Estimated reading time: 5 minutes

The Pulse: Caught in the Crosshatch—A Cautionary Tale of Detective Work

A chance meeting at a family wedding the other week led to a conversation about numbers, an introduction to a book entitled Humble Pi, and how numeric misinterpretation can lead to all kinds of unexpected outcomes, some just costly, others tragic. It’s a good and amusing read, and as a result of this conversation with someone I had previously never met, I feel somewhat (at least temporarily) enlightened. One of the takeaways of the book is that humans are born to think logarithmically, and linear math has to be formally educated into our brains. That got me curious for more.

Back in the World of Work
The Monday morning after that chance meeting, an email landed in my inbox from a colleague in our Singapore office: “One of our OEM customers is getting very different results when compared with his supplier, despite them both using identical Polar software. What could cause that?” As is often the case, customer confidentiality meant that limited information was available, and all I had to work with was a somewhat redacted cell phone picture of the two software instances side by side in an Excel sheet (Figure 1).

These scenarios usually lead to a degree of detective work to isolate the root cause of the discrepancy. So, I set about manually entering the data and checking on my own version of the field solver. The scenario involved modelling and using crosshatched (mesh) ground planes. I thought it would be good to revisit the rationale for using mesh planes and the dimensional (and other) considerations when deploying them.

Crosshatch Planes Revisited
From a signal integrity purist's point of view, crosshatched planes are non-ideal: A full copper plane provides a consistent return path. However, PCB engineering often involves a compromise between the mechanical and electrical worlds, especially with flex circuits. Meshing/crosshatching is extremely useful on flex as it keeps the flex, well, flexible. It’s no use having perfect signal integrity if the flex becomes brittle and fails owing to forcing solid copper plains in a stripline to act as an “I” beam, making the flex “inflexible.” Here, the use of crosshatch is vital if the flex circuit is to survive its design lifetime of flexing. This is clearly more important in flex, such as with hinged and moving applications, rather than flex-to-fit, which may only be flexed a few times during assembly and repair.

Crosshatch is also useful on thinner stackups, flexible or not, to achieve a given impedance whilst keeping the line width manufacturable. A meshed return plane will require a wider trace for a given impedance. This can be helpful when a fabricator is bouncing on the lower limits of process capability.

Considerations
It’s important to keep the mesh as small as possible. You can make large apertures in the mesh, but this will lead to EMI problems and, in extreme cases, cause impedance to vary along the line. Think of fiber weave effects on steroids. But if kept small in relation to the wavelength, the crosshatch is a viable and practical way to keep things flexible.

Single-ended and differential?: Interestingly, crosshatch has a greater effect on single-ended transmission lines than differential because in differential pairs (depending on the coupling percentage), much of the return current flows in the complementary side of the pair, so the closer the coupling, the less the effect of the hatch.

Alignment: Often, you will see that guidance is to align over the hatch at 45 degrees and over the center of the crosses. However, this is not always possible. Often, the hatch is specified to the fabricator and added at the CAM stage rather than in CAD. It’s best to know that much of the crosshatching in flex is added at the fabrication stage—that might be changing, and if you think otherwise, let me know. At Polar, we get requests to add crosshatch to flex impedance coupons in the CGen coupon generator, but for that reason, we haven’t done it, as it might mean the hatch on the coupon hatch is created differently from the hatch under the circuit itself.

How Did the Variation Arise?
I started by noting that a customer reported different impedance results from the same modelled data than from their supplier, so I took a long, hard look to see whether units were mixed or there had been some form of input error. But as hard as I tried, I could not replicate the discrepancy. My numbers only aligned with one of the predictions sent to me. Then, as with all detective work, a minor point struck me that I had missed on first look: One of the screenshots had a “+” and a “–” in the top left of the structure image (Figure 3).

The other image from the OEM’s supplier did not contain that control. A  look in the archives showed me that this change in our software dated one version at a release (at least 11 years older than the latest, maybe 15), which was in the early days of crosshatch modeling for Polar. Our tools are regularly updated and fine-tuned, and so I find myself now in a further search to see what has changed and what might have improved since those early days. A lesson, perhaps, in making sure that software is regularly updated, especially when working on critical designs.

So, as with many investigations, the original supposition turned out to be wrong, but it did point to further analysis and triggered the search for the cause of the difference.

Martyn is the author of The Printed Circuit Designer’s Guide to… Secrets of High-Speed PCBs, Parts 1 and 2, and The Printed Circuit Designer’s Guide to…More Secrets of High-speed PCBs.

This column originally appeared in the April 2026 issue of I-Connect007 Magazine.