In the previous article, we explained the design principles of differential pairs. In this article, which is FC34, we will cover crosstalk. In differential pairs, we intentionally utilized electromagnetic coupling between traces, but now we must suppress it between "unintended" traces. We will explain the physical mechanisms of crosstalk and specific suppression techniques in FC design.
What is Crosstalk?
Crosstalk is a phenomenon where unintended signal coupling occurs between adjacent traces. A signal flowing through one trace (the Aggressor) is induced as electrical noise in the other trace (the Victim).
In the previous differential pair discussion, we intentionally used the coupling between P and N lines to achieve noise cancellation. However, crosstalk is coupling that occurs between traces not intended by the designer (for example, adjacent traces of different signal systems) and is treated as a factor that degrades signal quality.
Crosstalk has two mechanisms: electrostatic coupling (capacitive coupling) and electromagnetic inductive coupling (inductive coupling), each with different physical origins.
Capacitive Coupling
Capacitive crosstalk is coupling that occurs through parasitic capacitance formed between adjacent traces.
The voltage change dV/dt of the aggressor line induces current in the victim line through the parasitic capacitance Cm between the traces:
I_induced = Cm × dV/dt
This induced current is converted into voltage noise through the impedance of the victim line.
Capacitive coupling is coupling via an electric field and is caused by the potential difference between traces. The closer the distance between traces and the longer the length they run in parallel, the greater the parasitic capacitance Cm becomes, and the stronger the crosstalk.
Capacitive crosstalk tends to appear in a form where the aggressor line and the victim line share the same direction of voltage change (noise of the same polarity at both ends).
Inductive Coupling
Inductive crosstalk is coupling that occurs through mutual inductance between adjacent traces.
The current change dI/dt of the aggressor line induces voltage in the victim line through the mutual inductance M:
V_induced = M × dI/dt
This induction is coupling via a magnetic field and is caused by the current loop (signal path and return path) flowing through the traces.
Inductive crosstalk can exhibit different noise patterns than capacitive crosstalk, sometimes appearing as noise of different polarities at the near end and far end of the trace (the difference between NEXT/FEXT described later).
On an actual PCB, both capacitive and inductive coupling exist simultaneously, and their relative strength is determined by the trace geometry (trace width, spacing, and distance to the reference plane). In common PCB materials like FR-4, capacitive coupling tends to be relatively stronger when trace spacing is narrow, while inductive coupling tends to be relatively stronger when traces are far from the reference plane.
NEXT (Near-End Crosstalk) and FEXT (Far-End Crosstalk)
Crosstalk is classified into Near-End Crosstalk (NEXT) and Far-End Crosstalk (FEXT) based on the location where the noise appears on the victim line.
NEXT is noise observed on the side of the victim line close to the aggressor line's signal source (the near end of the victim line). As the aggressor signal propagates along the trace, both capacitive and inductive coupling generate noise components directed toward the signal source.
FEXT is noise observed on the side of the victim line far from the aggressor line's signal source (the far end of the victim line, i.e., the receiving end). FEXT has the property of appearing as the difference between capacitive and inductive coupling, and in cases where both are perfectly balanced (homogeneous coupled lines), FEXT can theoretically be zero.
In homogeneous coupled lines (homogeneous medium), capacitive and inductive coupling have a nearly equal balance, so FEXT is small and NEXT becomes dominant. On the other hand, in non-homogeneous coupling environments such as striplines (described later), FEXT can appear more prominently.
In practical FC design, NEXT is more commonly an issue. In particular, NEXT tends to accumulate easily in long parallel traces (such as traces routed to connectors).
Design parameters affecting crosstalk
We will organize the main design parameters that determine the strength of crosstalk.
Trace spacing is the parameter with the greatest impact. The wider the spacing, the more rapidly crosstalk decreases. As a rule of thumb, it is considered that ensuring a spacing of at least three times the trace width (the 3W rule) can suppress crosstalk to a level that is not a practical problem.
Parallel run length is the distance over which crosstalk accumulates. The longer the parallel section, the stronger the coupling. In the trace routing design stage, a layout that avoids unnecessarily long parallel runs is effective.
The distance to the reference plane (trace height) affects how electric and magnetic fields spread. The closer the trace is to the reference plane (the thinner the dielectric), the more strongly the electromagnetic field is confined between the trace and the plane, reducing coupling to adjacent traces. This means that the stack-up design in FC31 and the impedance design in FC39 are also directly related to crosstalk suppression.
Signal rise time directly affects the strength of crosstalk. Signals with larger dV/dt or dI/dt (steeper edges) induce stronger crosstalk. The same principle as the trade-off between MOSFET switching speed and EMI explained in ESC11–ESC17 also applies to signal traces on a PCB.
Practice of the 3W rule
We will organize the practical meaning of the 3W rule, which states to "ensure a trace spacing of at least three times the trace width."
The origin of the 3W rule is the empirical observation that the spread of the electromagnetic field between traces is concentrated within a range approximately equal to the trace width. By taking a spacing of three times the trace width, the influence of the electromagnetic field on adjacent traces is significantly attenuated (an empirical rule stating that crosstalk is reduced by about 70%).
In more rigorous designs, the 5W rule (spacing five times the trace width) is also used, which can be expected to reduce crosstalk by a few more percent, but the trade-off with board area becomes greater.
As for the priority of applying the 3W rule, it is practical to apply it intensively to combinations of crosstalk-sensitive signals (analog signals, low-level signals, clock signals) and noisy signals (DC-DC switching-related traces, high-speed digital signals) rather than applying it equally to all traces.
Guard Trace
A proactive measure to suppress crosstalk is the guard trace.
A guard trace is a technique of adding a trace connected to GND (or VCC) between two signal traces where crosstalk is a concern. This guard trace shields the electromagnetic field to some extent and reduces coupling between the signals on both sides.
The effect of a guard trace is maximized when the guard trace is connected to the GND plane with vias at a sufficient frequency. If the guard trace itself is in a floating state, it does not provide a sufficient shielding effect. The principle of via stitching explained in FC25 applies here as well.
As a limitation of guard traces, if basic spacing such as the 3W rule is already achieved, the additional effect of guard traces may be limited. When board space is tight, it is necessary to evaluate whether increasing spacing or adding guard traces is more efficient.
Guard traces are sometimes used around analog sensor signals (such as IMU analog outputs or barometer analog interfaces) to protect them from adjacent digital signals.
Crosstalk countermeasures around sensors
We will organize the areas where crosstalk is particularly problematic in FC design.
Around the IMU (see FC6–FC8), SPI clock signals (high-speed, steep edges) can cause crosstalk in adjacent analog or low-speed control signals. Ensure the spacing between SPI clock lines and other signal lines is at least the 3W rule, and add guard traces if necessary.
GNSS antenna feed lines (see previous series) handle extremely weak signals (-130 dBm) and are therefore very sensitive to crosstalk from adjacent high-speed digital signals. It is important to design with the widest possible keep-out zone around the antenna feed line and avoid running it parallel to other signal traces.
Around the compass (magnetic sensor) (see FC9), in addition to electrical crosstalk, there is also magnetic interference (a mechanism different from crosstalk, which was covered in the previous series' compass article), so consideration from both perspectives is necessary.
Crosstalk simulation and measurement
We will organize the design verification methods for crosstalk.
Simulation using a field solver is a specialized tool for calculating the strength of crosstalk (coupling coefficient, magnitude of NEXT/FEXT) from the cross-sectional structure of the wiring (width, spacing, dielectric thickness). Tools such as Saturn PCB Toolkit and Polar Si9000 support this type of calculation, and the same toolset used for impedance calculation covered in FC39 can be utilized.
Actual measurement using TDR (Time Domain Reflectometry) is a measurement technique for evaluating transmission line characteristics of high-speed signals and can also be applied to measuring crosstalk. It requires a dedicated high-bandwidth oscilloscope and TDR module, and while it may be omitted in the standard FC design workflow, it is worth considering for industrial and high-reliability designs.
As a practical evaluation method, many FC designs take the approach of adhering to rule-of-thumb-based design guidelines like the 3W rule rather than strict simulation, and confirming final performance (sensor noise floor, communication error rate) through actual hardware evaluation. EKF innovation statistics explained in FC14 and the GNSS CN0 evaluation in FC11 serve as indicators that indirectly reflect the effectiveness of crosstalk countermeasures.
Crosstalk and DRC
The limitations of DRC explained in FC30 also apply to crosstalk.
Standard DRC can check the minimum value of wiring spacing (from a manufacturability perspective), but whether that spacing is 'sufficient from a crosstalk perspective' (whether it follows the 3W rule) is not automatically checked unless defined separately as a custom rule.
In tools like KiCad, you can partially automate constraints equivalent to the 3W rule by setting the minimum spacing according to the trace width as a custom rule. This is a concrete practical example of 'expanding the parts that can be automated' discussed in FC30.
However, context-dependent judgments such as parallel run length and signal importance (analog vs. digital) remain areas that require review by the designer.
Summary
Crosstalk occurs through two mechanisms: capacitive coupling (electric field) and inductive coupling (magnetic field), and appears on the victim line as NEXT (near-end) and FEXT (far-end). Wiring spacing (3W rule), minimization of parallel run length, and management of distance to the reference plane are the primary suppression means, and guard traces function as additional protection for particularly sensitive signals. Strict crosstalk countermeasures are especially necessary around high-sensitivity sensors in FC design, such as GNSS antenna feed lines and the IMU area. Verification combining quantitative evaluation with simulation tools and sensor performance evaluation on actual hardware is a practical approach. In the next issue, FC35, we will explain power plane design—the Power Distribution Network (PDN) and the suppression of plane resonance.