[TowerTalk] DX Engineering's "Non-Inductive" Resistors

[Tom Rauch]

Hi Jim and all,

> I measured at 1 MHz intervals  between 1 MHz and 10 MHz
(down to 2 MHz on the
> MFJ). The AEA and MFJ agree within reasonable experimental
error that the
> resistor shows considerable reactance at even the lowest
frequencies.
> Interestingly, the magnitude of the impedance stays fairly
close to 470 ohms over
> this range of frequency on both analyzers, and the typical
phase angle is on the
> order of 17 degrees.  And while the phase angle does
increase a bit with
> frequency, the increase is not anything close to
proportional.

On a network analyzer with near-zero lead length, those
resistors are virtually flat at frequencies up to 150MHz. In
contrast a typical MOX resistor of the same rating can have
20-30 ohms (or more) reactance at 30MHz.

The reactance in the metal composition resistors is a very
tiny part of overall resistance, so I'm pretty sure your
load is good with the exception of transmission line effects
of the connector body and resistor. I'll try to measure one
in a PL259-SO-239 combination on my network analyzer and see
how it is and let you know.

What you are seeing is a combination of three things:

1.) The bridge in these analyzers (even the AEA) is 50 ohms.
They are not reliable for reading impedances 10X the basic
bridge impedance

2.) The bridge is located a small but important distance
from the resistor, and neither analyzer compensates for
transmission line effects in that distance. That effect is
small at 50 ohms, but considerable at high or very low
resistances.

3.) The A/D conversion is very sensitive to even a single
bit of error, in particular at impedance far from 50 ohms.
The primary effect of that error is to "confuse" the
analyzers into thinking something is reactive when it is
not.

In the MFJ analyzer, each diode detector converts through an
8 bit A/D conversion. The voltage range is 0 through 255
bits, or 0.4% resolution. These analyzers determine if a
load is reactive by comparing the "bits" of voltage across
series detector (Vs), voltage across the load (Vz), and
source voltage feeding that divider. When the sum of Vz and
Vs is greater than source (255 bits), the load is considered
reactive.

One single bit of error causes a large jump in reactance.
One bit error is about 10 ohms jump at 50 ohms, and about
100-ohm jump at 500 ohms resistance! I decided to smooth
that jump in the MFJ by comparing SWR to R and Z and
weighting the difference (SWR and Z can be used to calculate
R and X) but that only works well around 50 ohms, where the
SWR bridge is accurate!

I don't have any 259's handy, so I just measured a resistor
on a 12 bit prototype of the MFJ 269 I normally use.  I get
to about 4Mhz and reactance jumps from zero ohms to about
100 ohms in a 1MHz increase. I'm not sure what firmware it
has. Even with 12 bits, it is "froggy" with 470 ohms.

Is 470 ohms simply too far from 50 ohms for the CIA and MFJ
to
> give acceptable measurements for X and the phase angle?

Yes.

> The 200 ohm load I assembled from four 1/8w composition
resistors to run through
> Tom's calibration reads 200 ohms resistive at 16 MHz, and
begins to show some
> inductance above 25 MHz on the CIA. The MFJ begins to see
inductance well
> below that.

A good non-inductive 200 ohm load should show capacitance
above a certain frequency, because the dominant rectance is
shunt C of the ~50 ohm connectors and foil traces/bridge. At
low resistances (say 12.5 ohms) the series reactance will
dominate and the load should look inductive.

The CIA guesses at the sign of the load, since like the MFJ
it only makes scalar measurements. Move frequency of the 259
slightly higher and see if reactance increases. If it does,
call it inductive. If reactance decreases, call it
capacitive. Now you just made the 259 guess the sign of the
reactance! It might be correct, it might be wrong.

In this case it will be a wrong guess, because the error
increases with frequency so it looks like reactance is
increasing. This tells the analyzer it is inductive, even th
ough with the shunt C of the connectors and other parts of
the system the bridge is actually seeing a very small
increease in capacitance.

Pretty interesting how these cheap scalar impedance
detectors work, and what the limits are. You have found a
very good example of the limits created when a cheap
detector is used to measure impedance. They still are useful
tools, but not for every application.

73 Tom