budgie wrote:

On Wed, 01 Dec 2004 16:35:36 -0800, Roy Lewallen wrote:


Duncan Munro wrote:

The kind of measurement frequencies we are talking about are in the order
of 700kHz. At that frequency, the inductance of the 'indicated' 4.5uH is
19.8 ohms, not a million miles from the 22 ohms of the resistor itself -
this is not what I would call a reasonable 'Q' value. Fair play to AADE,
it's designed to measure the inductance of inductors, not other components
;-)

If the reactance is much lower than the resistance, it's generally
inconsequential in a practical application. I think that's almost always
the case for carbon film resistors, and I suspect it's nearly always the
case for metal film resistors.

Probably, if the Q is so low as to make measurement difficult, it's
probably low enough that the X isn't important in a practical application.


(adds) ..... at the test frequency

Sure. Any statement about a frequency-dependent property like X or Q
applies only at the frequency at which the component has that particular
X or Q.

As you raise the frequency, the X of course increases while the R stays
relatively constant. But other effects like shunt C and the physical
length of the part eventually start coming into play, making the
simplistic model of a series RL inadequate. My general experience has
been that I can ignore the inductance of leaded carbon film resistors up
to a frequency where the leads and component length become a problem,
and I need to go to chip components. I've never seen significant
reactance from the trim cuts on a thick film chip resistor -- the shunt
C across the narrow cuts pretty much makes them invisible.(*) I suspect
that carbon film resistors likewise have a narrow cut. But I don't have
much experience with metal film resistors. I assume the base material
has less resistivity, so is probably cut into thinner strips with more
"turns" and more spacing between "turns". So there might be a
combination of R and frequency where the reactance is objectionable,
below the frequency where you need to abandon leaded parts. I'm watching
this thread with interest for any good measurement results. I could try
making some measurements up to 1.3 GHz with my 8505A network analyzer,
but I wouldn't trust the results. I think the measurements probably
would have to be done on a system with good, computer-directed
calibration, a good set of calibration standards, a decent and
well-characterized test fixture, and an operator who's very familiar
with the many traps you can fall into when making subtle measurements
like these -- and I have none of the above.

(*) I've used thick film resistors at frequencies up to 20 GHz or so, in
very sensitive time-domain applications. In those applications, I
modeled nearly every component as a transmission line or a pi or tee
approximation to a line, with the R in one or two lumps. Those models
agreed quite well with actual results.

Roy Lewallen, W7EL