An early version of the following article was published by the University of Oxford Dept. of Continuing Education.
A guard trace, or guard track, is a pcb trace that is installed parallel to an existing high-speed signal. Guard traces are usually installed in the hope of reducing crosstalk. Often you will see them installed on both sides of a sensitive signal. Typically, guard traces are grounded at both ends.
Guard traces appear extensively in analog designs. At audio frequencies, on a simple two-layer board, guard traces can reduce crosstalk by an order of magnitude, or more. The digital world acts differently. On a digital pcb that already has solid reference planes, guard traces provide little benefit.
To understand why, consider a long wire dangling in space, carrying a fast-changing digital signal. Somehow, the current associated with that signal must return to its source, so there must be another wire, somewhere, carrying the returning signal current. These two currents flow in opposite directions.
Current on the first wire induces a certain pattern of magnetic fields in space. Current on the return wire does the same thing, with one important difference. Current in the return wire flows opposite to that that in the signal wire, so the field pattern generated by the return wire must be the exact negative of the field pattern for the first wire. If the two wires were co-located in space the two field patterns would precisely cancel with the result that no field energy, and no crosstalk, could possibly emanate from the pair of wires.
A practical application can never co-locate two wires. There always exists some small space between them. From that space flow all of our difficulties with crosstalk.
To see why, it helps to exaggerate the problem. For example, imagine you are standing in a football stadium, on the sidelines, at the 50-yard line. Walk to your left until you arrive at the end zone. Place a signal generator at that point. From the signal generator, run one signal wire down the goal line to the opposite side of the field, 53 yards away. Connect a light bulb to the wire at that spot. It will be the circuit "load". Now connect a return wire to the other side of the bulb, but instead of taking the shortest path back to the source, run the return wire clockwise all the way around the field until you come back to the source, making a rectangular loop that covers the whole field.
Obviously, the huge separation between the signal wire and the return wire wipes out any hope of magnetic field cancellation, so the system radiates prodigiously, creating a lot of crosstalk for nearby victims.
Suppose I install a new, third wire in the system, on the 1-inch yard line, right next to your signal wire. On the far side of the end zone my new wire connects to the return side of the light bulb, and it comes back to the ground terminal of the signal generator, just like your return wire, but it takes a different path. My wire runs parallel to the signal wire, separated from it by a gap of only one inch. The new wire exists in parallel with the original return conductor, soldered at both ends to the same terminals used by the original return.
The new wire provides returning signal current a choice of pathways leading back to the source. The returning signal current can travel along the old wire, straying out 100 yards away from the signal conductor, or it can return along the new wire, keeping very near the outbound signal all along the way, never more than one inch apart.
Given that choice, returning signal current will flow mostly on the new wire. That's a terrific outcome, because the magnetic field associated with the new return conductor now lies very close to the field created by the signal conductor. The closer you place the two field patterns, the more perfectly they cancel, and the less crosstalk they produce. Just by having the second pathway available, current naturally chooses it, and crosstalk drops.
In an old analog layout, circa 1970, if the available paths for returning signal current include some power-supply wiring, a couple of random cable shields, and a pair of guard traces intentionally laid 0.010 in. on either side of the main signal, where will most of the returning signal current will flow? It will return along the guard traces. The guards keep the returning signal current close to the outbound current and that reduces crosstalk.
Now, adjust your thinking to a digital application. Your pcb already has a solid ground-plane layer (maybe more than one). That plane provides every trace an infinite continuum of possible returning current pathways, all packed tight up against the outbound signal trace, mere thousandths of an inch away. That plane is the biggest, fattest, best-looking guard trace you will ever see. It is your best resource for reducing crosstalk.
Unless you can lay out a guard trace closer to the signal trace than the solid reference plane, and make it wider and better than the plane, the guard trace provides almost no benefit.
On a pcb that already has solid reference planes I never use guard traces. I keep my signals close to a solid, continuous reference plane everywhere they go, and I keep in mind that crosstalk between parallel traces changes approximately in proportion to the square of trace height and inversely with the square of separation. When it comes to controlling crosstalk, height and separation are your greatest friends.
I've included some pictures below to help to cement the guard-trace effect in your mind. Each picture shows a trace configurations on the left with one aggressor and one victim. On the right, I show a 2-D cross-section view of the corresponding magnetic field pattern. Crosstalk at the victim varies according to the number of magnetic lines of force that pass under it, and is rated in terms of the dimensionless Near-End Crosstalk coefficient, or "NEXT coefficient".
These pictures are now included in the latest version of my basic course, "High-Speed Digital Design." I'll be teaching that, and also my "Noise and Grounding" class at Oxford and in Canada in June. Hope to see you there.
Dr. Howard Johnson