Hi Eric,
> >How did you model the trees, wires, cables, trucks, cars,
> >busses, towers, poles, and other things in your program?
>
> I didn't. I didn't need to because I wasn't trying to
> characterize their effects on the test. Hopefully, when others
> repeat this effort, they will have selected a protocol that makes
it possible to eliminate or minimize the effects of site clutter.
Then it doesn't mean anything as to my comments, because you
did not include the effects I was talking about.
> >Also, did you model the accuracy of a spectrum analyzer used on
> >a 10 dB log scale to see if it really can resolve fractional dB
> >differences, and a Bird meter used to read true power and SWR
> >operated at 1/4 scale?
>
> Nope. I wasn't trying to evaluate the absolute accuracy or even
> the absolute meaningfullness of the testing. They did what they
> did, measured what they measured, documented the process well
> enough for it to be evaluated by others (certainly gave you
> plenty of anmmunition), and reported the results.
First let's look at the Bird meter. A Bird is + or - 5% of full scale
anywhere on the scale. That includes reflected power, along with a
directivity rating of 30 dB **IF** the slug has been properly
calibrated.
Ignoring the directivity issue, accurately resolving the difference
between a 1:1 SWR and a 1.5 :1 SWR with a single slug of 100
watts operated at 100 watts would be impossible.
With a 2:1 actual SWR and a single slug set at full scale forward,
the Bird could indicate anywhere between 1.5 to 1 SWR and just
over 2.5:1 SWR.
Besides the feedline loss errors, there would be a relative reference
power delta of almost 26 watts with a 100 watt slug and an actual
working SWR of 2:1 .
That's an error of plus or minus one half dB *IF* the Bird is used for
repeating power levels (not measuring absolute power) with SWR's
of 2:1 maximum. That error ignores feedline loss errors. It also
assumes the meter is set for full scale in forward power readings.
The difference between two things measured, including power
applied to two dummy loads or antennas, could change by as
much as a dB if a single slug is operated at full scale and turned
around.
That isn't absolute power, it is resetable power.
Now I appreciate all the work everyone does, but we have to face
reality. There is the potential for a few dB of error in any
measurement like this unless the proper equipment is used the
proper way, and that ignores potential site errors.
Just because something can be read to .1 dB, or calculated to .1
dB, does not mean it is accurate to anywhere remotely near a dB.
> Now, I believe that it would be far more useful to refine the
> test protocol for the next time someone is willing to do the
> (tremendous amount of) work to try and verify or refute the
> results by measurement than to argue about the test results of
> the first test themselves. Once this has been repeated a few
> times and the protocol has been refined, we will be able to look
> back on this particular data pont with the benefit of hindsight
> and determine how meaningful it was or wasn't.
My point exactly. One suggestion would be to use an accurate
SWR measurement device, and read reference power at full scale
on a large scale meter. Do the power corrections manually, or use
a low power slug (properly calibrated) near full scale to read
reflected power.
That would reduce resetable power error to a fraction of a dB.
I'd also use a meter, NOT a spectrum analyzer, and a calibrated
pad for FS reference. That would take out another potential error of
a dB or more.
Finally, I'd get a clear test site with a reasonable distance, maybe
500 feet or so, for the measurement. I'd verify the pattern of the
dipole to be sure it was normal, and do the FS measurement in at
least two distinctly different directions to check for multipath.
Better yet, I'd just use a reference antenna with a similar pattern
and known gain to the antenna's under test.
I'd validate the measurement with a different protocol and different
people and instruments.
At that point, I'd bet we'd be within a half dB of absolute in the
comparison of on antenna to another.
Understand I'm not saying the present results are right or wrong,
just that they have the potential for some serious error. It would be
nice to reduce that error as much as reasonably possible.
It is poor form to ignore the errors, and proclaim accuracy to .1 dB
in a test setup like this. I think most people know that. It's
especially worrysome that all similar antennas but of different feed
systems have patterns (in the first test) all show a large amount of
skewing to one side. Worse yet, it is always the same side!!
That is clearly a multi-path warning indicator, meaning gain
measurements are also likely off. That flag might not be red, but it
is at least bright yellow. As are the uses of instruments.
73, Tom W8JI
w8ji@contesting.com
--
FAQ on WWW: http://www.contesting.com/towertalkfaq.html
Submissions: towertalk@contesting.com
Administrative requests: towertalk-REQUEST@contesting.com
Problems: owner-towertalk@contesting.com
Search: http://www.contesting.com/km9p/search.htm
|