We're using the DP83848 in a previous design, and the LAN8742A from Microchip in a different design.
Both run off 25MHz crystals.
And both exhibit emissions at 50,100,150,200,250,300,350,500,550MHz (harmonics).
We failed prelim EMC qualifications at specifically 200, 250 and 550MHz.
How do I get rid of these?
The only significant difference between our design and the reference design for the Microchip part (LAN8742A) is that:
1. The distance between micro-controller and PHY is about 4 times less on dev kit. We will move uC and PHY closer in next spin.
2. The RMII_MDIO and RMII_MDC lines seem to have been routed as a differential pair on dev. We will route them as impedance-matched diff pair in next spin.
3. The external magnetics they use do not have any ground or power planes beneath it, and our connector with built-in magnetics is on top of the system GND and 3V3 supply planes. We will clear the planes below it and connect system GND and chassis GND around it symmetrically with DNP resistors to be able to experiment with EMI test results.
4. One of the reference designs have 100R inline resistors on each and every RMII trace between uC and PHY.
Any other suggestions you have?
