Note that inaccuracy is only relative to gain with real grounds.
For example, if the model is well designed and has enough segments, AO
perfectly deals with metal ground screens that assimilate a perfect ground
(i.e. DDRR antennas) or with close in elements like in open sleeve yagis.
I wouldn't even think to deal with a low horizontal antenna (less than 0.2
WL) over a poor ground without a metal screen or radials.
... as usual this is another story, thread: "magical" antennas.
----- Original Message -----
From: "Guy Olinger, K2AV" <firstname.lastname@example.org>
To: <email@example.com>; "Pete Smith" <firstname.lastname@example.org>
Sent: Sunday, June 09, 2002 5:38 AM
Subject: Re: [Antennaware] Re: NVIS -- modeling, mininec vs nec-2 disparity
> Finally dug up the specifics of the disparity.
> To say that Mininec is "inaccurate under .2 wavelength" because ground
> loss inaccuracies can become "significant" really doesn't do the job
> about when to switch to another method.
> The following snippet from the Doc for AO. The methods are Mininec
> methods, which the optimizing shell in AO uses as a core.
> "You should exercise caution when modeling antennas over
> ground. AO uses ground characteristics only to determine the
> ground-reflection factor for the far-field tabulations and
> patterns. It uses a perfect-conductivity groundplane when it
> calculates wire currents. This implies that ground-current
> losses are not accounted for. While these losses normally are
> negligible for horizontal antennas higher than about 0.2
> wavelength, they can be significant for low horizontals and for
> verticals fed against poor ground systems."
> Or as you go past .2 wavelength, start worrying whether you are
> ignoring near-field ground losses. It's not like the model falls off
> the table at some point approaching ground. It DOES fall off the table
> on some other issues like parallel wires closer than something like an
> inch. Go past that bogey and it gets quickly strange.
> Frankly, this little exercise is the first AO generated solution of
> mine that had anything in it that showed any diff with NEC other than
> minimal center frequency or impedance diffs.
> Apparently seven feet off the ground for the reflector is
> detuning/lossing up the reflector.
> Whether that is a loss related issue or AO picking a length that ought
> to be something else, remains to be seen. AO stuck the reflector down
> there because it thought it was hooking some ideal ground
> reinforcement. Seven feet not all that close. Suspect there is a
> better length/height for the reflector.
> It is much harder to generate iterations through EZNEC looking for a
> solution than to let AO go looking for it automagically.
> Nevertheless, I owe you a beer or two, Pete...
> Will constrain the height of the solution and see what that reflector
> is at 28 feet.
> ----- Original Message -----
> From: "Pete Smith" <email@example.com>
> To: <firstname.lastname@example.org>
> Sent: Friday, June 07, 2002 7:17 AM
> Subject: [Antennaware] Re: NVIS -- modeling, mininec vs nec-2
> > At 09:51 PM 6/6/02 -0400, Guy Olinger, K2AV wrote:
> > >Going to post this to antennaware, with the figures to see if the
> > >gatheren already know about this.
> > >
> > >The antenna in question...two element reflector beam pointing
> > >up.
> > >
> > >7.025 MHz, upper wire fed in center
> > >
> > >Upper wire = 67.5 feet long, 47.7 feet high (21 segs on eznec, 16
> > >segs/hw mininec)
> > >Lower wire = 68.4 feet long, 7.4 feet high
> > >
> > >AO (mininec) optimizes to these numbers, gets 11 dbi at 75 degrees
> > >with 11.4 dbi straight up.
> > >
> > >Anyone know about this one and care to comment?
> > I'm not sure how much of the context was on the reflector, so let me
> > add a little. The same model, with NEC-2 and Sommerfeld-Norton
> > shows a VERY different result. Instead of peak gain being at the
> > it is at 41 degrees and maximum is only 5.99 dBi. With what EZNEC
> > MiniNEC ground, results from NEC-2 are close to those Guy reports
> > AO. Since doubling the number of segments produces very little
> change in
> > these results, I presume they represent good convergence.
> > The question is, which modeling result is closer to reality?
> > 73, Pete N4ZR