[TowerTalk] steppIR Gain--test method

[Jim Lux]

At 11:56 AM 3/16/2005, Rick Karlquist wrote:
>I believe someone else said they used a step attenuator.
>Secondly, they were doing differential measurements between
>a dipole and a SteppIR, not absolute measurements of power.
>
>In any event, the 141 series can be quite accurate if you know
>what you are doing and let the instrument fully warm up.  There are
>several adjustments (eg gain and offset) that need to be made.  If these
>are properly made they are really quite accurate over a narrow amplitude
>range (much better than a couple of dB).  I have adjusted many of them and
>checked them with calibrated signal generators which in turn had been
>verified with power meters.  If you read the specs on the 141 series, it
>may seem like they are not so accurate.  Having worked for HP/Agilent for
>25 years, I can assure you that may considerations go into setting
>those specs, and in many cases they are very conservative.
>
>Rick N6RK

I agree with Rick here.  They were doing a classic substitution 
measurement, and the required dynamic range was around 10dB (i.e. the 
difference in gain between the reference dipole and the AUT), at least for 
the max gain.  Given that the F/B is probably no better than 20dB (in 
situ), the total measurement dynamic range was really in the 30dB range, 
and almost any spectrum analyzer will be good to better than a dB, in the 
top 30 dB of the scale at a fixed frequency, if you don't change any of the 
settings.  You're really looking at the accuracy of the log detector, and 
that's fairly good in most cases.   Nobody should be trusting F/B numbers 
to an accuracy of better than a couple dB anyway(except for very low gain 
antennas).

If they had a step attenuator with calibrated <0.1dB steps (no easy feat to 
come by one of those, by the way), and used the "match levels" approach 
(which would remove calibration of the spectrum analyzer as a source of 
error), the other errors in the range measurement would dominate.

One can get pretty good absolute gain measurements by using the "three 
antenna" substitution technique, assuming you correct for the other range 
effects.

If I had to guess, without doing detailed analysis, the two big sources of 
error that they would have are:

1) Measuring at ground level, with only one probe height, so you can't 
quantify the effects of the ground reflection, and convert to a free space 
number. The textbook by Krauss (and other books as well) have a good 
explanation of how to do range measurements on a ground reflection 
range.  (this is the "null at the horizon" problem Tom mentioned)

2) Multipath effects on the reference antenna (not the probe, but the one 
being used to replace the AUT), because the directivity of the reference 
antenna is quite different than the AUT.

Other, more minor problems would be things like polarization and weather 
variability which is usually an issue on outdoor ranges: humidity and 
temperature affect the RF properties of the soil and/or vegetation, which 
in turn affects near field loss and multipath and polarization 
effects.  Something else which may not have been well controlled is the 
transmitted power, which could be affected by a whole host of factors 
(power supply voltage, rig temperature, etc.)

Making good antenna measurements, other than main beam gain and 3dB 
beamwidth, is fairly challenging, especially when you're more than 20dB 
down from the main beam.  Getting the multipath sources below 40 dB is very 
challenging.  To get 1 dB accuracy at -20dBc, you need to make sure that 
the interference source is no more than -26 dBc.  To get 0.1 dB accuracy at 
-20 dBc, the reflections need to be -36 dBc.  It's left as an exercise for 
the reader to determine just how big a chunk of metal you need on the range 
to scatter -36 dBc (it's not real big).

In decent anechoic chambers, you typically can get the wall reflections 
down to -40 dB or so.  This is good enough to get measurement uncertainties 
around 1dB when you're down 30dB from boresight.

The summary is that range effects will likely be the dominant source of 
error, even if you don't have a very sophisticated measurement receiver.