When you look at EMC analysis and prediction
you see nearly every possible formulation brought
into play depending mostly, I suspect, on the
prejudice of the formulation guru. Now that you
simply buy an Ajax Storm Door EMI Prediction Program
that runs on your PC the situation is even worse
because the assumptions used in selecting the
algorithms are never stated, and although the
same old trash is included, you dont know
I dont mean to say that the formulation
is wrong. Most of it comes from a
text somewhere and ultimately from someone like
Kraus who knew exactly what was going on and why
but had to get simple answers so that his students
could understand his text. So he made carefully
described assumptions like you were so far from
the source that the waves are plain and have only
frequency, amplitude and direction of arrival.
If you look at the source you see a point of light.
The stars are genuinely in the far field except
for one of them the sun. It is not in the
far field. You dont see a point of light
you see the shape of the source. Youre
in the near field. It may be close enough to far
field to suit you, but it is near field.
The EMC folks, I hesitate to use the term engineers,
settle for having the largest of the near field
terms be the same size as the far field terms.
This guarantees at least a 6dB difference between
the exact solution and the calculated value. Why
not just guess based on experience? Why fool around
My next favorite stupid machination is using a
half-wave dipole to measure the emissions from
a distance of 5 meters at say 40MHz. When I drive
out of my driveway, I face my neighbors
house on top of which he has an 11-element two-meter
Yagi antenna. It works. He gets about 10dB gain
out of the thing because the elements are affecting
each other. Each element is close to a half-wavelength
long and about a quarter-wavelength from its nearest
neighbor. If the whole mass were scaled to 40Mhz,
it would be about 15 meters long and every one
of the elements would affect the current flowing
in every other element. Thats why the darned
thing works. If the antenna is an array, not a
dipole, you need to be even farther away.
The way we calculate the E-field set up by EMI
currents in a culprit circuit is fascinating.
First, we figure the current based on Ohms
law probably okay if we include the inductance
of the wires or clads in the circuit, often the
current-limiting factors in circuits above about
But then we fly blithely into never-never land.
We get out the antenna book and pull out a formula.
The problem is that antenna books are written
about standing wave structures dipoles
and short circuit loops because they work
best as antennas. By and large the currents in
the culprit circuits or on its circuit boards
are traveling wave structures, that is the currents
are in conductors connecting a generator to a
load and are not standing waves. We almost always
use the wrong equations. The equations are technically
fine but they are the wrong equations for the
situation. Its like trying to explain the
behavior of an octopus based on the known responses
of a turkey.
Also, we tend to forget the generator impedance
when we are calculating EMI noise. Ive seen
people predict thousands of volts per meter based
on curves which supposedly predict E-field based
on noise currents. Their argument
goes like this: The open circuit noise voltage
is up-dee-dink volts, so if we connect the generator
to a circuit with an inductive impedance of a
micro ohm, we get up-de-diddle mega amperes of
noise current, and the curves say this produces
kilovolts of radiated E-field. Bzzzatt
wrong! The noise generator has an internal impedance.
If it is one ohm, then the maximum current that
can flow in the circuit is, according to Ohms
law, one ampere even if the external loop is a
short circuit. And by the way, who said Murphy
is smart enough to go around matching incidental
noise generators to their accidental loads?