Hi guys,
I am developing a product for machine vision using a high-speed video link over a distance of about 15 meters (~ 50 feet).
The differential signal lines are coupled by 100 nF ceramic SMT-capacitors on both sides, transmitter and Receiver (refer to channel link III, http://www.ti.com/general/docs/datasheetdiagram.tsp?genericPartNumber=DS92LX1621&diagramId=SNLS327H).
As connectors passive (without magnetics) RJ45-jacks are used, jack chassis and cable shield are galvanically connected to functional earth (FE) and capacitively coupled to signal GND by 10 nF X2 capacitors. As cable a Cat5 industrial ethernet cable with 100 Ohms differential impedance is used.
Functional earth and signal ground are galvanically isolated, only capacitively coupled.
Everything works fine and stable, so we started EMI immunity testing, starting with burst.
If we are coupling burst capacitively to cable shield using a coupling clamp as one pole and earth as the other pole, the SerDes connection is not influenced, the burst impluses seem to be drained by the cable shield to functional earth.
But if we are coupling the burst capacitively to the power supply of the unit as on pole and FE as the other pole, the SerDes connection is influenced and having transmission errors with the smallest possible voltage level of burst impulses of the burst generator (200 V).
At the input of the unit under test 3.3 nF X2 capacitors from supply line and signal ground line to FE are used to drain interferences to FE.
In parallel to the video link there is a 100 MBit/s ethernet Interface with signal lines coupled by magnetics and also shielded by FE. It is not influenced by burst to levels higher than 2 kV.
If signal ground and FE are short-circuited, everything works fine, but due to safety issues this is not acceptable for the application.
If the cable shield is galvanically coupled to signal GND and the connection between shield and FE is interrupted, the SerDes connection works fine for burst on supply but fails on burst on shield.
Has anybody an idea or explanation why ethernet is working fine in all cases of burst interference while the video SerDes connection fails?
Greetings
Steffen