#1 open air results off by order of magnitude

[Mark Whitis]

I tried to create a model for an open air line. More specifically, I
was trying to test the effect of a dialectric under the line.

So, I took your twin-wire-1.bmp and I spread the rods out and doubled the
image size so that they were separated by 6 diameters center to center
which should give an impedance of 300 ohms. The program reports about
600 ohm. Ok, I figured the image was too small, so I enlarged it to
around 500x1000 pixels (vs 30 pixels wires with 180 pixel spacing). Now,
my problem is that the calculated impedance is WAY too low (order of
magnitude). Seems that the far field may be given way too much weight.

Because of image size, I am only including the large file with dialectric,
it can be removed with "-d" option. The results are the same as
phsyically removing the dialectric from the image file.

I am using the version (4.6.0) that comes prepackaged for debian
linux. Single CPU.

I also notice an asymetry in the field result images.
#without dialectric
atlc twin-wire.bmp
twin-wire.bmp 2 Er= 1.00 Zo= 522.647 Ohms C= 6.4 pF/m L= 1743.4 nH/m
v= 2.998e+08 m/s v_f= 1.000 VERSION= 4.6.0

#with dialeectric
atlc twin-wire-2.bmp
twin-wire-2.bmp 2 Er= 1.41 Zo= 439.677 Ohms C= 9.0 pF/m L= 1743.4 nH/m
v= 2.522e+08 m/s v_f= 0.841 VERSION= 4.6.0

# with dialectric
atlc twin-wire-3.bmp
twin-wire-3.bmp 2 Er= 1.42 Zo= 23.231 Ohms C= 171.3 pF/m L= 92.4 nH/m
v= 2.513e+08 m/s v_f= 0.838 VERSION= 4.6.0

# without dialectric
atlc -d ffff00=1.0 twin-wire-3.bmp
twin-wire-3.bmp 2 Er= 1.00 Zo= 27.706 Ohms C= 120.4 pF/m L= 92.4 nH/m
v= 2.997e+08 m/s v_f= 1.000 VERSION= 4.6.0

Mail to the *.kirby@onetel.net addres bounced.

[Mark Whitis]

twin-wire.bmp, twin-wire-2.bmp, twin-wire-3.bmp

[Dr. David Kirkby]

Logged In: YES
user_id=160407
Originator: NO

I suspect the program is having problems computing the fields when the fields are not defined on the boundaries. It may be possible to try to improve this by setting the field to be zero at the outer edge, as long as the edge is sufficiently far from the conductors. I can't really off too much suggestion on this. I am aware the accuracy decreases under such circumstances, but don't really know the best way to fix it.

[Dr. David Kirkby]

Sorry for the delay in replying - not quite 3 years. I never see your bug report until today.

I'm not too surprised by the problems you report with open-wires. I've not really tested them much. The problem is much more difficult, as the fields extend to infinity. This leaves one the problem of what are the boundary conditions.

At the time ATLC was developed, I did not have the computing power to handle very fine grids. Since machines have become much faster, I've investigated using much finer grids. In theory they should be better, but in practice if made sufficiently large they can become inaccurate, even on fully closed lines like coax. I believe the reason is that rounding errors propagate badly if there are too many points. The code determines the voltage at a point, which for a simple case is just the average of the voltages around that point. But for very fine grids, any errors at one point will induce errors at another. It's probable that for a sufficiently large number of points, the rounding errors in the floating point processor may become significant.

I don't have time to investigate this any further now. Part of the problem is that there are not any exact equations on which I can make comparisons with atlc, like there is for closed structures with one dielectric.