Core PCB Design Issues and How to Resolve Them

Board failures rarely come from a clever circuit that did not work. They come from a handful of recurring issues that appear in almost every design: a sensitive part placed next to a heat source, a plane that turned into a capacitor, a via chosen without thinking about its effect on the layer beneath, or a design rule check that was run once and then ignored.

This article sets out those issues in the order a PCB design meets them, explains why each one matters, and describes the practical response. The goal is not a set of rules to memorise but a way of seeing a layout that makes the problems visible before the first panel is built.

Placement: the Decision Everything Else Follows

Placement determines the routing, the thermal behaviour and the electromagnetic performance of a board at the same time, which is why it deserves more time than it usually receives. Components that form one function belong together, but that grouping has to be balanced against the need to keep heat sources away from temperature-sensitive parts and noise sources away from quiet ones.

<img src="https://www.gopcba.com/wp-content/uploads/2020/12/ptt_history.jpg" alt="Ground plane and signal layer relationship in a stack-up” />

A converter that raises the local temperature by twenty degrees will change the behaviour of anything within a few centimetres of it, and it will do so before any electrical problem appears. Placing two converters that generate different noise on opposite sides of the board is a simple measure that reduces both the thermal load and the mutual interference, and it costs nothing except the time to decide it.

Heat and Noise Sources on the Same Board

Temperature-sensitive devices include sensors, references and any part whose accuracy is specified at a particular ambient. Heat-generating devices include power converters, linear regulators, power transistors and anything carrying a significant current. The layout should treat the two as opposing forces and keep them apart, with the sensitive parts placed where the airflow is coolest.

Noise sources overlap heavily with the heat sources, which is convenient because the same separation helps both. A switching supply is simultaneously the hottest and the electrically noisiest device on most boards. Giving it its own area, with its input capacitor and its return path confined to that area, addresses thermal spreading and electromagnetic interference with the same piece of copper.

Stray Capacitance Between Neighbouring Features

Two conductors separated by an insulator form a capacitor, and a layout contains thousands of them whether or not anyone intended it. Two traces running parallel on adjacent layers are the usual case, and the effect grows quickly as the layers get closer. The consequences are a change in the propagation velocity and an increase in the energy drawn by the trace.

The remedies are geometric. Shortening the distance over which two traces run parallel removes most of the coupling, and where the parallel run cannot be avoided, routing one of the two on a different layer for part of its length breaks it up. A ground plane between the layers shields the coupling but introduces its own capacitance to both. That trade is worth understanding rather than applying a rule blindly.

What a Ground Plane Really Does

A ground plane is often described as a shield and behaves partly as a capacitor. Every trace above it is capacitively coupled to it, which slows the signal slightly and increases the current the driver must supply. The benefit is that the same plane provides a continuous return path directly under the trace, which minimises the loop area and therefore the inductance and the radiation.

The ground routing and power trace planning that goes with the plane is what turns it into a usable reference.

The net effect is almost always positive, but the trade-offs are real. On a two-layer board, a small ground plane may provide a return path without the layer count of a full plane, at the cost of a less complete reference. On a multilayer board, placing the plane close to the signal layer increases the capacitance and reduces the impedance, which is desirable for high-speed nets and irrelevant for slow ones. The choice between solid copper and a mesh follows from the same reasoning.

Vias and the Cost of Changing Layers

A via is a discontinuity, and it is also a structural feature that the fabricator has to produce reliably. Electrically, it adds capacitance to any plane it passes through and inductance in series with the signal. Mechanically, it consumes routing space on every layer it crosses and reduces the strength of the laminate in that region.

The practical discipline is to question every layer change. A signal that changes layer once and then runs to its destination is generally fine. A signal that hops between layers repeatedly accumulates both the capacitance and the inductance, and it does so in a way that is invisible on the schematic. Keeping high-speed nets on one layer from driver to receiver, and using vias only where the path genuinely requires them, removes a whole class of problems.

Design Rule Check and Design for Manufacture

A design rule check is the mechanism that turns a fabricator’s capability into a constraint the layout tool enforces. It flags traces that are closer than the process allows, minimum-width violations, annular ring problems and short loops where two conductors pass too near each other. Its value depends entirely on the rules being set from the real process rather than left at the tool defaults.

Design for manufacture goes one step further and considers how the board will be built, assembled and tested. Panel utilisation, solder mask dams between fine-pitch pads, the clearance needed for the router, and the placement of fiducials for automated assembly all fall into this category. Checking them at the layout stage costs an hour; discovering them at the assembly house costs a re-spin.

Manual Intervention Where Automation Fails

Trace width, spacing and via limits should all come from the fabricator rather than from the tool, as set out in the design guidelines that keep a board manufacturable.

Automatic placement and automatic routing are useful starting points and poor finishing tools. An auto-placer does not know which part is a heat source, and an auto-router does not know which net is sensitive, so the results pass the electrical checks while missing the design intent. The productive approach is to let the tool do the bulk work and then review the high-risk areas by hand.

The areas that deserve manual attention are the ones where several constraints meet: a converter input, a connector breakout, a clock distribution network or an analogue front end. These are small regions of the board that carry most of the risk. Reviewing them carefully, and leaving the automatic result in place elsewhere, is a far better use of engineering time than routing the whole board by hand or accepting the automatic result everywhere.

FAQ

How much separation is needed between a heat source and a sensor? It depends on the thermal resistance between them and on how accurately the sensor must read. As a starting point, keep the two on opposite sides of the board where possible, and use a quick thermal estimate rather than intuition to confirm the placement.

Is a ground plane always worth the extra layer? On a design with fast signals or a sensitive measurement, usually yes, because the return path is the reason rather than the shielding. On a slow control board, a two-layer layout with a carefully routed ground can be perfectly adequate.

Should automatic routing be used at all? Yes, for the low-risk portion of the board. Its output should then be reviewed where constraints interact, and any high-speed or analogue net should be routed deliberately rather than automatically.