5

I'm using a Micrel KSZ8051 Phy in a 100mbps design. The datasheet specifies that it should have a small ferrite bead feeding the power.

Phy with Ferrite bead

However, the datasheet doesn't suggest any specifications for the ferrite. What should I consider when choosing a ferrite? Do I simply select the component with the highest impedance at 100MHz which can take the current?

Rocketmagnet
  • 26,933
  • 17
  • 92
  • 177
  • I would say it depends on the characteristics of your 3.3V power rail. If there is a large amount of switching noise at a certain frequency, you would want to use a ferrite that blocks them. – dext0rb Sep 16 '13 at 15:36
  • @dext0rb - Oh, I would have thought the ferrite was doing the opposite job, preventing switching noise from the Phy coming back into the 3.3v rail? – Rocketmagnet Sep 16 '13 at 15:47
  • Ah, yeah, that's probably true as well. I always forget about stuff coming back out. Either way, I think it might take some experimentation to determine what type of bead. – dext0rb Sep 16 '13 at 15:52
  • You might consider Vishay [ILHB0805ER601V](http://www.mouser.com/ds/2/427/ilhb-239956.pdf) ... 600 ohm @ 100Mhz, rated to 2A – Tut Sep 16 '13 at 16:41
  • @Tut - Would you say that the aim is simply to find the ferrite with the highest impedance at 100MHz, as long as it can supply enough current (with 300% margin)? – Rocketmagnet Sep 16 '13 at 16:52
  • That is basically my understanding; but being no expert in this, I posted as a comment. The part mentioned is what we use, but our design has yet to be taken through compliance testing. There is also a reference design at http://andybrown.me.uk/wk/2012/09/01/ethernet-phy-stm32f107/ which uses a KSZ8051 and they specify [Murata BLM18PG221SN1D](http://www.farnell.com/datasheets/5148.pdf) which is 220 ohm while 470 ohm is available on the same datasheet. – Tut Sep 16 '13 at 16:59
  • I would presume that the higher impedance is better, and the 22uF and 0.1uF capacitors should take care of the current needs at 100Mhz. – Tut Sep 16 '13 at 17:02
  • 1
    I think a smaller impedance could be called for if you experience excessive overshoot or ringing ... ref: http://www.altera.com/literature/an/AN583.pdf ... It might be good to have a range of values on hand for initial board testing. – Tut Sep 16 '13 at 17:52

1 Answers1

1

It's old school good practice. If you redesign your schematic, you will see a PI filter composed by two capacitors before, two capacitors after the inductor. It will prevent inrush high speed vdd fluctuations from the digital part of the chip, vddIO, to the analog part, vdda. At 100mbps any small ferrite will do. Pay attention on wire geometry in this region!

m33600
  • 78
  • 3