Experiments In EMC:
How Common Mode Currents Are Created
By Glen Dash, Ampyx LLC, GlenDash at alum.mit.edu
Copyright 2000, 2005 Ampyx LLC
“I’ve all ready read the
books on EMC and visited a lot of home pages... But all these references did
not mention anything about the physical phenomenon that causes common mode
currents... Are common mode emissions inherent in any physical system? Can I model them?”
--Overheard on the ‘Net
It’s by no means a trivial
question. And, in spite of decades of
hand waving by authors and consultants, the principal mechanism by which common
mode currents are created in digital devices was not well understood until the
decade of the 90s. In this article,
we’ll explore the physics behind the creation of common mode currents, and
perform some experiments to verify our understanding.
We begin with the simplest of
circuits, a signal source driving 10 cm of 300 ohm twin lead shown in Figure
1. In one way or another, all wire line
communication has as its goal transmitting signals faithfully from a source to
a load. Here the load is matched to the
line, and good fidelity can be expected.
(Note that since the transmission line is matched to the load, there
will be no reflection at the load end.
Therefore, it is not necessary that the source be matched to the line.)
Figure 1: Our analysis starts
with a simple circuit. A voltage source
drives a short length of 300 ohm twin lead, terminated in a 300 ohm load.
The
radiation that could be expected from the circuit in Figure 1 is relatively
small. We can simulate the circuit on
our Method of Moments simulator (Reference 1).
It predicts the radiation at 3 meters for the circuit in Figure 1 to be
approximately 1200 uV/m at 3m (in free space).
We will make the circuit a
bit more challenging by adding two wires to the return path as shown in Figure
2. I1, the “forward”
current, will now be split at Node 2, some of it returning via the twin lead to
the source, and some of it moving down the wire now attached at that node. Current moving down the added wire, which
acts as an antenna, will partially be reflected back when it reaches the end of
the wire, and partially radiated into space.
Figure 2: Adding wires to the
return in Figure 1 create a more complicated circuit. Radiation increases dramatically because of common mode currents
driving the wires attached.
Can I3 be readily
predicted? It can, using a few
simplifying assumptions. Suppose the
wire “antennas” of Figure 2 are resonant at the drive frequency, 300 MHz. These two wires will have the characteristic
impedance of a dipole antenna at resonance: 73 ohms. Each “arm” of the antenna can be modeled as a resistance to earth
ground of one half of that figure, or 36.5 ohms. The twin lead itself, being very much shorter than the
wavelength, can be modeled electrically by the use of two inductors, L1
and L2 as shown in Figure 3.
Each is approximately one half the total inductance of the loop shown in
Figure 1. The formula for the
inductance of a short strip of twin lead is well known and the inductance is
approximately 140 nH. If we assume that
half of this inductance can be assigned to L2 , then this “partial
inductance” L2 is approximately 70 nH. Knowing that, we can readily calculate I3 from the
circuit model of Figure 3. We predict
that I3 =I4 = 1.2 mA.
These currents are “common
mode” currents. Here’s how common mode
and differential currents are defined:
Where:
I diff = Differential current
I cm = Common mode current
Figure 3: The radiation from the circuit in Figure 2
can be calculated using the circuit model of Figure 3. To calculate radiation accurately we will
need to know the partial inductances of the forward and return conductors. The added wires of Figure 2 are assumed to
be at resonance.
Rearranging terms, we find
that:
A loop’s differential and
common mode currents are defined as follows.
Since , by definition I 1 diff = -I 2 diff and I 1
cm = I 2 cm:
Our Method of Moments program
can be used to predict the radiation from the circuit of Figure 2. It is 27,500 uV/m at 3 m in free space, a
gain of 27 dB over the circuit in Figure 1.
Adding a couple of wires to the return of Figure 1 increases radiation
dramatically, and that radiation is due to common mode currents.
Figure 4: Any I/O cables attached to a circuit are
directly or incrementally connected to return wires (or planes). How much radiation results is strongly
affected by the length and layout of the attached wires. Even a circuit whose source is bonded to a
perfect ground plane will exhibit some radiation.
Figure 5: We built this assembly
to measure radiation from the circuit shown in Figure 4b.
Digital devices, of course,
are far more complex than our simplified model. For one, the various I/O cables (which act as antennas) vary in
length and geometry. Any variation in
length or geometry quite clearly will affect emissions. Even devices with a good earth ground (such
as a solid metal plate beneath the circuit whose smallest dimension is on the
order of a wavelength at the lowest frequency of interest) will not necessarily
result in lowered emissions. For
example, in Figure 4b we show schematically the assembly of Figure 5, one in
which one end of our return wire is bolted solidly to our earth ground, and the
other end is connected to a wire which is run up a short mast creating a vertical
antenna.
We built and tested this
assembly. Our assembly consisted of a
50-ohm source driving 300 ohms of twin lead that was terminated in its
characteristic impedance. The circuit
was suspended .5 inch (1.27 cm) above a large ground plane, and one end of the
return wire was bolted to that ground plane.
Another wire was connected to the far end of the return wire and run up
a short pole, creating the vertical antenna.
At resonance, a vertical has one half the impedance of a true dipole,
36.5 ohms. Vertically suspended above a
large ground plane, it also produces an image antenna, the net result being a
dipole with vertical polarization.
Measurements made with a
Tektronix CT1 current probe showed that 1.2 mA of current was flowing into base
the vertical wire. Simulations with our
Method of Moments software yielded a similar result, 1.34 mA flowing into the
base of the vertical wire with predicted field strength of 27,500 uV/m.
Therefore, we can show that
even simple circuits, well-matched in terms of their load and suspended a short
distance over a wide ground plane can still produce radiation when wires are
connected to their return structure.
Whether a two-sided or multi-layered board is used, there will always be
some partial inductance in the return, and therefore some voltage driving wires
attached, even those wires attached to what is sometimes (incorrectly) called
“signal ground.”
Figure 6: One sure fire way of
reducing emissions dramatically is to wrap our circuit in a shield. This forces return currents to flow on the
inside of the shield.
How can such radiation be
avoided? One method is shown in Figure
6. Here a 360 degree shield has been
thrown around our circuit. Note that we
have not connected that shield directly to our ground plane, or to any portion
of the circuit, except to its return wire.
The measured current in the vertical wire falls dramatically -- to 80
uA. The conclusion? Wrapping a complete shield around a circuit
will work just about every time.
Why does such a shield
work? It works because the return
currents travel on the inside of the shield. They don’t travel on the outside of the shield, and therefore,
there is no voltage drop between the source and the load. Said another way, a complete shield has a
partial return inductance of near zero.
That, in fact, is also why a
shielded cable works. The shield formed
by the braid or metalized foil traps nearly all the current inside the
shield. A perfectly shielded cable has
an effective return inductance of zero.
However, nothing being perfect, some current does leak through to the
surface of a shielded cable. That
current produces a small amount of “lost flux,” which, in turn results in a
small amount of radiated energy.
How do we use what we have
learned in practice? Designers have
used two methods to reduce emissions from circuits like that shown in Figure 2
through 4. First, they have abandoned the
return wire for the return plane of a multi-layer board. The effective inductance of a plane is far
lower than a wire, but is by no means zero.
Even in multi-layer boards, significant radiation can result. Wrap a tight shield around the multi-layer board,
however, and the radiation will drop dramatically.
Figure 7: I DD
currents drive supply planes as surely as our source in Figure 1 drives its
transmission line. Even though the
planes are close together, wide and uniform, they still exhibit an effective
inductance that causes radiation. The
use of planes alone may reduce emissions, but is not as effective a technique
as one that would trap returning currents, such as the shield shown at the
bottom of the figure.
Sources of RF currents do not
refer just to clock drivers and the like.
Each time an IC switches, it creates a pulse across its supply. These pulses can be a nanosecond or less in
time and tens of amps in amplitude for devices such as a microprocessor. These pulses are referred to as I DD
currents or I DD noise. How
does I DD noise become common mode radiation? Take a look at Figure 7. In essence it is no different than Figures 1
through 4 except that we have exchanged our 300 ohm twin lead for a transmission
line made of plates. Though of lower
characteristic impedance, plates form a transmission line nonetheless. Exactly the same effects illustrated above
will cause such a circuit to radiate.
The extent of the radiation can be calculated if the return plane’s
partial inductance is known. From
Reference 2, it is:
Where:
L
P = Partial inductance of the return plane in Henries
m0 = 4 p x 10 -7
l = Length of the return plane in meters
d = Distance between the planes in meters
w = Width of the planes in meters
That is why the use of a
multi-layer board, while in itself a powerful technique for mitigating
emissions, is not a perfect solution.
In most applications, some other technique has to be used to further
lower emissions. This may consist of
the use of ferrite cores over I/O cable (which effectively places an impedance
in series with the radiation resistance) or the use of a 360 degree shield
surrounding the whole circuit to which shielded I/O cables can be attached or
unshielded wires bypassed. A complete
shield will trap RF currents on the inside, leaving no voltage to cause
radiation across its surface.
References
1. EZNEC is available from Roy Lewallen, W7EL@teleport.com.
2. F. Leferink, “Inductance Calculations; Methods and Equations”
1995 IEEE Symposium on Electromagnetic Compatibility, page 16.