– Copied from customer issue –
I’m posting an interesting issue we’ve encountered with the GigaBlox SFP (BB GFP D 1) switch. Our setup consists of an ASRock NUC‑155H‑R2 computer, Raspberry Pi 5, Jetson Orin NX, a network camera, DVL and a copper 1 Gb SFP module. Initially all ports negotiated at 1000 Mbps (except the 100 M camera). After a few days of intermittent testing the NUC port started negotiating only 100 Mbps. Moving the NUC to another port resulted in 1 Gbps, while the original port stayed at 100 Mbps. We swapped in a spare GigaBlox SFP and everything went back to 1 Gbps, but a couple of days later the replacement developed the same behaviour. On one occasion the NUC NIC even stopped replying to ARP requests and pings; reseating the cable would briefly restore connectivity. Power is provided to all devices via a 24 V Vicor DCDC regulator with soft‑start, and the electronics are mounted on a 3D‑printed nylon chassis inside a pressure vessel that sees around 50 °C. We’re using the supplied network cables.
The original unit we removed was tested on the bench and one of its ports would only negotiate 100 Mbps, while the others still worked at 1 Gbps. That suggested a hardware issue on that port. I took the GigaBlox board under a microscope to inspect the magnetics and found a cracked common mode choke (L6) adjacent to one of the mounting holes. I suspect the mounting standoffs don’t provide enough clearance and are putting mechanical stress on that component. After replacing the choke with one of the same footprint, the port returned to 1000 Mbps operation. We plan to repair the second unit, adjust the bracket clearance and see if the problem is resolved.
Thanks for the support.
Andrew