th, 2008 at 2:43 pm
its the same dumb thing when you go to the results section and plot 1d results… on two axis