Jim Brown wrote:
> N6RK suggested that the MFJ259B would be a good way to measure coax loss with
> a reflection measurement at one end only. I responded I was looking for more
> precision than that. Several years ago, I carefully calibrated my 259B using
> W8JI's procedure. Today, I compared loss measurements of a 208 ft length of
> Commscope 2427K (the plenum version of 3227) made both with my 259B and with
> my HP generator/spectrum analyzer setup.
> The HP data has one digit better precision -- the spectrum analyzer readout
> is 0.01dB, while the 259B reads 0.1 dB. Within those limits, the MFJ yielded
> results that were usually within rounding error of the HP setup between 10MHz
> and 150 MHz, which is where I stopped measuring (because that's the highest
> frequency I cared about for my applications). For example, at 50, 100, and
> 150 MHz the HP numbers were 2.07 dB, 2.84 dB, and 3.37 dB; the MFJ yielded
> 2.1 dB, 2.8 dB, and 3.3 dB. At 3 MHz (the lowest frequency I could measure
> with the MFJ), the numbers were 0.62 dB and 0.6 dB.
> For obvious reasons, you really need the greater precision of the HP setup to
> get good data for frequencies and lengths where the loss is relatively small.
WHile the HP spectrum analyzer has a *precision* of 0.01dB, I'll bet its
uncertainty is bigger than that (0.1dB at least, maybe 0.3 to 0.5dB). If
you used it for a substitution measurement (i.e. you put a calibrated
attenuator and adjusted to match to the same reading), it would be better.
Even a new power meter (which is generally better than the log amp in
the spectrum analyzer) is probably only good to a few percent (0.1dB).
(the 8563E says the scale fidelity is 0.1dB/dB for RBW>300Hz)
A calibrated VNA can do sub 0.1dB over small dynamic ranges, mostly
because it can take care of the uncertainties with mismatch, which the
power meter and spectrum analyzer don't do so well on.
For this kind of thing at work, I use a 8902A measuring receiver, which
claims ±(0.015 dB + 0.005 dB/10 dB step)
or in other ways, precision of 0.01dB (0.01% of full scale in linear
mode), +/-0.02dB or +/-0.03dB linearity plus another 0.02dB uncertainty
if your signal is in a different range from the signal you're comparing
against (range is roughly 30db steps 0 to -30, -30 to -60, etc.)
I suspect the latest Agilent PNAs can do comparable, at least for low
loss measurements (measuring down at -100dBm, you've got noise floor
TowerTalk mailing list