EMC Problems in PCB Design: A Diagnosis and Fix Workflow

Electromagnetic compatibility problems are rarely solved by adding components at the end of a project. They are solved by identifying which of a small number of physical mechanisms is responsible, and then changing the geometry that produces it. The diagnosis is systematic, and the fix is usually cheaper than the component that was tried first.

Five Attributes Worth Establishing First

Before changing anything, establish five facts about the problem. The first is the physical size of the radiator, because an emitting structure whose dimensions approach a fraction of a wavelength is far more efficient than a small one, and the length of a trace is part of that geometry.

The second is the impedance relationship between the source, the path and the receiver. The third is the time behaviour of the disturbance: is it continuous and periodic, or does it occur only during a particular event such as a power-up, a key press or a network burst. The fourth is the strength of the source relative to the limit. The fifth is the frequency content, which is established with a spectrum analyser rather than by intuition, because the position of the problem in the spectrum identifies the mechanism.

Why Grounding Habits Do Not Transfer

Single-point grounding is an excellent technique for low-frequency circuits, because it prevents return currents from sharing conductors and creating common-impedance coupling. Applied to a radio-frequency circuit, the same technique fails, because at high frequency the return current does not follow the geometric path; it follows the path of least impedance, which is the one directly beneath the trace.

Designers who apply a low-frequency habit to every board encounter EMC problems that seem inexplicable, because the rule they are following is correct in a different frequency regime. The useful question is not which grounding style is better but which return path the current will actually take at the frequency of the disturbance.

Return Currents and Loop Area

Current flows in a closed loop, and the area enclosed by that loop determines how efficiently the circuit radiates and how well it receives. The return current takes the path of least impedance, which for a signal over a plane is directly underneath the trace. Anything that interrupts that path, such as a split in the plane or a via that forces the current to detour, enlarges the loop and worsens the problem.

Where the impedance between the source and the load is high, the return current may have several possible paths, and all of them contribute. That is why a supply net routed as a long thin trace is a worse radiator than the same net routed over a plane: the trace and its return enclose a large area, and the supply current contains the switching content of everything it feeds.

When a Trace Stops Being a Trace

The impedance of a conductor includes resistance and inductive reactance, and above roughly a hundred kilohertz the inductive term dominates. A trace that was a connection at low frequency becomes an inductor, and a trace that carries audio-frequency or higher content can behave as an antenna.

The practical expression of this is the rule that a conductor should not be a significant fraction of a wavelength at the frequency of interest, because a conductor approaching a quarter or a half wavelength is an efficient radiator. A trace that accidentally reaches that length turns a trace into an antenna, and the resulting emission is difficult to diagnose because it appears to originate from the circuit rather than from the layout.

Layout Decisions That Decide the Result

Board dimensions are the first decision. A board that is too large lengthens every trace, which reduces the immunity of the system and increases the cost; a board that is too small concentrates heat and encourages coupling between functions. Neither extreme is a good answer, and the useful question is what the routing actually requires.

Placement of the clock sources comes next. A clock trace should not run beneath a plane that carries a critical signal, and it benefits from being kept away from the board edge and from connectors. The general principle is that components belonging to one function should be placed near each other, so that the loop formed by the current in that function is as small as the circuit allows.

A Workflow for the Fix

The workflow that produces results is to measure the emission, identify the frequency, and then decide whether the source is a clock, a switching supply, a data interface or an external coupling path. Once the source is known, the geometry that couples it can be addressed: the loop that carries the current can be shrunk, the reference under the trace can be restored, a filter can be placed at the connector rather than at the device, or the offending net can be moved away from the structure that radiates.

Adding a component without identifying the mechanism leaves the problem in place and consumes the design margin that the fix would have needed. Our EMI immunity design notes describe the coupling mechanisms, the high-speed design rules explain the return path requirements, and the design release checklist covers the review that should happen before the board is built.

Distinguishing Conducted From Radiated Paths

Two boards can show the same emission on a test report for entirely different reasons. A conducted path travels along a cable or a power connection, so it responds to filtering and to the impedance presented at the connector. A radiated path leaves the board through the air or through a slot in the enclosure, so it responds to the geometry of the loop that produces it rather than to anything placed in the circuit.

The distinction is established by measurement rather than inspection. Disconnecting a cable, wrapping it in an absorbing material or temporarily enclosing the board changes the result for a conducted path and leaves a radiated one largely unaffected. Making that determination first prevents the common mistake of adding filters to a problem that is leaving the board as a field, where no amount of filtering will help.

FAQ

Should the board be made smaller to reduce emissions? Not automatically. A smaller board shortens the traces, which helps, but it also brings the noisy and sensitive sections closer together and reduces the area available to spread heat. The useful approach is to size the board to the routing requirement, then organise the layout so that each functional block has its own region and its own return path.

Why does a problem appear only during a particular operation? Because the disturbance is an event rather than a continuous signal. A switching event, a motor start, a network burst or a display refresh produces a broadband disturbance that is absent at other times, and a measurement taken during idle conditions will not reveal it. Recording the emission while the product performs the operation that causes the problem is what makes the diagnosis possible.

What does gopcb review for EMC? We review the stack-up and the return paths, the placement of the clock and switching sources relative to the connectors, the separation between functional blocks and the treatment of the connector ground. Those are the structural decisions that determine the emission behaviour, and they are all made before the board is fabricated rather than corrected afterwards.