MIME-Version: 1.0 Content-Type: multipart/related; boundary="----=_NextPart_01D1CA31.8090B330" This document is a Single File Web Page, also known as a Web Archive file. If you are seeing this message, your browser or editor doesn't support Web Archive files. Please download a browser that supports Web Archive, such as Windows® Internet Explorer®. ------=_NextPart_01D1CA31.8090B330 Content-Location: file:///C:/D04BBC92/atlc2.htm Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset="us-ascii"
6/14/16
This is the manua=
l for
atlc2
Arbitrary Transmission Line Calculator
This program was inspired by the atlc program written by Dr. David=
Kirkby, G8WRB.
You submit a drawing showing the cross-sect=
ion
of a transmission line with any geometry.&=
nbsp;
From that, the program will use numerical methods to find Z0
and the other transmission line parameters. Atlc2 is free.
Download atlc2 for Win32=
now (2.4M bytes)
Download atlc2 for=
Win64
now (4.2M bytes)
(The Win32 version=
will
run on a 64-bit Windows system. The
only advantage of the Win64 version is it allows more than 13500 equations.=
)
=
Dif=
ferences
between atlc and atlc2
1. Atlc2 applies
Faraday’s Law to determine the current distribution inside the
conductors. From that it calc=
ulates
the inductance and skin effect resistance.=
Atlc used only the laws of electrostatics.
2. Atlc2 is well tho=
ught
out for unshielded lines, solving them rapidly and accurately. (Surrounding them with a ground shield just
slows down the program.)
3. The size of the
simulation is not related to the size of the drawing. The E field simulation is always 3=
200 x
3200 pixels.
4. If there is a thi=
rd
conductor, it can either be pinned to ground potential or it can float to an
unforced voltage level.
5. Atlc2 is a Windows
program. Atlc was a Unix prog=
ram,
not especially user friendly.
6. Atlc2 can interna=
lly
generate some simple geometries. In
such cases no drawing file is necessary.
Atlc2 will accept bitmaps created for atlc.=
Gen=
eral
description
Atlc2 computes Z0, vf, L, C, RS,
GP, and VO for any geometry. (VO is the offset volta=
ge
necessary for no radiation.)
There are formulas for Z0 for si=
mple
geometries, but they are not always accurate. If the geometry is extreme or
complicated, has regions with dielectrics, or if the skin-effect resistance=
is
wanted, then a program like this one is your best hope.
If used properly, atlc2 results are accurat=
e to
1% or better. But it is not
fast. It is compute intensive=
. Some speedups are employed that wo=
rk
well for some geometries, but not others.&=
nbsp;
It can be hard to predict whether a run will take seconds or hours.<=
span
style=3D'mso-spacerun:yes'> Run atlc2 on your fastest computer=
.
Microsoft Paint is a good program for creat=
ing
the drawing, but any such bitmap editor will work. There must be no de-aliasing, so
Photoshop is out and JPEG files will not work. BMP format is assumed but TIFF and=
PNG
files will also work.
Normally red, green, blue, and cyan are
conductors and all other colors are insulators. But other color schemes are
allowed. Color definitions ca=
n be
changed at the keyboard or they can be loaded from a file. The line conduct=
ors
can be any 2 of the 3 primary colors, but if all 3 appear then green is ass=
umed
to be ground. Most commonly r=
ed is
+1 volt (actually +sin ωt), blue is –=
;1,
green is 0, and cyan is varying.
Unlike atlc, atlc2 requires a frequency, a
definite size, and a conductor resistivity. But you can specify anything if th=
ey
have little effect on the results.
Full red, green, and blue (x0000FF, x00FF00, and xFF0000) are now
copper. But slight off-shades=
are
defined for other common metals.
Atlc2 has 45 predefined colors for the common materials.
Cur=
rent version
of atlc2: 1.00 6/14/16
This is the first version I am proud of. It has no bugs, so far as I know.<=
span
style=3D'mso-spacerun:yes'> If you see anything that might be =
a bug,
please report it to me at kq6qv@aol.com . I will fix it quickly.
Previous versions had problems with charge prediction that somet=
imes
caused them to run very slowly.
This has been mostly fixed.
Some small changes to the program:
1. A new checkbox =
allows
E-field prediction (charge prediction) to be skipped. This makes the run time different,=
but
does not change the ultimate numerical result. E-field prediction usually speeds =
up the
program, but for some very large examples makes it slower.
2. Gp calculation =
is
new. (atlc2 log.txt is affect=
ed.)
3. If you created a
MoreColors.txt file, it should be changed to include tan(delta) values. If you do not know your dielectric=
's
tan(delta) or don't care, specify a 0.
4. I decided to ch= ange the name of "C and vf" to "C and Gp". This is a purely cosmetic change.<= o:p>
5. Clicking on
"File>Save displayed bitmap as..." now saves the full 3200 x 3=
200
map, with no zooming.
6. PNG and TIFF fi=
les
now work again. Saved images =
will
have the same file format as the Usermap.
7. C and Gp now em=
ploys
up to 32 threads.
8. Rs
prediction is now accurate to 5 %.
9. Rescaling is ne=
w,
allowing Usermaps from files to be resized. (It looks like zooming, but actual=
ly
changes the number of usermap pixels.)
10. C and Gp predi=
cted L
and Zo by assuming vf=3D1. L =
and Rs predicted C and Zo by assuming vf=3D1. Both have been changed to make a g=
uess
about vf. The guess averages =
the permittivities of all the dielectric pixels, weighing=
them
equally. This is often a rath=
er
poor guess, but it is usually closer than assuming vf=3D1.
11. Scripting
changes: The entire script fi=
le is
read into RAM and closed before the script is executed. The "erase" command no l=
onger
terminates atlc2. The command
"restrict" has been replaced by "box 1", and
"CVF" has been replaced by "CGP". "Conductances.txt" is
new. Scripts can call other s=
cripts
(key F8).
12. E-field lines =
is new. =
Line
density is adjustable.
Ins=
talling
atlc2
Atlc2 is a Windows program. It will run on Window 2000 and
newer. It might run on some o=
lder
platforms. The 64-bit version is required if you need more than 13,500
equations.
Try=
outs: atlc2 can be executed directly upon
download. It does not have to=
be
saved to disk.
There is no formal installation procedure.<=
span
style=3D'mso-spacerun:yes'> It is a single .exe file and does not
require any other files. It i=
s not
a registered program and does not show up in the Windows list of installed
programs.
You probably should create a new folder and=
put atlc2.ex=
e in that folder. We shall call it the “execut=
ion
folder”. It can have an=
y name
but should not be read-only. =
Atlc2
will record the results of any run that takes longer than 1 minute. Those results go into a file calle=
d altc2 log.txt in the execution folder.
Drawing files and any files produced by atl=
c2
can reside in any folder, but the open and save dialogs will always start at
the execution folder. So it i=
s most
convenient to keep your atlc2 files there.
You may want to create a file that declares
colors as conductors or insulators, and assigns physical characteristics to
them. This file must be calle=
d MoreColo=
rs.txt and it must be in=
the
execution folder.
Use=
rmap: Preparing the drawing
(De-aliasing is a blurring of edges that ma=
kes
them less jagged. But it crea=
tes
unintended colors that atlc2 will not recognize. So you must use an editor that avo=
ids
that.)
Commonly red and green are used for fully
shielded lines, such as coaxial cable.&nbs=
p;
Otherwise use red and blue.
When red and blue are both present, they are adjusted by adding V
The drawing can be in any folder and have a=
ny
name. But finding the file is
easier if you put it in the execution folder and give it the name
When the checkbox “Restrict to skin
depth” is checked, the program automatically blackens out conductor
pixels more than 3δ from the surface, where δ is the predicted sk=
in
effect depth. (So setting the
frequency too low makes the program run longer.) (Using this feature causes a decre=
ase in
accuracy of less than 1%.)
The pixels at the edge of the Usermap are
special. They are replicated
outward until the map is 3200 x 3200.
(The four corner pixels are doubly special. They are replicated in two dimensi=
ons.)
Floating wires cannot carry a net current, =
but
they can carry local currents (eddy currents). That is, there will be return curr=
ents
in other floating pixels (so that they sum to zero).
All +1 pixels (usually red) are electrically
connected to a voltage source. Thus
they are connected to each other.
The same is true for -1 wires (usually blue) and ground wires (usual=
ly
green). But floating wires are
connected to each other only if they are exactly the same color. So eddy currents will not link flo=
ating
wires of different color.
A grounded third wire (usually green) can c=
arry
a net current. But if it does=
then
the stated impedance is not a true characteristic impedance. If the ground current is more than=
a few
% then the impedance the program reports should be disregarded.
Red Gree=
n Blue =
&nb=
sp;
use &n=
bsp;
resistivity
permittivity
tanδ permea=
bility name
█π=
8;█ 255 0 0 x0000FF +1 1.7241 1 =
1 copper
█`=
08;█ 0 255 0 x00FF00 0
1.7241 1 =
1 copper
█`=
08;█ 0 0 255 xFF0000 -1 1.7241 1 =
1 copper
█`=
08;█ 0 255 255 xFFFF00 float 1.7241 1 =
1 copper
█&=
#9608;█ 224 31 31 x1F1FE0 +1 2.62 1 =
1 aluminum
█&=
#9608;█ 31 224 31 x1FE01F 0 2.62=
1 =
1 aluminum
█&=
#9608;█ 31 31 224 xE01F1F -1 2.62 1 =
1 aluminum
█&=
#9608;█ 31 224 224 xE0E01F float 2.62 1 =
1 aluminum
█&=
#9608;█ 224 31 0 x001FE0 +1 1.62 1 =
1 silver
█&=
#9608;█ 31 224 0 x00E01F 0 1.62=
1 =
1 silver
█&=
#9608;█ 31 0 224 xE0001F -1 1.62 1 =
1 silver
█&=
#9608;█ 31 255 224 xE0FF1F float 1.62 1 =
1 silver
█&=
#9608;█ 224 0 31 x1F00E0 +1 2.44 1 =
1 gold
█&=
#9608;█ 0 224 31 x1FE000 0 2.44=
1 =
1 gold
█&=
#9608;█ 0 31 224 xE01F00 -1 2.44 1 =
1 gold
█&=
#9608;█ 0 224 224 xE0E000 float 2.44 1 =
1 gold
█&=
#9608;█ 224 63 63 x3F3FE0 +1 9.71 1 =
1 * steel
█&=
#9608;█ 63 224 63 x3FE03F 0 9.71=
1 =
1 * steel
█&=
#9608;█ 63 63 224 xE03F3F -1 9.71 1 =
1 * steel
█&=
#9608;█ 63 224 224 xE0E03F float 9.71 1 =
1 * steel
█&=
#9608;█ 224 0 63 x3F00E0 +1 11.4 1 =
1 tin
█&=
#9608;█ 0 224 63 x3FE000 0 11.4=
1 =
1 tin
█&=
#9608;█ 0 63 224 xE03F00 -1 11.4 1 =
1 tin
█&=
#9608;█ 0 255 224 xE0FF00 float 11.4 1 =
1 tin
█&=
#9608;█ 24 63 0 x003FE0 +1 14.5 1 =
1 60/40 P=
bSn
solder
█&=
#9608;█ 63 224 0 x00E03F 0 14.5=
1 =
1 60/40 PbSn solder
█&=
#9608;█ 63 0 224 xE0003F -1 14.5 1 =
1 60/40 PbSn solder
█&=
#9608;█ 63 255 224 xE0FF3F float 14.5 1 =
1 60/40 PbSn solder
███ 0 0 0 x000000 insul 1000000 1 0 1
███
255 255 255 xFFFFFF=
insul 1000000 1 0 1
█&=
#9608;█ 255 202 202 xCACAFF=
insul 1000000 1.0006 0 1 air
█&=
#9608;█ 130 53 239 xEF3582 insul 1000000 2.07 0.00020 1 teflon<=
/span>
█&=
#9608;█ 142 142 142 x8E8E8E insul 1000000 2.2 0.00090 1 duroid<=
/span>
5880
█&=
#9608;█ 255 0 255 xFF00FF insul 1000000 2.26 0.00064 1 polyeth=
elene
█=
9608;█ 255 255 0 x00FFFF insul 1000000 2.5 0.00033 1 polystyrene
█&=
#9608;█ 239 204 26 x1ACCEF insul 1000000 4.5 0.011 1 polyvinylchloride
█&=
#9608;█ 188 127 96 x607FBC insul 1000000 3.3350 0.03 1 epoxy resin
█&=
#9608;█ 223 247 136 x88F7DF insul 1000000 3.7 0.018 1 FR4 PCB
█&=
#9608;█ 26 239 179 xB3EF1A insul 1000000 4.8 0.018 1 fiberglass/epoxy PCB
█&=
#9608;█ 105 105 105 x696969 insul 1000000 6.15 0.00270 1 duroid<=
/span>
6006
█=
;██ 220 220 220 xDCDCDC=
insul 1000000 10.2 0.00230 1 duroid<=
/span>
6010
█&=
#9608;█ 213 160 77 x4DA0D5 insul 1000000 100 0 1
█&=
#9608;█ 100 200 255 xFFC864 insul 1000000 75 0.157 1 distilled water
█&=
#9608;█ 176 224 200 xC8E0B0 insul 1000000 3.78 0.00006 1 quartz
█&=
#9608;█ 153 255 153 x99FF99 insul 1000000 5.0 0.00540 1 glass
* Atlc2 cannot presently handle ferromagnet=
ic
materials.
The tanδ values above are rough typical
values. The program assumes
tanδ is constant. But fo=
r most
materials, tanδ varies slightly with frequency and manufacturer. (Tanδ specifies the dielectri=
c loss. It is used only in the calculation=
of
Gp.)
Best
practices
The Usermap can have any dimensions. A larger Usermap (a map with more =
pixels
per conductor) will run slower, but a smaller map will not represent fields=
or
curved surfaces as well. The =
gap
between conductors must never be less than 2 pixels, and accurate results
usually require a gap of at least 5 pixels.
There is no need or benefit to having empty
space around the transmission line.
(This is true for shielded and unshielded lines.) The Usermap can be the smallest ma=
p that
describes the line.
Usually the overriding goal is a Usermap wi=
th
1000 to 3000 conductor pixels, which will solve for L and Rs in a couple minutes.&=
nbsp;
Execution time rises dramatically with the conductor pixel total
(roughly the cube of the equation total).&=
nbsp;
Solving for L and Rs ta=
kes:
200 conductor pix=
els - =
1
second &=
nbsp; &nbs=
p; =
on
a 1.5 GHz Pentium 4
400 conductor pix=
els - =
4
seconds =
&nb=
sp; on
a 1.5 GHz Pentium 4
800 conductor pix=
els - =
1.4
minutes =
&nb=
sp; on
a 1.5 GHz Pentium 4
1600 conductor pi=
xels - 13.0
minutes =
&nb=
sp; on
a 1.5 GHz Pentium 4
3200 conductor pi=
xels - 100.6
minutes =
&nb=
sp; on
a 1.5 GHz Pentium 4
3200 conductor pi=
xels - 7.0 m=
inutes &=
nbsp; &nbs=
p; on
a 3.33 GHz Xeon, 1 thread
3200 conductor pi=
xels - 1.2 m=
inutes &=
nbsp; &nbs=
p; on
a 3.33 GHz dual Xeon, 12 threads
6400 conductor pi=
xels - 9.0 m=
inutes &=
nbsp; &nbs=
p; on
a 3.33 GHz dual Xeon, 12 threads
12800 conductor p=
ixels
- 90.0 minut=
es &=
nbsp; &nbs=
p; on
a 3.33 GHz dual Xeon, 12 threads
25600 conductor p=
ixels
- 14 hours &=
nbsp; &nbs=
p; on
a 3.33 GHz dual Xeon, 12 threads (64-bit version)
So it is best to let the “pixels per
inch” be determined by whatever scale gets a reasonable execution
time. Execution times to solv=
e for C and Gp are more variable, but r=
uns
can be ended early if convergence is seen.
C a=
nd Gp finds the capaci=
tance
by summing the energy in the field, so some capacitance is missed if a
considerable field extends beyond the 3200 x 3200 area. Thus while Usermaps as large as 32=
00 x
3200 are allowed, C and Gp bec=
omes
inaccurate if the conductors are not confined to the central 1600 x 1600
region. Ground planes and ful=
ly
shielded lines are exceptions to that.
Get=
ting
an accurate Rs
If you violate this, atlc2 will report the =
Rs value using a red font as a warning. To work around that, you will need=
to
make a larger model (with more conductor pixels) and then hollow it out to
reduce the pixel count.
Atlc2 might be the only program anywhere th=
at
can predict Rs within 5% for any wires with any
cross-sections.
Run=
ning
atlc2
Atlc2 begins by showing one of its internal=
ly
generated examples. The normal
procedure is to click File>Open=
and select the BMP file that you have prepared. You must type in a “Pixel
width” when using an external bit map. Then click Solve>Solve fully.
Or File>New will sel=
ect a
different internal example.
Wherever atlc2 asks for a numerical value, =
the
entered value can be an integer or floating point number. The numerical value can be followe=
d by a
power-of-ten suffix in “e” form. For example, 13e-3 would be 13
thousandths of an inch. Addit=
ional
suffixes are allowed:
in &=
nbsp; (inches)
m &=
nbsp; (meters)
cm &=
nbsp; (centimeters)
mm &=
nbsp; (millimeters)
AWG =
(American
wire gauge)
Inches are assume=
d if
there is no suffix.
Some keyboard com=
mand
keys are available:
U &=
nbsp; Show
without E or V field
V &=
nbsp; Show
the V field intensity
E &=
nbsp; Show
the E field intensity
D &=
nbsp; Show
the D field intensity&=
nbsp;
(where D is εE)
L &=
nbsp; &nbs=
p; Show
the V field as a lines plot. =
(These
lines are often perfect circles.)
N &= nbsp; Show the E field as a lines plot (lines of force direction)<= o:p>
B &=
nbsp; Show
both the E and V field lines
J &= nbsp; &nbs= p; Show J-field, not the Usermap (The Usermap substitutes when the J-field is not yet found.)<= o:p>
H &=
nbsp; Show
all fields high intensity (employs a square root function)
S &=
nbsp; Stop
solving =
&nb=
sp;
+ or =3D &=
nbsp; Zoom
in  =
; &n=
bsp;  =
; &n=
bsp;  =
;
- &=
nbsp; &nbs=
p; Zoom
out
] or PgUp  =
; Rescale
larger
[ or PgDn  =
; Rescale
smaller
> or Insert
< or Delete
Enter &=
nbsp; redraw
plot
&J &=
nbsp; Write
the entire current map to a text file
&V &=
nbsp; Write
the entire voltage map to a text file
=
&nb=
sp; (The &L, &T, &K, and &a=
mp;G
commands are undocumented. Th=
ey are
not of general interest.)
Color definitions=
can
be modified using the keyboard. If
the “Show all standard colors” box is checked, the changes are =
to
the internal color table and persist until the program terminates. Otherwise the changes persist only=
until
the next Usermap is loaded.
Mor=
eColors.txt
This file does not have to exist. Any text editor can be used to cre=
ate
it.
A sample MoreColors.txt<=
/b> file:
|red: green: blue: use: Ohms: E=
r: tanDelta: Mu:
name:
250
20 &nb=
sp;
20
+1 &nb=
sp;
7.3 =
1 0
30 255 30 0
20.6 1 0
180
180
120
insul 0
30 &=
nbsp;
30 &nb=
sp;
30
insul 0
100
140
170
insul 0
Each line in the file declares and defines =
one
color. All parameters must ap=
pear
in order. The parameters are:=
1. Red value. (0 to 255)
2. Green value. (0 to 255)
3. Blue value. (0 to 255)
4. Function. (must be +1, -1, 0, float, or insu=
l)
5. Resistivity in
Ohm-Centimeters. (ignored for
insulators)
6. Relative permitti=
vity
(ignored for conductors)
7. Loss tangent
tanδ. (dielectric loss,
ignored for conductors)
8. Relative
permeability. (for future use=
)
9. Material name.
If a color duplicates an existing color, the
new definition prevails. Atlc2
loads Mor=
eColors.txt only once, when =
it
begins. Then the user can use=
the
keyboard to further redefine colors.
The program will hold up to 500 colors, but only 256 can appear in a
Usermap.
All parameters in this file are delimited by
blanks. A tab is treated as a
blank. Consecutive blanks are
treated as one blank. A
“|” character denotes a comment, and all further characters on =
the
line are ignored. Blank lines=
are
allowed.
Imp=
lementation
details
Atlc2 is written in Delphi XE10 Pascal (abo=
ut
8000 lines). It uses 64-bit
floating point for all calculations.
Atlc2 uses multiple threads (cores) during E
field prediction and relaxation (C=
and
Gp), and equation solving (L a=
nd Rs). It
assumes all of the computer’s processors are idle. The program can manage up to 32
threads. One "master&quo=
t;
thread parcels out work for the others, which sit in infinite loops when th=
ey
have nothing to do. So the Ta=
sk
Manager graphs do not necessarily show actual work being done. Any interruption that slows down a=
ny
thread slows down them all. Y=
ou
should reduce the atlc2 thread count if your system does a lot of other
concurrent work. On the autho=
r's dual-Xeon workstation, running 12 threads is about 6 t=
imes
as fast as a single thread with the other 11 cores idle. L a=
nd Rs This part of the program works by solving
equations, one equation per conductor pixel. The equations force the total net
current to equal zero. But th=
ere
are transmission lines in which this is not true. Some geometries (including coaxial
cable) encourage an unbalanced current, which produces an excess charge tha=
t is
balanced by a virtual charge infinitely far away. Such lines will radiate radio wave=
s at
all frequencies if nothing is done about this. Grounding the offending conductor =
might
be sufficient to fix this. A =
driver
circuit that implements VO will certainly fix it. If there is a grounded third wire (green) t=
hen
“Ignd” gives the
ground current compared to the red current. Analytically this number is not ve=
ry
useful. It is provided as a
convenience. If I=
gnd=
is below 4% then=
it
might be safely ignored. Othe=
rwise
maybe your whole approach should be reconsidered. Infinite plane extensions are ignored durin=
g L and Rs. Only the conductor pixels in the U=
sermap
contribute toward the solution. The math used in L and Rs is described at LandRs.html. Atlc2 correctly models low frequen=
cy
dispersion.  =
; &n=
bsp; Low
frequency dispersion  =
; =
C a=
nd Gp A frequency and pixel size are not required
when solving for C and Gp. The
method used is the same as in atlc.
So C is a D.C. surface capacitance.=
The A.C. capacitance is generally the same at all frequencies. C a=
nd Gp can be thought o=
f as a
soap-film-stretching program.
Imagine high structures (high voltage) and low structures (low volta=
ge)
with a fabric or soapy film stretched between them. The white V-lines (often circles) =
are
elevation lines, like those on a topographical map. The method is a “relaxation
algorithm” that continuously sets the voltage at each pixel to the
average of its four closest neighbors.&nbs=
p;
This will eventually stabilize and will be a perfect solution to
Gauss’s Law. How long it
takes depends on how the program initializes the array. The better the program can predict=
the
final field, the quicker the relaxation algorithm will finish. Prediction uses the formula <=
/sub>=
after distributing the surface char=
ge. Surface charge distribution is by successive
approximations. The E field a=
t each
surface pixel is found by summing the contributions from the charges at all=
the
pixels. The field component n=
ormal
to the surface predicts a new value for the charge at that pixel. About 20 iterations seems to work
well. Presently the prediction cannot guess the
effects of dielectrics. So ex=
amples
with dielectrics take longer for an accurate result. Most other examples finish quickly=
. Another slow example is any transm=
ission
line with a net charge. This =
will
be the case if you use green instead of blue for unshielded lines. Such lines will radiate, and the
reported Z0 is not a true characteristic impedance. Almost certainly you are making a
mistake. If you want to desig=
n an
antenna, you need a very different kind of program. For unshielded lines, the Usermap is extend=
ed
outward by adding pixels. The=
se
pixels are of two different sizes.
Usually the Usermap is extended outward by 100 pixels that are the s=
ame
size as the Usermap pixels. T=
hen
the map is further extended using pixels that are 8 times larger (256 time
larger in area), until the size of the map is equivalent to 3200 x 3200 of =
the
Usermap pixels. Note that if =
you
surround the conductors with a lot of empty space (rather than use a smaller
Usermap) then more of the smaller pixels are employed, slowing down the run
with probably no improvement in the accuracy. (The region with smaller pixels is=
never
smaller than the Usermap.) L a=
nd Rs versus C and Gp Atlc2 is essentially two programs bundled t=
ogether. They do roughly the same thing by =
two
different methods. Sometimes =
you
need to run both, but for other cases, one is sufficient. L
and Rs tends to be more accurate, but its execution time rises with the
cube of the conductor pixel total, which can be prohibitive. L
and Rs cannot handle multiple dielectrics. C
and Gp loses accuracy when the conductors are close together. (Close conductors will spike the
capacitance, which a coarse grid portrays poorly, but the inductance remains
well behaved.) C a=
nd Gp is itself two pr=
ograms
bundled together: A charge-sh=
ifting
program and a relaxation program.
Either could do the whole job.
The former is usually faster, but the latter is more accurate. So C
and Gp employs both. In c=
ases
where charge-shifting is slower, you might want to skip it, which is done by
checking the box “Skip E prediction”. Execution time for prediction rise=
s with
the cube of the model size, while relaxation is roughly the square. If you must use a map too large for L and Rs, try this trick: Run C
and Gp a second time but with all permittivities=
span>
set to 1, and use the L that it reports.&n=
bsp;
Z0 =3D sqrt(L/C) A
discussion of 3-wire transmission lines Multi-wire transmission lines have multiple
characteristic impedances, none of which truly correspond to that of a 2-wi=
re
transmission line. Any 3-terminal linear network can be modele=
d as
3 impedances in either a “Y” or “delta”
configuration. We shall use a
“Y” to describe a 3-wire line as seen from the end. If the green wire is made to float, it will
carry no current. Upon running
atlc2, let’s designate the resulting impedance Z=
oGCZ
(green current zero). T=
wo
more runs of atlc2 will give us ZoRCZ and ZoBCZ.&nb=
sp;
It is evident that: Z=
oRCZ
=3D ZoB + Zo These can be solved for ZoR,
ZoG, and ZoB: ZoR=
=3D (ZoGCZ + Zo=
BCZ
– ZoRCZ)/2 ZoG
=3D (ZoRCZ + Zo=
BCZ
– ZoGCZ)/2 ZoB
=3D (ZoRCZ + Zo=
GCZ
– ZoBCZ)/2 ___________________________________________=
_____________________________ A quarter-wave directional coupler is a com=
mon
3-wire transmission line: If the green wire is made to float
then there is no green current.
Whatever float voltage atlc2 reports will be equal to the center
voltage, VC. Thus:=
Z=
oB
=3D ZoGCZ * (Vfloat+1) / =
2 ZoR=
=3D ZoGCZ – ZoB ZoG =3D ZoR ZoODD =3D 2 * Zo=
R ZoEVEN =3D ZoR=
sub>
/ 2 + ZoB So it is possible to get a complete
characterization with a single run of atlc2. In theory this works. In practice, the results are often=
off
by more than 5%. The problem =
is
that the asymmetry in the excitation results in an unbalanced charge, which
makes ZoGCZ inappropriate for determ=
ining ZoODD.&nb=
sp;
You could derive C from the L and Rs run. But it is probably best to make tw=
o runs
that find ZoODD and ZoEVEN
directly. Scr=
ipting
Facility Although I added this feature for my own us=
e,
you might find it useful. If you enter ZZZ in the Name box =
and
then select File>Run Script,
atlc2 will open the file When the atlc2 application begins, it looks=
for
the file StartupScript.txt. If found, atlc2 will execute it. A sample script file: | A sample script file twinlead box 1 F |
restrict to skin depth total 4000 separation 6mm diameter 2mm frequency .001 sweep frequency 25 4 The file format rules are the same as for <=
/span>MoreColors.txt.
Only the first 3 characters of each command name are examined. When the argument is a directory o=
r file
name, the case and any blanks are preserved. Otherwise the case does not matter=
but
there must be no blanks within an argument. Most commands are simple string
moves. Very little error chec=
king
is performed. A detected erro=
r will
abort the script. The
"Stop" button will abort the script. No files are kept open during solv=
es or
script execution. All=
owed
commands: ZZZ.txt (arg1, arg2, and=
arg3
are the command arguments)
1 &nbs=
p; Restrict
to skin depth 2 &nbs=
p; Unending
run 3 &nbs=
p; Skip
E prediction 4 &nbs=
p; Vo
zero 10 &nb=
sp; Add
top plane 11 &nb=
sp; Add
bottom plane 12 &nb=
sp; Add
side planes 13 &nb=
sp; Insulation
vertical 14 &nb=
sp; Insulation
horizontal 15 &nb=
sp; One
wire 20 &nb=
sp; Add
floating wire 21 &nb=
sp; Add
grounded wire 30 &nb=
sp; Air
dielectric 31 &nb=
sp; Show
all standard colors