Radiation Paths and cable-to-cable coupling . This is the 4th article of our EMC awareness series. The former articles, after a broad overview of the EMC subject, reviewed the principal Civilian and Military Norms and test methods, insisting on the legal inforcement of these verifications in Europe, where they are turned into mandatory, « must-comply » laws. Given that source/coupling path/victim concept is the basic approach to EMC, most of the time it is the coupling path between the culprit source and the victim equipment wich is the crux of the problem, hence of its solutions. The 5 essential coupling mechanisms were listed, by which EM Interference take place. Although any equipment can be alternately the victim, or the source, of an EMI problem, we focused on EM susceptibility as being the manifestation that appears first in the designer’s or field engineer’s worries.
Nevertheless, emissions problems sooner or later may show-up, but since coupling mechanisms are reciprocal, the author has taken the choice of always reviewing susceptibility situations first, because once understood, the comprehension of emission mechanisms would follow easily.
The 3rd article treated the very common mechanism of Common Impedance Coupling, one that probably ranges at the top of the list for causing EMI problems. The present, 4th article is reviewing two coupling paths where the culprit and victim circuits are not in physical contact and the parasitic effect occurs through radio propagation or near-induction coupling (Crosstalk).
1. Field-to-cable Coupling
Although neither considered as antennas or designed for, cables and printed circuit traces are unintentional antennas. Any piece of conductor carrying a current (or excited by a voltage) will radiate an electromagnetic field. Reciprocally, any piece of conductor illuminated by an ambient electromagnetic field will exhibit a current flow and a related voltage. Such unintentional, antennas can be broken-down in two simple forms: closed loops or open-ended wires.
A loop is intuitively regarded as magnetic field (H-field) sensitive, but in fact it responds to both E and H fields, with a predominant ability to privilege one or the other depending on its orientation versus the field vectors and propagation. Open-ended wires are generally regarded as E-field sensitive, yet both shapes can pick-up signals from an electromagnetic field, expressed in Volts/m.
Most often, real-life circuits and cabling configurations are neither purely open wires or perfect loops, but somewhere in-between. Circuits that are terminated into high impedances (> 377Ω) tend to behave as dipoles or whip antennas, while those terminated in low impedances (< 377Ω) tend to behave as loops.
Three factors are playing a role in the efficiency of these fortuitous antennas to pick-up an ambient, time-varying field (alternating or simply pulsed):
- the dimensions of the loop or wire
- the frequency (or rise time for a pulse) of the radiated threat
- the amplitude of the field
One good measure of the ability of the unintentional antenna to convert Volts/m into Volts is by comparing its physical length with the incident field wavelength « λ ». Given that the relationship of λ to frequency is:
λ (m) = 300/F(MHz)
we know that a wire (or loop) whose dimension is reaching λ/2 (λ/4 for a whip) has its maximum ability for field-capture. But even below this resonance, any conductor of length « ℓ » has a pick-up efficiency proportional to its 2ℓ/λ ratio. At the frequency of 10MHz, that is λ = 30m, a wire whose length is only 1 meter starts being a fairly able antenna since it represents 1/15th of a perfectly tuned dipole.
Few examples of frequently found incidental antennas are shown on Fig. 2. Notice that in (a), the loop is NOT the intentional path in the link. The cable can be a bundle of many signal or power wires, including their respective return (or «ground») conductors. The voltage induced in the loop by the incident field is a Common Mode (CM) voltage appearing in the loop, which tends to push a current in the same direction in all the wires of the pair or bundle.
We have already seen such kind of parasitic voltage developing in Common Impedance Coupling (CIC) of Article #3. How much of it actually reaches a sensitive circuit at either end of the link depends on the way the receiving circuits are balanced (symmetrical) or not, or if the cable is shielded or not. A line-to-ground CM voltage, if high enough, like hundreds or thousands of volts can also simply destroy sensitive, unprotected circuits.
It is interesting to note that this voltage relates to the flux linkage in the physical area (m2) of the loop, REGARDLESS the loop is electrically closed or not. Simply, if the loop is open, like floating one of the 0V ref of the electronic circuit, in theory no current could flow in the cable. But we have already seen (Art. #3) that such solution, called star or single-point grounding progressively loses its efficiency when frequency increases, because of the parasitic capacitances to ground in the equipments. (Se left-hand box, Fig 2.a)
In captions (b) or (c) the Differential Mode (DM) loop is of smaller size, like two wires of a same pair, or PCB traces, forming a loop with a few tens cm2 area. But this is a differential scheme, where the picked-up voltage can directly upset the analog or digital ICs attached to the pair.
Three simple formulas give a good estimate of the worst-case loopinduced voltage Vi:
a) for F(MHz) < 100 / ℓ (m): Vi = ℓ.h(m2) x E(V/m) x F(MHz)/50
b) for F(MHz) ≥ 100 / ℓ (m): Vi = 2h(m) x E(V/m), independant of F and ℓ
c) for F(MHz) > 100 /h(m): Vi ≈ 120/F(MHz)
where ℓ and h are the loop length en height
Fig. 3 shows field-to-loop coupling factor for a few simple sizes, along with the E-field pick-up ability for a typical « whip » antenna configuration, where the base of the antenna can be a grounded equipment while the I/O cable is connecting a small, totally isolated access-control device, surveillance camera or alarm device, grounded to nowhere.
What is the voltage induced by a 3V/m ambient at 100MHz in a 1.50m cable loop 0.80m above ground? For this geometry,
Eq. a) applies up to F = 100/1.50 = 66MHz
Eq. b) applies from F ≥ 66MHz up to F = 100/0.80 = 125MHz
Eq. c) applies for F ≥ 125MHz
100MHz is > 66, thus Eq. B applies: Vi = 2h(m) x E(V/m) = 3V/m x 2 x 0.80m = 4.8V
2. What ambient E-M fields can be a serious threat for modern electronics?
If we let aside radio, TV and other RF receivers where jamming can occur at very low levels (1mV/m or less) but this implies an illicit source tuned on the same radio station, most consumer / industrial or medical products are tested to withstand at least 1 to 3V/m. Automobile or airborne electronics are designed to tolerate 10 to 30 times more. Consequences of a field exposure may also depends on:
a) its frequency; most critical frequency range is in the tens to thousands MHz, since related wavelentghs are 30m down to 0.30m respectively. Thus cables as short as a meter are already efficient pick-up antennas.
b) its quasi permanent, or transient nature : A 50kA lightning strike at 30m distance will cause a 100kV/m field, that is a tremendously high value. But if the installation is reasonably protected, the consequence may simply be a temporary upset, that can be self-recovered, or at worst require a reset or re-power-up. On the opposite if a constantly occuring RF field is causing errors, the system will be down all the time.
Table 1 gives an idea of common radiated exposures, with their frequency-distance variables.
3. What are the solutions against field-to-cable coupling?
Assuming the radiating source is beyond our control, there are many solutions to reduce its effects. We can reduce the dimension of the capture area, shield the conductors or circuits that behave as receiving antennas, or filter the RF currents before they reach the victim circuits inside the equipment. As we did for the former coupling path (Article #3 Common Impedance Coupling), next are several simple rules for attenuating the effects of radiation coupling.
Notice that these rules are not countradictory with those for conduction coupling. In fact, they are both complementary, since conduction problems manifest most at freqencies below tens of MHz where radiatiion coupling is rather unefficient, and vice-versa. This is why these rule are numbered in sequence after those of Article #3.
Rule 6: Reduce by all means the geometrical dimensions of the capture areas existing wire-to-wire (or trace-totrace) and cable-to-ground.
– For installations, try to run the external cables as close as possible to a metallic structure or reinforced concrete slab ; the steel grids or rebars make a decent HF reference, reducing the induced CM voltage.
– For wire pairs, keep close to each other the positive and return wires of a same signal. Twist them together. For PCBs, keep the positive and return traces close (side-by-side) or on top of each other. It reduce induced Diff.M voltage
Rule 7: If you cannot bring the conductor down to a ground plane, bring a ground plane up to the conductor.
– For installations, run the cables in solid or perforated metal raceways, making sure all elements are continuously bonded / screwed from endto-end and to the equipments enclosure. A large “U” shaped raceway enveloping the cables acts as a substitute to ground plane.
– For installations with no metal ground plane accessible, install a large (at least 2 m2) metal plate underneath the equipment cabinets, extending generously beyond the equipment footprint.
– For PCBs, use multilayers with plain (no slots) ground planes
Rule 8.a: Use shielded cables with metallic connectors making a tight, peripheral contact with the shielded jacket.
Install them such as the shield make a good, direct contact at the penetration in the metal enclosure of the equipment (if there is any …). If the equipment has no metallic (or metallized plastic) enclosure, try to ground the I/O cables connectors to the PCB ground plane, as close as possible to the point of entry.
Rule 8.b: Cables shields must be connected directly to the equipment at both ends.
Stay away from old, die-hard myths that say «never ground a shield both ends, because this creates ground loops». Open-ended shields do not provide an alternate path for cable-loop current, therefore there would be no change in the CM current flow on your signal link. The only case where grounding a cable shield both ends could be detrimental is when a large, low frequency (like 50 or 400Hz) ground shift exist between two equipments on a site. But in this case, the shield with one end floated will remain worthless against field coupling into the cable loop at any frequency. Only low frequency capacitive (E-field) coupling will be attenuated.
Rule 9: If none of rules 6, 7, 8, 9 is feasible or controllable at design stage, let the antenna (capture loop) exist and block the RF currents by EMI filters right at the I/O ports.
Cable-to-cable Coupling, or Crosstalk?
The term Crosstalk, inherited from the early days of telephony, describes a situation where during a conversation between two subscribers, one could hear the talk of a third party. Technically speaking, Crosstalk is the mechanism by which two wires, belonging to different circuits and running in parallel can couple tranversally: a certain % of the « culprit » wire voltage appearing on the « victim » wire.
Since the two circuits involved are not in physical contact, one could think of a radiation coupling. However there is a difference: with radiation, the souce circuit is emanating a field in a 3-D space, where it can be received by anybody. With Crosstalk, the coupling occurs in the limited spacing by the wire-to-wire capacitance or mutual inductance; no RF propagation is involved.
Thus, as shown on Fig.4, Crosstalk happens because of the distributed capacitance C1-2 (Capacitive Xtalk) or mutual inductance M1-2 (Magnetic Xtalk) that exists between two conductors belonging to different circuits. This coupling can take place between two traces on a PCB or ribbon cable, between 2 pairs (Diff. mode Xtalk) or between 2 cable bundles above a ground plane (C. Mode Xtalk).
Crosstalk is governed by several parameters belonging to the circuits configuration. It aggravates when:
a) Ratio h/s increases
b) Both conductors are embedded in a dielectric (more capacitive Xtalk)
c) Victim impedance is high (more capacitive Xtalk)
d) Culprit impedance is low (more primary current, hence more magnetic Xtalk)
e) The parallel run of victim/culprit wires is longer
Crosstalk is defined as the ratio of the victim-induced voltage to the culprit voltage:
Xtalk = Vvictim/Vculprit, that is a Volt/Volt, dimensionless number.
It can be expressed in frequency domain, for sinewave signals, or in time domain for pulsed signals. The frequency diagram Fig. 5 shows that Xtalk never exceed or even reach a 100% ratio (that is 0dB): victim’s induced voltage cannot equal or exceed the culprit voltage.
When expressed in frequency domain (sinewave situation like a single radio frequency or a discrete harmonic of a digital signal):
Xtalk coefft (Capacitive) = Rv / (Rvict + 1/C1-2ω) ≈ RvC1-2ω if Rv < 1/C1-2ω
Xtalk coefft (Magnetic) = M1-2ω / Rculp if M1-2ω < Rculp
with, C1-2, M1-2 = wire-to-wire coupling capacitance and mutual inductance
Rvict= parallel combination of victim near-end and far-end resistances
Rcullp = culprit circuit resistance
ω = 2 π F
Notice that the capacitive Xtalk is governed by the victim’s resistance, while magnetic Xtalk is driven by the culprit’s load, since this latter dictates the current in the primary loop. When conductors are embedded in a dielectric, like discrete wires or PCB traces, capacitive Xtalk is increased by the dielectric constant (ԑr = 3 – 3.5 for PVC, 4.5 for epoxy). Magnetic Xtalk is not affected.
In time-domain representation (when the rise time Δt of culprit signal is known):
Xtalk coefft (Capacitive) = Rv C1-2 / Δt if Rv C1-2 < Δt
Xtalk coefft (Magnetic) = (M1-2 / Rculp) /Δt if (M1-2 / Rculp) < Δt
Unaware designers tend to regard Crosstalk as a minor, “oh, by the way” nuisance that will be taken care of if it happens. The following numerical example gives some measure of this insidious coupling that can cause digital upset and eat-up the EMC margin. It can also export internal circuits noise to external cables that will in turn cause violations of radiated EMI limits.
Examples of crosstalk
Solutions against Crosstalk
Essentially, all solutions that work against field-to-cable coupling are efficient against Crosstalk. Furthermore, besides Rules #6, 7, 8, 9 seen before, a few specific solutions can be applied:
Rule 10: against capacitive or magn. Xtalk, increase the culprit-victim wires separation.
Generally, try to increase the s/h ratio, since the mutual capacitances and inductances fall-off rapidly when s/h ratio becomes > 1 . A s/h ratio ≥ 10 is a guarantee that Xtalk will never exceed 1% (-40dB) at any frequency.
Rule 11: against Xtalk in PCB s, use the « poor man’s » shield, a simple ground trace between culprit and victim traces.
This simple, unexpensive precaution easily reduces both Xtalks by 20dB (a x10 factor)
Needless to say, all the coupling mechanisms seen above and their reduction techniques are perfectly reciprocal, in that they will prevent radiated emissions as well. This aspect will be reviewed more in detail in other EE Magazine articles to come.
Michel Mardiguian EMC Consultant, France firstname.lastname@example.org