There are eight different variables that affect the way the Fletcher-Reeves algorithm used by OptiMatch works. The Default Utility allows you to change any or all of these eight variables on a "permanent" basis. We divide the default variable values into two groups. We call the first four the primary defaults, and the last four the secondary defaults. You can save any changes you make, and, unless you change them back, they will be in place each time you run OptiMatch.

The primary defaults are:

                Exponent       :     2
                Weight          :      1
                Goal              :     1
                Average        :     Y

The secondary defaults are:

                Search limit  :     100
                Error           :     1E-5
                Step            :     -0.1
                Delta           :     1.001

This utility will let you change the current unit set, or restore the original default unit set shown above.


When you choose "(4) Default Utility" from the Main Menu, either by pressing "4," or by using the "F4" function key, you'll enter the Default Utility. After a brief introductory message, the screen display will look as follows:

Primary Defaults             Secondary Defaults                 1.  Exponent      ( 2)         5. Search Limit    ( 100)                 2.  Weight           ( 1)         6. Error                  (1.0E-05)                 3.  Goal               ( 1)          7. Step                  (-0.1E+00)                 4.  Average        ( Y)          8. Delta                 (1.001)                 Change  (1-8, R[Reset],  [<Enter>=Quit] ? _

The values shown above are what we call the default variable set. We've found them to be optimal in most cases. To make a change, type "1,2,3," etc. and then enter the new unit value.

After making a change, you're then asked:

                Save changes (Y/N) ? _

If you choose "Y," you'll see:

                The 'Default File' has been saved.


Primary Defaults
The exponent is the power to which the error function is raised. It must be an even integer to avoid negative values.

Sometimes, since the error function may become quite small, we can add a weight factor to raise the error function up out of the numeric noise floor.

The goal is usually set equal to 1, since that represents a perfect VSWR.

Finally, we may form a weighted average by dividing the sum of the squared errors by N, where N is the number of frequencies in the optimization band.


Secondary Defaults
The search limit sets the number of linear searches allowed, without a reduction in the error function, before a search limit termination occurs.

The error value controls the fine grain optimization limit. When each variable reaches the point where it is changing by less than 1E-05, the search will stop. If the error value is set too large, a global minimum may never be found. On the other hand, if it is set too small, the algorithm may "rattle around in the numeric noise floor."

The step, actually the initial step, has to do with the expected change in the error function as a result of taking a first step. It has been observed that 10% is a good number. However, if you have need of a really fine grain search right near the global minimum, you might want to reduce the default value.

The delta is the change in element value used when computing a partial derivitive by the method of finite differences. The default value of 1.001 allows a change of 0.1% in element value to be made. A delta of 1.0001 would allow a change of 0.01%.


Normal vs Modified
When the default variable set is in place, a look at the System File, reached by using "5" from the Main Menu, will read as
Normal. If the default set has been modified, you'll see the word Modified.

It's easy to restore the default set. All you need do is access the Default Utility, then type "R" for Reset. Be sure to save the changes.


Auto vs Expert Mode
There are two different ways to use OptiMatch. The first, what we call Auto mode, makes use of a standard error function, and uses the default variable set. This set is one we've found best in most situations. However, the VSWR of some circuits may be improved by doing things slightly differently. One thing we've noticed is that it is sometimes possible to create a VSWR vs frequency response that is equi-ripple instead of parabolic in form. What happens is that the error function can force the VSWR at mid-band to get slightly worse, while improving it at the band edges. This can sometimes be brought about by simply changing the exponent, which normally "squares" the error function, to a higher even order value.

If you elect to use the Expert mode, to try and improve the optimization results, you can temporarily alter any, or all, of the four primary variables by entering the values you wish to use. If you want to make these changes "permanent," use the Default Utility to do so.