For the last several weeks I’ve been fully immersed in the design of a custom carrier board for the LattePanda N100 compute module. This isn’t just another board project — it’s also the backbone of a new course I’m recording, which I hope to release in October.
Right now, the design is at an exciting stage. The layout is essentially complete. As you can see in the screenshot below, the major routing is done, the copper fills are in place, and all the components are positioned. What remains is a careful pass through the Design Rule Check to make sure no issues slip through, some refinements to the silkscreen, and then off the files go to the manufacturer.
What makes this board interesting is that it brings together a mix of interfaces: USB-C, HDMI, Gigabit Ethernet, and M.2 expansion, alongside more familiar low-speed buses like I2C and SPI. That combination is what turns this into such a valuable learning project. It’s one thing to design a breakout board for a sensor on I2C, and quite another to route HDMI at hundreds of megahertz or USB at multi-gigabit speeds.
Let me give you an example. If you mis-route an I2C bus — say you make one trace longer than the other or you snake it around the board — chances are it will still work. I2C runs at a few hundred kilohertz or a couple of megahertz at most, and the signals don’t “see” the PCB traces as transmission lines. The worst that might happen is a little noise or a missed clock edge.
Now compare that to USB-C. If one of the differential pair traces is just half a millimeter longer than its partner, the signals can arrive at slightly different times, causing skew. That skew may mean the receiver no longer interprets the data correctly, and the port either runs unreliably or doesn’t enumerate at all. Or imagine you forget to maintain a solid ground reference plane underneath the USB pair. The return current has to find another path, which increases electromagnetic interference and weakens signal integrity. These are the kinds of mistakes that don’t matter in I2C or SPI but will completely break HDMI or USB.
This is where impedance matching comes in. At high speeds, the width of the trace, the spacing between differential pairs, the thickness of the board, and even the copper weight all determine the impedance of the signal path. If the impedance doesn’t match what the transmitter and receiver expect, you get reflections that degrade the signal. In practice, that might mean an HDMI monitor that flickers, or a USB drive that keeps disconnecting.
In the course, I’ll show you exactly how to prevent these issues. You’ll learn how to set up KiCad’s design rules so the tool itself enforces differential pair spacing and skew tolerances. You’ll see how to use an impedance calculator to select the right trace widths for the stackup. You’ll practice routing differential pairs while length-tuning them so they arrive together, and you’ll understand why copper pours and stitching vias are critical for providing clean return paths.
I think this is why the course is so important. If you’ve only ever designed boards with low-speed interfaces, you might not have encountered these problems yet. But as soon as you step into real-world, high-speed design, the rules change. This project gives you a guided introduction to those rules. By the time you finish, you’ll not only understand how to design a LattePanda carrier board, you’ll also have the confidence to take on any complex, high-speed PCB design challenge.
The next step for me is manufacturing. Once I get the boards back, I’ll assemble and test them in the lab, and then record the final lessons of the course. If all goes well, you’ll have access to the full course in October (I hope!!!).
I can’t wait to share the finished project with you — and I’m even more excited to see what you’ll build once you’ve mastered these techniques.