Well, to be fair, the 0/45/90 deg traces are a byproduct from early days of EDA CAD products. Aside high frequency busses (which you're doing EM analysis on anyways), traces can be all sorts of shapes. That 3v3 or 5v line can have 135deg traces - they might be ugly, but they're very doable. This is a case where early electronic CAD turned into a cargo-cult.

Prior to the early computing CAD days, the traces were really rounded and flowing, and had teardrop fills. They were also done freehand, with transparencies, markers, and tape.

This is a good example of hand-designed boards: https://www.worthpoint.com/worthopedia/vtg-early-1970s-ampex...

> Well, to be fair, the 0/45/90 deg traces are a byproduct from early days of EDA CAD products. Aside high frequency busses (which you're doing EM analysis on anyways), traces can be all sorts of shapes.

The constraint is still there. Curved traces make your GUI ridiculously unmanageable. Ever tried to manipulate font outlines? It's like that on a PCB, only an order of magnitude worse. You have to operate control points for every single segment rather than just dragging a trace. All for effectively no benefit on 99.9% of all PCB designs.

In fact, the only good argument for using non-Euclidean traces is in the ultra-high-speed arena (think DDR4 bus or multi-GHz RF). If traces are small enough that they fit between the PCB weave, different traces can have enough difference in electrical permittivity that it will screw up the matching (trace 1 has an FR-4 fiber directly underneath while trace 2 only has cross fibers underneath and so has about 50% air underneath instead of 100% fiberglass). Consequently, you have to swing the angles to weird things like multiples of 7 or 13 degrees to prevent that.

The premise of your comment is excellent - the fiber weave effect you describe is not often considered although the two dielectric materials contained are glass bundles and resin (instead of air). I would highly recommend Bert Simonovich's publications for anyone interested in more ([1] for example). It's worth noting that DDR4 isn't really fast enough to consider "non-Euclidean" traces, though when it comes to the weave you should always be working with your fabricator to use mechanically spread glass adjacent to these high speed signals.

For anyone with more questions in this area of high speed digital design, feel free to reach out to me (contact info in profile) - I love the topic.

[1] https://blog.lamsimenterprises.com/2011/01/08/fiber-weave-ef...

Whoops. You are totally correct. It is resin and not air. I completely spaced on that.

The other good argument is when designing FPCs (flexible PCBs) --- where gentle curves reduce stress concentrations that can cause cracking and delamination.

mitxela did a great video on this topic recently and wrote a plugin for kicad to make the regular geometric layouts more organic.

https://www.youtube.com/watch?v=euJgtLcWWyo