Been scratching my head over something similar on my narrowboat setup and this thread topic has genuinely stopped me in my tracks.
I've got a Cerbo GX talking to a pair of Fogstar Drift 200Ah batteries via BMS-Can, and every few days I'd get completely nonsensical state-of-charge readings — we're talking 4% one minute, 94% the next. Assumed it was a dodgy CAN cable or possibly a firmware quirk. Never once considered the Ethernet side of things could be involved.
Can someone explain the mechanism here? I'm struggling to understand how a duplex mismatch on the Ethernet port would corrupt data coming through BMS-Can — are they sharing some internal buffer or processing resource on the Cerbo that causes the knock-on effect? Or is it more that the Cerbo's CPU gets hammered trying to handle the Ethernet errors and essentially drops/garbles the CAN packets?
My Cerbo is connected via Ethernet to a TP-Link switch (nothing fancy), and I honestly never checked whether autonegotiation is behaving properly on either end. That feels like an embarrassingly simple thing to have overlooked for months.
- Has anyone confirmed this with specific firmware versions?
- Is forcing 100Mbps full-duplex on both ends the recommended fix, or does Victron suggest something else?
- Would the VRM logs show any indication of the Ethernet errors, or would you only catch it via the Cerbo's local diagnostics?
Interested whether anyone else running lithium setups in marine or shepherd's hut installs has hit this — feels like the kind of gremlin that would be incredibly easy to misattribute to the batteries themselves.
