[TowerTalk] Measuring coax loss with MFJ-259B

[Ian White GM3SEK]

Pete Smith wrote:
>I recently replaced my two main 190-foot feedlines to the tower with
>LDF4-50A hardline.  On the shack end, there is about 50 feet of fairly old
>RG-213 (12 years or so).  Today I used my 259B to measure the combined loss
>at 30 MHz, with the far end open-circuited.  I got 1.3 dB on one cable and
>1.4 on the other, which correlates very nicely with the rated attenuation
>of the cables.  However, there was one strange thing - strange to me,
>anyway.  with the 259B in coax loss mode, I found that relatively small
>frequency excursions would cause a relatively large change in the reported
>loss - going, for example, from 1.3 dB at 30 MHz to 1.6 at 29.6 MHz - and
>of course, if anything, you'd expect the attenuation to go down with
>decreasiong frequency.  The variation in reported loss is also cyclical as
>I tune further down.
>
>The manual for the 259B doesn't say anything about this sort of frequency
>sensitivity, but I remember that with my old Autek it was necessary to tune
>for minimum R and then apply a formula to determine loss.  Is this just an
>omission in the MFJ manual, or something I don't understand, or maybe a
>problem with the unit itself?
>

It is probably due to the limited directivity of the Wheatstone bridge 
circuit. If the bridge is not perfectly balanced, there will be some 
small "leakage" of the forward signal into the detector port that is 
supposed to be seeing only the reflected signal. This forward leakage 
will then interact with the reflected signal, and depending on their 
relative phases the two signals may either add or tend to cancel.

If the forward leakage adds to the reflected signal, the impedance match 
will look worse than it really is. That means the computed return loss 
and cable loss will look lower than they really are (the one-way cable 
loss is computed as simply one-half of the two-way return loss). If the 
leakage tends to cancel the reflected signal, the errors go the opposite 
way.

When you sweep the frequency, the true cable loss isn't changing by 
much, but you are rapidly changing the phase of the reflected signal. 
That's what is causing the cyclical behavior that Pete is reporting. The 
longer the cable, the faster the phase will change.

Whenever you try to measure something, you're measuring the performance 
of your test equipment as well. A single-ended measurement of 
return/cable loss will always be very error-prone, so Steve K8LX has got 
it right: "This is perfectly normal behavior". The only difference with 
the MFJ-259B is that the built-in computing power makes the error more 
obvious - but it is present in all other kinds of directional couplers 
too.

Steve continues: "I sweep feedlines on a regular basis with Anritsu 
Sitemasters, and usually use (highest peak + lowest valley)/2 as my 
cable loss value."  A simple average between maximum and minimum 
readings in dB will not give the correct value for the cable loss... but 
that's only a detail. The most important thing is to keep notes, so you 
can see if anything has changed since the last measurement.


-- 

73 from Ian GM3SEK
http://www.ifwtech.co.uk/g3sek