One thing not to do with NEC2

I had this bright idea, sort of. I plugged a commercially made USB WiFi adapter into a USB extension cord, clipped it to a window curtain up high, and started stationary wardriving. I was amazed at what I could see considering there are no other houses within 1/4 mile, but of course all the signals are weak and I need a better antenna. With the USB collinear I don't need to mess with expensive bulky coax and trying to connect it to a WiFi card, I can just wrap a reflector around the collinear for directivity and more gain. But what shape parabola? I wanted to model it to see.

I've got some 1/2 inch hardware cloth which is almost right (should be 3/8) so I wrote a program that generates a NEC file to model a reflector, added a simple dipole at the focus for now and let fly:

But when I tried to look at the RF coming out of it, all I can get is a big omnnidirectional ball (see bottom). Tried 2 different NEC2 versions, same thing. Tried doing it in Mmana-gal, can't use more than 512 wires with the free version. So I'm thinking "too many wires, too close together, I've heard of that".

Then I made the leap into using surface patches instead, a new frontier for me, but it wasn't that hard: now my program has 3 different output subroutines, same data. I've got some aluminum flashing I could use instead of hardware cloth.

But when I look at the gain pattern I still get this big stupid ball. I think the key is in the maxgain and vgain numbers, I think they're indicating an overflow condition.

If I were to move the wires farther apart it might get through NEC2, but then it would be useless as a reflector. Don't bother doing this, this is what you'll get. The antenna probably works, modeling it doesn't.

If you've got any use for it, the NEC file and C program to generate it are here: wo.tar.xz. I don't remember why I named it wo. I had in mind to set it up to take quadratic coefficients for a Y = ax^2 + bx + c equation to model different parabolas but I sort of abandoned that idea once I found I couldn't feed the output through NEC2.
Alan Corey, ab1jx

7/16/2016:
I actually did build this antenna, still have it. I used aluminum flashing for the reflector, the top and bottom are held in place by 1/4 inch plywood. I printed out 2 of the PDF above and glued to the plywood pieces then cut along the line with a saber saw. Half fits inside the reflector and one outside, they clamp it in place quite nicely. A pair of these at the top and another at the bottom.

But does it work? Not really. I didn't do any very quantitative testing but using WiFi Analyzer on a phone and watching the signal strength meter I can tilt the antenna all over the place and not see much change in the signal. It's just about as bad as modeled. I have no idea why. Here it is built. That's a commercial gain USB/WiFi antenna at the focus point, an LB-Link 6 DB gain collinear. I just plugged a USB extension cable into it to run it.

ab1jx / calcs