Hello. I’m using a GigaBlox Nano for my project. I have three switches connected on a single shared board, which integrate simultaneous use of a Raspberry Pi 5 and an Auterion flight controller that has its own KSZ8794 switch. The ground is common for all devices and clean, while the power supplies for the switches, Raspberry Pi, and flight controller are separated.
Capacitive coupling is used without transformers, and all devices are connected to the switches using the 100M standard. The differential pair routing on the board is carefully done, with minimal length and matched pair lengths.
The problem is as follows: after a random amount of time (from 4 minutes to 1 hour), there is a short loss of connection (interface reset) on the flight controller side only. There is no packet loss, and the ping remains minimal. Changing the port does not help. Connecting a transformer on the flight controller side was also tested, with no improvement.
Could there be any incompatibility between the RTL8367N and the KSZ8794? What would you recommend in this situation?
So we can verify if this issue is related to your baseboard layout or now, are you able to test the same connection using a GigaBlox Nano + GigaBlox Nano PicoConn (and GigaBlox Nano RJConn, if you have one).
This allows us to eliminate the possibility of this being a layout issue, from which we can them look at other issues.
I would say that the KSZ series of microchip hardware absolutely do have silicon bugs; we’ve seen it on some of our boards that use that series. It has to be fixed in firmware by changing some of the register settings on the chip. That would require support from Auterion, but let’s get to that point later.
Does the Auterion software allow any configuration of it’s onboard switch? Being able to toggle duplex mode, as well as turning off EEE (Energy Efficient Ethernet), and turning off auto-negotiation may help. In fact the issues you’re seeing do sound a lot like the kind of issues that EEE causes.
Thank you for the detailed and quick response. Yes, I performed a test without my board, using your ready-made solutions - GigaBlox Nano + GigaBlox Nano PicoConn and separate battery power for the GigaBlox Nano - and the issue occurred again with the same frequency. I suspected EEE might be the cause, and you’ve confirmed that. I’ll try contacting Auterion support. If there’s any additional information, I’d be glad to hear it.
I need to disable auto-negotiation for the switch in Auterion, which should also disable EEE, and then manually set it to 100M full, correct?
Greetings. We tried disabling EEE on the switch on the flight controller side. This did not help. We tried adding transformer isolation; the interval between connection drops increased, but they did not disappear. We were able to disable EEE on the Gigablox Nano side with a resistor between pin 81 of the RTL8367N and ground. The dropouts disappeared completely.
