Steve
I have also seen cases where EZNEC will not repeat the results of a
calculation after reopening a file and re-running it. It's sometimes
frusterating when you are preparing some data, need one more piece of
data, rerun the file and everything changes. I always thought it was
because EZNEC was changing the data it was using for the ground model.
It stores that data and doesn't necessarily recalculate it every time,
and I thought it was just making a different decision about which
ground-data file to use. Usually when it does this, the answer is only
different by tenths of a db, however in your case, seems it was making a
really gross error. Check your setting for "ground file tolerance".
Here is what the EZNEC help file says about it:
When High Accuracy ground is used, a Sommerfeld-Norton interpolation
table is calculated. Although this calculation is quite fast with modern
computers, the resulting table is saved on the disk. If future runs
require a table with similar values, an already-calculated table is read
and used instead of repeating the calculations. The key parameter which
determines whether an existing table is close enough is the ground's
complex permittivity (which is a function of the frequency and the
ground conductivity and permittivity). If a table is found for a
magnitude within a user-specified tolerance of the required magnitude,
it will be used. You can change this tolerance, if desired, from the
Control Center Options menu. A tolerance value of a percent or so is
adequate where horizontal wires aren't extremely close to the ground. In
situations requiring high accuracy, or where there are low horizontal
wires, a much smaller tolerance, perhaps 0.05%, is recommended.
Jerry, K4SAV
Steve Hunt wrote:
> Well, I'm pleased to report I'm not going senile - well, I don't think so!
>
> I'd forgotten that I saw a similar problem last year on a Hexbeam model
> and contacted Roy Lewallen about it. I just found the model again and
> can reproduce a very similar bug:
>
> Double click the file to open EZNEC
> Click FF Plot
> Azimuth plot shows Gain -3.47dBi, Front/Back 26.58dBi
> Click Ground Type
> Click Free Space
> Click OK
> Click Ground Type
> Click Real
> Click OK
> Click FF Plot
> Azimuth plot shows different results: Gain -4.04dBi, Front/Back 40.13dBi
>
> I get the first results consistently when I've loaded the file until
> such time as I select Free Space and revert to Real Ground; thereafter I
> consistently get the second set of results. The model is complex and
> produces a segmentation warning, but I can't understand why the results
> wouldn't be repeatable.
>
> Notice that I don't even need to do a computation run to correct the
> problem - simply selecting Free Space and then immediately re-selecting
> Real Ground is enough to correct it. I sent Roy the file in question but
> didn't ever hear back from him.
>
> If anyone would be interested to check whether this behaviour is
> repeated on their version of EZNEC - I'm on EZNEC+ v.5.0.20 - I'd be
> happy to forward the file.
>
> Steve G3TXQ
>
>
>
>
> Steve Hunt wrote:
>
>> Jerry,
>>
>> My head is spinning after 30 minutes of trying to compare our figures.
>> I'm sure you wont believe this, but I believe there is a bug in EZNEC.
>>
>> I loaded up the vertical model I was using last night - 132ft over
>> Real/MININEC average ground, Aluminium Wire Loss, and re-ran the
>> figures. My numbers were several dB higher than yours. I then switched
>> to Perfect Ground and Zero Wire Loss to check my Average Gain figure -
>> it was very low. Then I switched straight back to Aluminium Wire Loss
>> and Real/MININEC ground and got a completely different set of figures -
>> very close to yours.
>>
>> I repeated the exercise from a "cold start" (closing down EZNEC and
>> restarting it) about 6 times and consistently saw the same thing - an
>> erroneous set of figures at first, and then a correct set once I'd
>> swapped to Perfect Ground and back. I was trying trying to bottom-out
>> exactly which change caused the change in behaviour, when it stopped
>> doing it and I now can't reproduce the effect.
>>
>> I wonder if anyone else has ever seen this? Please don't suggest it was
>> "operator error" - once I'd first identified the effect I was very
>> careful to check and double-check every setting.
>>
>> Steve G3TXQ
>>
>>
>
>
_______________________________________________
_______________________________________________
TowerTalk mailing list
TowerTalk@contesting.com
http://lists.contesting.com/mailman/listinfo/towertalk
|