Consider a system like:
Black: basic diagram. Suppose there's...
Box 1 has earth connection, and a grounded, shielded output.
Box 2 is ungrounded/floating, but don't worry, we've run a separate earth connection to the cable to "terminate" that end of the shield. (Or left it completely hanging. Homework: how bad is each, or which is worse?). It has a grounded shielded output.
Box 3 has a grounded, shielded input, and earth connection.
All earths collect via facility wiring back to the breaker panel and nearest ground rod.
Red: suppose there's a magnetic field within the general enclosed area of this loop. What happens? Namely, what current flows in the loop, what are the loop characteristics (R, L, transmission line / antenna element length, etc.), what voltage drops are distributed where around it, etc.
Box 2 inset, red highlighted: the "extra GND" does absolutely nothing at radio frequencies, the return path is spaghetti, at least several µH but 10s or even 100s might happen in a tall wood-frame building.
Most of the loop voltage drops across the input, adding directly in series with the signal received from Box 1. Extreme noise has entered the system.
Note that, even if we're using differential signaling, we still have to worry about the common-mode voltage. No differential receiver is truly perfect: even transformers have limited CMRR (varies with frequency) and breakdown voltage, and most (semiconductor) receivers have a limited range, for example 10-30V for RS-422/RS-485/CAN, between supplies for most others (instrumentation amps, general purpose diff amps, LVDS receivers, etc.). So the CMR (CM range) might be hardly a volt (e.g. 3.3V supplies, biased in the middle) in a very typical (low voltage analog or digital receiver) case.
And, what we need to protect against, ranges from a couple volts CW RF (commercial immunity test levels), up to 10kV transients (ESD), give or take how much we need to withstand those transients -- maybe we don't mind if the signal gets momentarily corrupted, or even if the equipment gets disrupted (requires power cycle), but high-reliability applications might even mind the additive noise.
Box 3, purple highlighted: whatever loop the input cable is a part of (actually none as shown, due to the above gap, but we can imagine an "extra GND" applied in the same way at the Box 2 output if needed), its current is shunted to ground around the circuit, and noise voltage is greatly reduced. How much it's reduced, depends on the impedance of that shield-to-ground link. Down at audio frequencies, we might be concerned that nothing is inductive, i.e. we don't obtain coupling value from the cables, they don't behave as transmission lines, or transformers*, and thus the resistance of those paths (including the cable shield itself) all have some voltage drop that manifests as a V_N in series with the input signal.
*To a certain degree, transformers are transmission lines; certainly, they can be built directly using them. The space of possible (practical or otherwise) transformers is much wider than that of practical transmission lines, I would say, and so too the range of possible response or behavior; especially in terms of nonlinearity, when core materials are involved in various tricky ways. But we can restrict our consideration to the set of better-behaved transformers, that are transmission-line-ey, without much loss of generality for signal and EMC analysis purposes.
The most important thing about earthing, for EMC purposes, is, if the lead length is long, its effect is
negligible, and for example, breaking the connection between cable shield and circuit reference plane or enclosure, introduces the full amplitude of whatever madness is going on outside that closed system.
Put another way, the shield is an extension of the enclosure around the circuit, and presumably, the circuit is bonded to that enclosure to make best use of it (lest the circuit's internal current loops and voltage drops corrupt the signals exiting different points on the board). Topologically, it doesn't matter what the shape of the enclosure is, just as long as it's conductive and thick enough to enclose and contain fields within. This includes the act of opening holes into the enclosure, to a limited extent: we can treat an opening as a radiation port (port: a datum through which energy flows in and out; usually a pointlike node, such as a connector, or a cross-section on a cable), and rather than having total isolation as in the ideal case, we simply have a degraded but still largely satisfactory case (maybe the shielding effectiveness is min 40dB over frequencies of interest, but it's still fine given what the circuit does and what levels it's tested at). We can go further and open whole sides of the enclosure, exposing the PCB ground plane as its own enclosure wall; as long as signals and components are arranged at only short distances away from that surface, we get microstrip trace geometry with reasonable performance (for commercial levels). This is a transformational way to understand how a PCB can be useful by itself without additional shielding.
Tim
Home
Help
Search
Login
Register
