RE:NEC-LIST: 4NEC2 Optimizer

And, for most practical applications, gain isn't the only figure of merit
you're really interested in. Typically, you may be more concerned about
things like lack of radiation at low angles and low side or back lobe
levels, rather than eke'ing out the last 0.1 dB of forward gain.

There's also the problem of high element currents (and corresponding losses
in a "real" device) when the gain gets high. Consider something as simple
as a W8JK two element phased array. Get the elements real close together
and the gain gets pretty high, but, in a real antenna, the loss starts to
bite you, the feedpoint impedance starts to get ugly, and the bandwidth
starts to get narrow.

The other aspect of optimization to be concerned about is manufacturing
tolerances, aging effects, and sensitivity to outside environment and
disturbances. All of these are a bit tricky to deal with in a practical
simulation environment, other than by a sort of Monte Carlo approach where
you run lots of sims, with random variations in the parameters around your
current optimizer point. It does you no good to optimize to the world's
best 7 MHz antenna if it requires that the element dimensions be controlled
to 0.1 mm, and that it be operated in free space.

The variabilities in environment alone are probably greater than the change
in gain/performance from the last few passes of the optimizer.

At 11:20 AM 12/11/2002 -0800, you wrote:
>With larger Yagis at least I find it helps to think of the ensemble as a
>lowpass
>"filter". Leaving matching aside for a moment, consider the *possible* or
>achievable gain vs frequency function.
>
>On the down-frequency side (from maximum gain frequency) the achievable
>gain vs frequency envelope diminishes monotonically as frequency is reduced.
>In contrast on the up-frequency side (from max gain freq) once the WL gets
>short enough (and this happens very quickly with increasing frequency for
>multielement gain-optimized designs) one or more of the directors suddenly
>flips phase and thus become reflectors. As more of the "directors" become
>reflectors with increasing frequency, the sharper the cutoff, not unlike
>filter
>behaviour with increasing numbers of circuit elements. So the gain vs
>frequency
>envelope function is far from symmetric with frequency. The practical
>implication
>of this is important -- working towards the low side is "easy" but that
>upper end
>can be dangerous territory. Cutting Yagis (element length) on the
>"too-short" side
>is always safe, although lower gain performance, but when going for max gain
>one needs to be careful of that upside cutoff when pushing element lengths
>towards the "too long".
>
>It is under (within, or constrained by) this "global" gain envelope that
>matching is
>done. That is to say, the match bandwidth of large Yagis is usually a
>relatively
>narrow slice in frequency compared with the above global (achievable) gain-vs-
>frequency envelope. Once one thinks on this a bit, it is far easier to
>comprehend
>than it is to cogently describe in words. Well, I tried.

-- 
The NEC-List mailing list <nec-list_at_gweep.ca>
http://www.gweep.ca/mailman/listinfo.cgi/nec-list