OPTIMISE(1)
NAME
optimise - Yagi-Uda project antenna optimiser
SYNOPSIS
optimise [ -dhvwO ] [ -aangular_stepsize ] [ -bboom_extension ] [ -ccleanliness_of_pattern ] [ -eelements ] [ -fFBratio ] [ -gGA_optimi- sation_method ] -lpercent ] [ -mmin_offset_from_peak ] [ -ooptimisa- tion_criteria ] [ -ppopulation ] [ -rresistance ] [ -sswr ] [ -tlength_tolerance ] [ -xreactance ] [ -AAuto_gain ] [ -CCurrents_simi- lar ] [ -Fweight_FB ] [ -Gweight_gain ] [ -Kkeep_for_tries ] [ -Pweight_pattern_cleanliness ] [ -Rweight_resistance ] [ -Sweight_swr ] [ -Tposition_tolerance ] [ -WWeighted_algorithm ] [ -Xweight_reactance [ -ZZo ] filename iterations
DESCRIPTION
The program optimise is one of a number of executable programs that forms part of a set of programs, collectively known as the Yagi-Uda project , which were designed for analysis and optimisation of Yagi-Uda antennas. optimise attempts to optimise the performance of a Yagi antenna for one or more parameters that are considered important, such as gain, F/B ratio, VSWR etc. It does this by randomly changing the lengths and positions, of one or more elements, then comparing the performance before and after the change. Any improvements are written to a new file called filename.bes where filename is the name of the antenna description file created by input or first
When Yagi's are designed on paper, or using this program, its possible
that they will be almost impossible to build, if their performance
depends too critically on the dimensions. To determine if this is the
case with a design, we run optimise with just the options 't' and 'T'.
These specify the tolerance with which you can build the antenna,
expressed as a standard deviation in mm. In this case, instead of trying to optimise a poor design, optimise will calculate the minimum
gain, maximum VSWR, and minimum FB ratio of a number of designs, all
slightly different from the input file. 99.7% of the components lie
within 3 SD of the mean, so if you think you can cut elements to with 1
mm 99.7% of the time, specify t0.33. If you can put them in the boom to
within 3 mm 99.7% of the time, specify T1.
If while optimise is running using the methods that require weights to
be attached to the gain, FB, SWR etc, it becomes apparent, the weights
are not optimum, its possible to pause the program and re-adjust the
weights. If a file with the name of changes is created, the program
will pause, then request new weights are entered at the keyboard.
AVAILABILITY
OPTIONS
- -d Print the default values of all the configureable parameters to
- stdout. Typing this option with any option that changes a parameter (see below) will display the new value of the parameter, rather than the default.
- -h Print a help message.
- -v Print verbose status information.
- -w Instead of optimising at one fixed frequency (the design fre
- quency), this directs the program to optimise at 3 separate frequencies (lowest, design and highest) then to average data at
all 3. This option is better for wideband antenna. Note that the
input impedance printed is at the design frequency, *not* averaged over 3 frequencies. Averaging an impedance, is likely to
give a very misleading impression. The impedance averaged over 3
frequencies can be 50+i0 Ohms, even if the VSWR is very poor
over all 3 frequencies, as the following 3 pieces of data show.
Z=147 + j 300 SWR= 15.46:1
Z=2 + j 100 SWR= 125:1
Z=1 - j 400 SWR= 3250:1
note in the above three cases, the average impedance is 50 + j 0, but average SWR is 1130:1. - -O Over-optimisation allowed. By default, the program does not
- over-optimise a parameter. For example, an SWR of 1.01 is usually considered good enough and any change, as long as the SWR stayed good, typically below 1.1:1, would be allowed, even if the SWR rose. By default, FB's of 27 dB, VSWR's of 1.1 are acceptable. However, by using the -O option, you can insist the program always improves things, no matter how good they are.
- -aAngular_stepsize
- When optimimising by trying to get a clean pattern, specifies the step size to use when looking for features in
the pattern. If its set too small, the program runs
slow. If its set too large, the program may miss features
in the pattern, such as a sidelobe. Then the resulting
antenna will have poor sidelobe performance, even though
you think it will be good. The program attempts to calculate a sensible value, based on 1/10th the approximate 3
dB beamwidth, if you don't set. - -bboom_extension
- Generally speaking, the gain of a Yagi increases with
boom length. Hence the optimiser would often give you a
Yagi with a much longer boom than the input file. This
may not be what you desire due to space restrictions.
These long antennas often have high gain, but are very
narrow in bandwidth. The default limits the antenna to
10x the original length, which means effectively there is
no boom length limitation. You can adjust the percentage
by setting boom_extension to whatever you wish. -b30 will limit the boom to no more than 30% more than the
original length. - -ccleanliness_of_pattern
- Specify the number of dB down on the peak gain to aim to
get the pattern. Any antenna pattern cleaner than this
will not effect the fitness, nor will it be considered
any better when comparing to antenna designs. 20 dB seems
reasonable, so the default is 20, but this may of course
change if it's deceided too. Check the source code to be
certain (see REASONABLE_SIDELOBE in yagi.h). - -eelements
- is an integer which specifies the type of elements that
are changed in the optimisation cycle. Possible values
are:1 - alter only the driven element(s) length (useful to
bring to resonance)
2 - alter only the driven element position. Don't change
its length.
4 - alter only the reflector length. The position is
always at x=0.
8 - alter only the director lengths. Don't change positions.
16 - alter only the director positions. Don't change
lengths.
32 - randomly adjust one element length, then makes all
other the same. Don't change the positions.
64 - apply a linear taper to the lengths.
128 - Set the driven element to a resonate length. It
may/may-not be altered after the first run, depending on
the whether or not '1' is invoked too. Eg -e128 will
make it resonate and keep it there forever. However
'-e129' will bring to resonance, then alter to maximuse
performance. - The elements altered is made from a logical AND of the
above, so for example to alter everything, except the
driven element length, use -e30, since 2+4+8+16=30. The
default is equivalent to -e31 , which changes everything possible. Note the reflector position is *never* changed.
It's always at x=0. - -fFBratio
- When optimising an antenna, consider any FB ratio
greater than FBratio dB to be equal to FBratio dB. This avoids optimising to a very high FB ratio,
which is impracticable, as the bandwidth over
which this FB ratio will be maintained is very
small and mechanical considerations will prevent
you from constructing it with such a high FB ratio
anyway. If this was not prevented, you might just
happen to get an antenna with 100 dB FB ratio, but
poor gain and swr. Since by default all parameters
must improve, the optimisation routine will most
likely never being able to improve on the 100 dB
FB ratio, so no improvement will result. Most people would prefer to get a few extra dB of gain,
even if the FB ratio dropped to 30 dB. - -gGA_optimisation_method
- Use a genetic algorithm. With the genetic algorithm, the program does not take any account any
of the initial lengths/positions of elements specified in the input file. Rather it works by initialising a number of different antenna, then computing a 'fitness' value for each. The fitness
value can depend on the gain, FB, real part of the
input impedance, reactive part of the input impedance, VSWR or the level of the sidelobes. The
integer after the g tells the optimiser what to
consider. -g1 Use gain
-g2 Use FB
-g4 Use R
-g8 Use X
-g16 Use the SWR
-g32 Use the level of the sidelobes. - You can use a logical AND of these, so for example
-g49 will use a genetic algorithm, optimising for
gain, swr and sidelobe level, since
1(gain)+16(SWR)+32(sidelobe level)=49. - -lpercent
- is a parameter (floating point number) which specifies the maximum percentage change in the positions or lengths of an elements at each iteration.
If the option is not used, it will be set internally at 10% for the first 25% of the iterations,
1% for the next 25%, 0.1% for the third 25% of the
iterations and 0.01% for the last 25% of the iterations. If set to a positive number x (eg optimise
-l 0.3 145e10) then the percentage will be set at
x% for 25% of iterations, x/10 for the next 25%,
x/100 for the next 25 and x/1000 for the last 25%.
If set to a negative number y (eg optimise -l -0.5
145e10) then the paramters will stay fixed at y%
(in this example 0.5%) all the time. - -mmin_offset-from_peak
- Sets the minimum angle in degrees offset from
theta=90 degrees, where the side lobes start and
the main lobe finishes. The higher the gain, the
smaller it should be. It is set internally if not
set on the command line. - -ooptimisation_criteria
- 1 - Assume better if the gain has increased.
2 - Assume better if the front to back ratio has
improved.
4 - Assume better if the real part of the input
impedance is closer to the value that the program
was compiled for, or set using the '-Z' option.
This will usually be 50 Ohms, but you may wish to
set this to 12.5 Ohms if you use a 4:1 balun.
Generally you can get higher gain from a Yagi if
you allow the input impedance to fall, but of
course feeding it becomes more difficult.
8 - Assume better if the magnitude of the reactive component of the input impedance is lower
(ie. the antenna is nearer resonance).
16 - Assume better if the VSWR is lower.
32 - Assume better if the level of all sidelobes
is lower. - The optimisation_criteria may be formed from a
logical AND of these numbers, so for example
choosing -o19 will only consider a revised antenna
better than the previous, if the SWR, gain and F/B
ratio have all simultaneously improved. - Clearly an antenna which originally had 12 dB gain
and 1.01:1 VSWR but then changes to 20 dB gain @
1.02:1 VSWR, would to most people be better, even
though the VSWR has increased. By default, optimise only optimises to sensible maximums, so to
not let the optimisation stall prematurely. By
running optimise with no arguments, the program will list the limits of acceptability. These
might be typically F/B ratio > 27 dB, VSWR <
1.1:1, magnitude of input reactance less than 5
Ohms and the real part of the input impedance
within 5 Ohms of Zo. Choosing -o19 (1+2+16=19)
will optimise for gain (since G=1), FB (since
FB=2) and SWR (Since SWR=16), but would consider a
higher gain and FB ratio antenna better than a
previous one, even if the SWR rose, as long as it
stayed below 1.1:1 (or as was set during compilation). The default behaviour (no options) is
equivalent to -o37 which optimiseas for gain(1),
the real part of the input impedance(4) and sidelobes(32) but this may be changed at any time, so
type optimise -d to check the current settings. If you insist on the program optimisang for the very
best of all selected parameters, use the -O option
too, but be warned the optimisation will probely
stick once it gets one parameter really good. - -ppopulation
- This determines the initial population used with
the genetic algorithm. - -rresistance
- When optimising an antenna, consider any input
resistance closer to Zo (usually 50 Ohms) than
resistance Ohms to be acceptable. This avoids optimising to an input resistance too close to Zo,
which is impracticable, as the bandwidth over
which the input resistance could be maintained is
very small and mechanical considerations will prevent you from constructing the antenna with such
an ideal input resistance. If this was not prevented, you might just happen to get an antenna
with an input resistance of 50.000001 Ohms, but
poor gain, FB and possibly even a poor swr, if the
antenna is well away from resonance. Since by
default all parameters must improve, the optimisation routine will get most likely never being able
to improve on the antenna, whereas we might be
happier with a few more dB gain, if the input
resistance went to 50.1 Ohms. It should be noted
that the default optimisation routine never uses
the input resistance directly (only VSWR), so this
option cant be used without the '-o' option to
optimise for other than the default parameters
(gain, VSWR and FB ratio). - -sswr When optimising an antenna, consider any SWR less
- than swr to be equal to swr This avoids optimising
to a very low swr, which is impracticable, as the
bandwidth over which such a low swr could be maintained would be very small and mechanical considerations will prevent you from constructing such
an antenna anyway. If this is was not prevented,
you might just happen to get an antenna with an
swr of 1.000000000001:1, but poor gain, FB ratio.
Since by default all parameters must improve, the
optimisation routine will most likely never being
able to improve on the antenna, even though in
practice you would like to get a few extra dB of
gain if the SWR would rise to 1.02:1. The default
was equivalent to -s1.1 but run optimise -d to display this and any other defaults. - -tlength_tolerance
- length_tolerance is the standard deviation in mm
of the accuracy with which you can cut elements.
Since 99.7% of elements will be with 3 standard
deviations of the mean length (stats theory says
this), set -t0.2 if virtually all (well 99.7%) of
elements are within 3x0.2=0.6 mm of the correct
length. This option *must* be used with the '-T'
option and can't be used with any other options
apart from '-Z', '-v' and '-d'. - -xreactance
- When optimising an antenna, consider any input
reactance of less than reactance to be reactance. This avoids over optimising the reactance, at the
expense of something else. - -Aauto_gain
- When the auto_gain option is used. the program
maximes the gain of the antenna (ignoring all
other parameters such as SWR, FB ratio etc) by
adjusting the length (not position) of one element
only. -A-1 will maximuse the gain, by adjusting
the length of the reflector, -A0 will maximise the
gain by adjusting the length of the driven element. Its generally *not* a good idea to maximise
the gain by adjusting the driven element, but the
program lets you do it, but using the option -A0.
Using -A1 will maximise gain by adjusting the
length of the first director, -A2 the second
director and so on, up to the last director. You
must check carefully that the input impedance in
particular does not fall to silly values if you
use this option. On a yagi with many elements (>
10 or so), you can pretty safely maximise the 8th
or more director, but doing it on the reflector,
driven element or early directors often leads to
silly input impedances - so beware! Note, no matter how many iterations you specify, this process
is only done once.Its unlikely you will be able to
do it again, without things going out of hand, but
if you must do it, you must re-run 'optimise'
again. - -Ccurrents_similar
- If this option is used, where currents_similar is
an integer, the program looks to make the currents
in the last currents_similar elements as similar as possible. It computes the sum of the squares of
the deviations of the absolute values of the element currents from the mean. If this falls, and
the criteria specified with the -W option is also
satisfied, the antenna is considered better. If
currents_similar is three less than the number of directors, it tries to make the currents in the
the directors (but ignoringing the first 3) all
similar. If currents_similar is equal to the number of directors, it tries to make all the directors have similar currents. If currents_similar is one more than the number of directors, it tries to
make all the directors and the reflector have similar currents. If currents_similar is equal to the total number of elements, then it fails with an
error message. - -Fweight_FB
- is the floating point number (default 1.0) specifying the weight to attach to the FB ratio of the
antenna when using the '-W' option, which calculates a fitness for the antenna based on one or
more parameters (FB, gain, input resistance, input
reactance, SWR, cleanliness of antenna pattern).
The '-F' option is similar to the options -G, -P,
-R, -S, -X (which specify weights for gain, pattern cleanliness, input resistance, SWR and input
reactance). When using the -W option the exact
algorithm used to compute the fitness (and hence
the effect of this parameter) is best checked by
looking at the source code (see perform.c). This
is one area of constant program improvement/changes/development, so its difficult to say
exactly the effect the parameter has. However,
increasing the weight of a parameter (using the
-F, -G, -R, -S or -X options) will make the associated parameter have a greater effect on the fitness. However, unless you optimise for a high FB
ratio with the -W option, then setting the -F
option will have no effect. For example, setting
the options -F2.5 -W1 is a complete waste of time.
There you have used the -W1 option to optimise
only for gain (see -W option section of man page)
but have changed the weight of the FB ratio from
its default 1.0 to 2.5. If you are not optimising
for FB ratio, the weight you attach to it is
irrelavent. - -Gweight_gain
- is the floating point number (default 1.0) specifying the weight to attach to the gain of the
antenna when using the '-W' option, which calculates a fitness for the antenna based on one or
more parameters (FB, gain, input resistance, input
reactance, SWR, cleanliness of antenna pattern).
The '-G' option is similar to the options -F, -P,
-R, -S, -X (which specify weights for FB ratio,
pattern cleanliness, input resistance, SWR and
input reactance). When using the -W option the
exact algorithm used to compute the fitness (and
hence the effect of this parameter) is best
checked by looking at the source code (see perform.c). This is one area of constant program
improvement/changes/development, so its difficult
to say exactly the effect the parameter has. However, increasing the weight of a parameter (using
the -F, -G, -R, -S or -X options) will make the
associated parameter have a greater effect on the
fitness. However, unless you optimise for gain
with the -W option, then setting the -G option
will have no effect. For example, setting the
options -G2.5 -W2 is a complete waste of time.
There you have used the -W2 option to optimise
only for FB ratio (see -W option section of man
page) but have changed the weight of the gain from
its default 1.0 to 2.5. If you are not optimising
for gain, the weight you attach to it is
irrelavent. - -Kkeep_for_tries
- keep_for_tries is the number of tries for the
optimise to persist using the original data file
as the starting point for optimisation. By default
it is 1, which means the program immediately looks
from a new position once a better one is found. It
is theeoretically possible that this might result
in a quick, but poor local maximum. If however,
keep_for_tries is 1000, it will stay at a position for 1000 iterations after finding the last best
result, before considering this to be a global
optimum. Then it starts for the new position. In
practice, I have found this option to make matters
worst in most cases. It was added to avoid the
local-minimum problem, but it appears the optimisation surface is pretty smooth, so it just slows
the program, without gaining much. Anyway, it can
stay as an option, but check the results
with/without carefully before using extensively. - -Ppattern_cleanlyiness
- is the floating point number (default 1.0) specifying the weight to attach to the cleanness of the
antenna pattern when using the '-W' option, which
calculates a fitness for the antenna based on one
or more parameters (FB, gain, input resistance,
input reactance, SWR, cleanliness of antenna pattern). The '-P' option is similar to the options
-F, -G, -R, -S, -X (which specify weights for FB
ratio, gain, input resistance, SWR and input reactance). When using the -W option the exact algorithm used to compute the fitness (and hence the
effect of this parameter) is best checked by looking at the source code (see perform.c). This is
one area of constant program improvement/changes/development, so its difficult to say
exactly the effect the parameter has. However,
increasing the weight of a parameter (using the
-F, -G, -R, -S or -X options) will make the associated parameter have a greater effect on the fitness. However, unless you optimise for a clean
antenna pattern with the -W option, then setting
the -P option will have no effect. For example,
setting the options -P2.5 -W1 is a complete waste
of time. There you have used the -W1 option to
optimise only for gain (see -W option section of
man page) but have changed the weight of the pattern cleanliness from its default 1.0 to 2.5. If
you are not optimising for a clean radiation pattern, the weight you attach to it is irrelavent.
With appropiate use of the -W option (eg -W49 for
gain, SWR and a clean pattern), the computer program finds the level of the most significant sidelobe, wherever it may be outside the main bean. It
then optimises to reduce this. The -P option tells
it how much weight to put on reducing this sidelobe. - -Rweight_resistance
- is the floating point number (default 1.0) specifying the weight to attach to the obtaining an
input resistance close to Zo on the antenna when
using the '-W' option, which calculates a fitness
for the antenna based on one or more parameters
(FB, gain, input resistance, input reactance, SWR,
cleanliness of antenna pattern). The '-R' option
is similar to the options -F, -G, -P, -S, -X
(which specify weights for FB, gain, pattern
cleanliness, SWR and input reactance). When using
the -W option the exact algorithm used to compute
the fitness (and hence the effect of this parameter) is best checked by looking at the source code
(see perform.c). This is one area of constant program improvement/changes/development, so its difficult to say exactly the effect the parameter
has. However, increasing the weight of a parameter
(using the -F, -G, -R, -S or -X options) will make
the associated parameter have a greater effect on
the fitness. However, unless you optimise for an
an input resistance close to Zo, with the -W
option, then setting the -R option will have no
effect. For example, setting the options -R2.5 -W1
is a complete waste of time. There you have used
the -W1 option to optimise only for gain (see -W
option section of man page) but have changed the
weight of the resistance from its default 1.0 to
2.5. If you are not optimising for an input resistance close to Zo, the weight you attach to it is
irrelavent. - -Sweight_swr
- is the floating point number (default 1.0) specifying the weight to attach to the SWR of the
antenna when using the '-W' option, which calculates a fitness for the antenna based on one or
more parameters (FB, gain, input resistance, input
reactance, SWR, cleanliness of antenna pattern).
The '-S' option is similar to the options -F, -G,
-P, -R, -X (which specify weights for FB, gain,
pattern cleanliness, input resistance and input
reactance). When using the -W option the exact
algorithm used to compute the fitness (and hence
the effect of this parameter) is best checked by
looking at the source code (see perform.c). This
is one area of constant program improvement/changes/development, so its difficult to say
exactly the effect the parameter has. However,
increasing the weight of a parameter (using the
-F, -G, -R, -S or -X options) will make the associated parameter have a greater effect on the fitness. However, unless you optimise for SWR with
the -W option, then setting the -S option will
have no effect. For example, setting the options
-S2.5 -W1 is a complete waste of time. There you
have used the -W1 option to optimise only for gain
(see -W option section of man page) but have
changed the weight of the SWR from its default 1.0
to 2.5. If you are not optimising for SWR, the
weight you attach to it is irrelavent. - -Tposition_tolerance
- position_tolerance is the standard deviation in mm
of the accuracy with which you can cut elements.
Since 99.7% of elements will be with 3 standard
deviations of the correct position (stats theory
says this), set -T2 if virtually all (well 99.7%)
of elements are within 3x2=6 mm of the correct
position.This option *must* be used with the '-t'
option and can't be used with any other options
apart from '-Z', '-v' and '-d'. - -WWeighted_algorithm
- Try to get an antenna which is better according to
a weighted combination of parameters, rather than
require them all to improve. The integer specifies
what to consider in the weighted parameters.
W1 Gain.
W2 FB
W4 R
W8 X
W16 SWR
W32 SIDE_LOBE
You can logically AND these together, so for example -W3 will optimise using a weighted combination
of gain and FB. -W49, will use a weighted combination of gain, swr and sidelobe leve, since
32+16+1=49. - -Xweight_reactance
- is the floating point number (default 1.0) specifying the weight to attach to achieving a low
input reactance on the antenna when using the '-W'
option, which calculates a fitness for the antenna
based on one or more parameters (FB, gain, input
resistance, input reactance, SWR, cleanliness of
antenna pattern). The '-X' option is similar to
the options -F, G, -P, -R and -S (which specify
weights for FB ratio, gain, pattern cleanliness,
input resistance, and SWR). When using the -W
option the exact algorithm used to compute the
fitness (and hence the effect of this parameter)
is best checked by looking at the source code (see
perform.c). This is one area of constant program
improvement/changes/development, so its difficult
to say exactly the effect the parameter has. However, increasing the weight of a parameter (using
the -F, -G, -R, -S or -X options) will make the
associated parameter have a greater effect on the
fitness. However, unless you optimise for a low
input reactance with the -W option, then setting
the -X option will have no effect. For example,
setting the options -X2.5 -W1 is a complete waste
of time. There you have used the -W1 option to
optimise only for gain (see -W option section of
man page) but have changed the weight of the reactiance from its default 1.0 to 2.5. If you are not
optimising for a low input reactance, the weight
you attach to it is irrelavent. - -ZZo
- Zo is the characteristic impedance used when evaluating the VSWR, reflection coefficient and other
similar calculations. The optimiser usually tries
to bring the input impedance of the antenna to
this value. It is set by default to 50 Ohms, so
the default is equivalent to -Z50 but may be set
to any positive number. Set to 12.5 Ohms if you
are going to feed the antenna with a 4:1 balun.
Generally speaking, the gain of a Yagi can be
higher for low input impedances, but of course
such antennas are more difficult to feed. - filename
- This is the name of the file containing the
antenna description. It is expected to be in a
format created by either input or first - two other programs in the Yagi-Uda project. This is an ASCII text file. - iterations
- is an integer specifying the number of iterations
for the optimiser to perform to try to get the
best antenna. Time will limit the number you
choose. 1000 iterations of a 1ele yagi takes about
5 seconds, a 6ele approximately 60 seconds, an 11
element 350 seconds, a 20 element 1030 seconds, a
33ele 2440 seconds, a 50element 5400 seconds,
100ele 21320 seconds all on an old 25MHz 486 PC
with no external cache. When using the -A option
the iterations is automatically set internally so only one attempt is made. When using the '-t' and
'-T' options, iterations specifies the number of iterations to attempt to get a poorer design, to
check the sensitivity of the design to small manufacturing tolerances.
EXAMPLES
Here are a number of examples of using optimise.
1) optimise 5ele 1000
Here the file 5ele will be optimised using the default
system for 1000 iterations. The default might typically
require gain, FB and SWR to all improve, but this may be
changed at any time. In any case, the program tells you
what its optimising for. By default the program will only
optimise to the selected parameters are good, not overoptimising any one at the detrement of the others.
2) optimise -b30 -f50 -s2 5ele 1000
This is similar to above, but the boom can not extend by
more than 30% from its original length, FB ratios above
50 dB are considered acceptable, as are SWR's less than
2:1. The optimised resultant antenna is likely to have
better FB ratio, but poorer SWR than in (1) above.
3) optimise -o1 5ele 1000
This will simply optimise 5ele for maximum forward gain.
The resultant antenna may have a poor FB ratio and is
likely to have an unacceptably low input impedance and
hence high VSWR. This is not a very sensible method of
optimisation.
4) optimise -W49 -l7 5ele 10000
This will optimise the file 5ele using for 10000 iterations. It will require that the weighted performance of
the antenna in three important parameters (gain, sidelobe
level and SWR) improves from one design to the next. One
or two parameters can actually get worst from one design
to the next, but the weighted performance is better. The
positions of the elements or lengths of elements will not
change by more than 7% in each iteration.
5) optimise -g -S30 -G50 -F20 -p1500 5ele 10000
This will optimise the file 5ele using a genetic algorithm. 1500 antennas will be randomly designed. The performance of each of these will measured using a 'fitness'
function, weighted 30% to SWR, 50% to gain and 20% to FB
ratio. The probability of breading from a pair of antennas is proportional to the fitness function.
6) optimise -w atv_antenna 10000
This will optimise the file atv_antenna for a best average performance over a wide band. The progrram calculates
the gain, FB and SWR at three frequencies, then computes
an average (mean) performance of the antenna over the
band. N iterations will take 3x as long to execute as N
iterations on the same antenna without the '-w' option.
7) optimise -t0.1 -T1 good_design 100
This will take the file good_design and make 100 different antennas from it, to simulate the effects of building
tolerances. Each element is assumed to be cut so that the
mean error of all elements is 0 mm, but a standard deviation of 0.1 mm, so 68.4% of element lengths are within
0.1 mm, 95.4% within 0.2 mm and 99.7% with in 0.3 mm. The
accuracy of placing elements along the boom is much
lower, so here we have specified a standard deviation of
1.0 mm, so 68.6% of elements are placed within 1 mm of
the correct position, 95.4% within 2 mm of the correct
position etc. The program will report the *worst* performances achieved. If the performance dips too mush,
then you either need to build them better, or get a
design that's less critical!
STOPPING
Optimise will stop after the number of iterations specified in the parameter iterations. It will also stop if a
file stop exits in the current directory of the executable optimise This file can of course only be created
using a multi-tasking operating system such as Unix. It
is *not* advisable to stop the program by hitting the DEL
key (Unix) or CONTROL-C (DOS), as one of the files may be
open at the time, resulting in an empty file. Files are
not open for any longer than necessary (they are closed
immediately after writing to them), so this is not a
likely occurrence, but can still occur.
LIMITATIONS
I'm not aware of any limitations, apart from that filenames, including full path, can't exceed 90 characters.
FILES
filename Antenna description, created by input or first.
filename.up Update file, listing achievements of optimise.
filename.bes Best file, containing the best design to date.
changes File that causes the program to pause to re-adjust weights.
stop File that stops optimisation process.
SEE ALSO
first(1), input(1), output(1), yagi(1), first(5),
input(5) output(5) and optimise(5).
PLATFORMS
Both DOS and Unix versions have been built. The DOS version as distributed requires a 386 PC with a 387 maths
coprocessor.
Although I have altered the source to make it more compatible with DOS (reduced file name lengths etc), my wish
is to build a decent program, rather than fit the program
to an outdated operating system. If there is a *good*
reason to use code that is incompatible with DOS, this
will be done.
Since optimise takes a while to optimise an antenna
(I've optimised one design for a week), it is obviously
more sensible to build this program under a multi-tasking
operating system, as otherwise a PC can be tied up for
days.
BUGS
Bugs should be reported to david.kirkby@onetel.net. Bugs
tend actually to be fixed if they can be isolated, so it
is in your interest to report them in such a way that
they can be easily reproduced.
The program will dump core (crash) if asked to optimise a
1ele beam, without any arguments. This is because a 1ele
beam has no parasitic elements and by default the program
only changes parasitic elements.
Some of the options are not checked for sensible values,
although most are now checked and report if they are out
of range.
If the user specifies very large manufacturing errors
using the '-t' and '-T' options, its possible for elements to overlap or for element lengths to become negative. This will cause numerical errors. Any reasonable
values will not cause this.
On long Yagi's (50 elements) optimise can go a bit silly.
It can optimise say a 1296MHz Yagi to get 20 dB at 1296
MHz, but less than 0 dB at only 1 MHz away. Needs some
thought!
The level of the sidelobes is not computed with the GA or
some other optimisation types. This will be corrected
later.
All those I don't know about.
AUTHORS
- Dr. David Kirkby G8WRB (david.kirkby@onetel.net). with
help with converting to DOS from Dr. Joe Mack NA3T
(mack@fcrfv2.ncifcrf.gov)