_____ From: Price, Ed [mailto:[email protected]] Sent: Tuesday, February 28, 2006 6:22 AM To: [email protected] Subject: Ethernet Decoupling
I have a twisted-pair Ethernet receive and a twisted-pair Ethernet transmit set of 4 wires penetrating what would otherwise be a shielded box. Internal box noise, as low as 35 MHz, is getting out along this path. I put a 4700 pF feedthrough capacitor on each of the four lines. That worked pretty well, so well that the Ethernet shut down. Does anyone have a guideline about how much capacitance, L-L & L-G, that is typically OK? I want to thank everyone for the many instructive comments and tips about Ethernet problems! It's obvious that Ethernet interfaces must give commercial compliance engineers a lot of grief. In my case, I had the advantage of machined aluminum equipment cases, shielded interface cables that looked like small fire hoses, and MIL-grade multi-pin circular connectors that cost as much as your mortgage payment. So, what went wrong? It all came down to a bad wiring practice by a sub-module vendor; they routed the 4 Ethernet wires, and the cable shield, through the pins of several connectors. Even worse, they didn't ground the shields at each connector. As a result, Ethernet data was coupling onto all the other wires in those connectors, as well as getting onto the outside of the Ethernet cable shield. A few changes got us a 35 dB reduction of the 100 MHz noise. Ed Price [email protected] WB6WSN NARTE Certified EMC Engineer & Technician Electromagnetic Compatibility Lab Cubic Defense Applications San Diego, CA USA 858-505-2780 (Voice) 858-505-1583 (Fax) Military & Avionics EMC Is Our Specialty
