For a future project, I need to conveniently probe an SFP transceiver while it's operating. To do so, the SFP needs to be outside of its cage with both sides accessible. Rather than having the SFP fixed to the host, such as in this implementation, this extender uses a mini-HDMI to HDMI cable to extend all signals on the SFP connector. HDMI cables are especially well-suited for this application for multiple reasons:
- Cheap and readily available cables and connectors
- Shielded differential pairs capable of at least 1Gb/s
- Just enough pins for low speed signals
Apparently, I'm not the first one to recognize the benefits of using HDMI connectors and cables for non-HDMI applications. So far, I'm aware of it being used in JTAG test equipment and for stacking switches.
As usual, these boards are made with Horizon EDA.
This part plugs into the SFP receptacle, so it has to comply to the SFP specification. This dictated the use of a mini-HDMI connector since a HDMI connector doesn't fit within the width of an SFP or would block adjacent ports if mounted outside of the SFP cage. I initially considered using a micro-HDMI connector, but was unable to fan out all pins while still meeting clearance and via diameter rules.
Install R1 to short VccT to VccR. This is intended to balance the current on the HDMI cable in case the module draws much more current on one Vcc than on the other. This is unlikely to be relevant for most transceivers as VccT and VccR usually connected internally anyways.
It's important that this board is 1mm in thickness to fit an SFP receptacle. The differential pairs are designed for JLCPCB's 7628 stackup. I specified to chamfer the edges of the SFP edge connector since it didn't cost anything extra, but this resulted in the copper being pulled back a bit from the edge to meet their design rules, so manually sanding the edges or just leaving them as-is is probably the better option.
The board slides into into a 3D-printed shell and locks in place with integral clips. There is no latch boss, as it's probably too small to be made on an FDM printer and I was too lazy to design an unlatching mechanism. Apart from that, the friction of the edge connector provides sufficient retention.
- TE 2013978-2 mini-HDMI connector
- 0 Ohm 0603 resistor (optional, see text)
In addition to interfacing the HDMI connector to an SFP receptacle, this board fans out all low-speed signals and power supplies to 0.1" headers for convenient probing or interception. The two left columns are intended to be connected with jumpers to connect all required signals. One can remove the jumpers as needed to intercept signals or measure current consumption. The rightmost column can be connected to an oscilloscope or logic analyzer. If interception or current measurement isn't required, the signals can be connected permanently with zero-ohm resistors on the bottom side.
To get to the the normally-inaccessible bottom side of an operating transceiver, one can cut the board at the silkscreen line and omit the SFP cage.
Apart from access to the the high-speed RX/TX signals, this board should provide all that's needed for developing, debugging or reverse engineering SFP transceivers.
The differential pairs are designed for JLCPCB's 7628 stackup.
- Molex 2086581051 HDMI connector
- Amphenol UE75-A20-3000T SFP receptacle
- Amphenol U77-A1114-30L1 SFP cage
- 0.1" pin headers
- 0 Ohm 0603 resistors (optional, see text)
CERN Open Hardware Licence v1.2