HDMI Routing for CM4 Carrier Board

Thread Starter


Joined Apr 16, 2024
I am trying to build an ultra-slim Raspberry Pi compute module 4 carrier board with as small of a footprint as possible. Currently I am trying to stick to the footprint of the cm4 with the connectors mounted on the bottom of the board. I am looking to have power over usb-c, a usb-a connector for peripherals and an HDMI connector for display. Most of the routing is fine and I don't expect issues however HDMI has stumped me.
The CM4 connects to the board with Hirose connectors which have pins positioned perfectly to suit an HDMI port mounted on the top of the board in line with the CM4. However, that means a board that is much wider than I would like. I have been looking for HDMI connectors with reversed pinout so they can route <5mm straight through vias to the CM4 pins, however I haven't had any luck. I have also read that HDMI is very sensitive to bad impedance matching and is infeasible on a 2-layer board.
I had a few questions that I was hoping someone could answer.
1. Does anyone know of an HDMI connector that has an inverted pinout for mounting on the underside of the board?
2. I made a (bad) route using a standard HDMI connector, if the traces are <40mm long how critical is impedance matching and the PCB only being 2 layers?



Joined Jan 27, 2019
Welcome to AAC.

This application note from NXP for one of its HDMI level shifters includes some excellent information that might help. The gist is that with HDMI you are dealing with transmission lines as you might with antennas, and keeping the impedance as close to 100Ω as you can manage is important for limiting waveform distortion.

It comes down to:

  • eliminating stubs caused by vias that aren’t blind (impedance mismatch)
    (not a concern on 2-layer boards, but an important principle for sotting potential trouble with your layout)

  • leaving a ground plane window around the pads for an SMD HDMI connector because of the impedance clobbering form factor of the pads (impedance mismatch)

  • routing the signal lines as differential pairs as close to each other as you can make them while leaving the largest air gap you can manage between pairs (crosstalk)

The following is speculative and I am not certain it would work but it might be possible to do something with a through hole connector such that some connections are made on the obverse of the board and some on the reverse. This would allow for slightly less tangly routing as you mirror the pins, and routing differential pairs alternating top and bottom to give the best gap for crosstalk elimination.