Hello Patrick and group,

Before considering any filtering just look at the problem. Theoretically what 
you have is a switch, followed by a load. The switch is a differential switch. 
So only RF-components should appear as a differential mode voltage between the 
two supply lines. A simple Differential mode filter ( 2 x L + X-cap) with 
sufficient roll-of should fix the problem. (draw it if it seems unclear)
Now the EMC test. We measure in Common Mode; that is between one of the supply 
lines and EARTH. For this we take earth of the measuring site, connected to the 
enclosure of your equipment if it is metal.  As earth is not connected to your 
circuit supply lines, there should not be any voltage between them.(the circuit 
loop is open) If you have a CM voltage problem, than it means that there is a 
connection between one of the supply lines and the enclosure. This connection 
is capacitive most of the time, unless you choose one of the supply lines to be 
earthed (see 2). Suppose both supply lines are equally capacitive to the 
enclosure then the enclosure would take a RF-voltage level in between the 2 
supply lines, and both lines would have an equal but opposite voltage. After 
Differential mode filtering (= short circuiting for RF ) no CM voltage on the 
supply lines would result (pos V +  neg V = 0).
In the case the capacitive coupling would be asymmetric, a common mode voltage 
would result ( <> 0). This voltage has to be filtered by a Common Mode 
filtering system ( 2 x L + 2 Y-caps).  

Conclusion: start to prevent any asymmetry in you switch mode power supply 
lines/switch/transformer circuit ( f.i. use equal length / with PCB traces ). 
Just Differential mode filtering (in essence a C) would then do the job. As 
full symmetry is impossible, some CM voltage will stay, and a small filter 
consisting of a current compensated CM coil in both phases and a couple of Y 
capacitors to earth would do the job. Make sure however that both caps should 
have their earth potential essentially the same as  the measurement set-up. 
This means that they should be connected to the metal enclosure. (Or to the PCB 
where it contacts to the enclosure)
No metal enclosure means that you have to be even more symmetric, because you 
can use a CM coil only, with less efficiency of course. 

(2) If you selected an asymmetric switcher, using one of the supply lines as 
ground (used often in DC/DC converters), you cannot filter the earth supply 
line. The problem is reduced now to a differential mode filtering problem, 
because one of the supply line is connected to measurement earth. However, any 
filtering coil in the other supply line ( the switcher) should be able to 
resist the peak DC/AC current without saturation. This requires bigger coils, 
even bigger then CM-coils. So not such a good idea. If you choose this solution 
however you should design your PCB as asymmetric as possible, give surface to 
the earth trace, and reduce length and with of the switching supply line as 
much as possible.

Keep in mind also that capacitive couplings to earth will carry RF-current, if 
you try to short circuit some supply line to earth this current can be 
substantial.. This gives the risc of DM magnetic emissions and voltage 
differences on your enclosure (V=IxR). Be sure, that every filter capacitor to 
earth only carries residue current previously weakened by an inductor.

Any questions,

Ask me

Ing. Gert Gremmen

emc/safety consultant
==
CE-test, qualified testing, 
Consultancy, Compliance tests for EMC and Electrical Safety
15 Great EMC-design tips available !
Visit our site  :  http://www.cetest.nl 
The Dutch Electronics Directory http://www.cetest.nl/electronics.htm
==


-----Oorspronkelijk bericht-----
Van:    Patrick Lawler [SMTP:[email protected]]
Verzonden:      woensdag 11 februari 1998 18:00
Aan:    [email protected]
Onderwerp:      Looking for info on designing EMI filters

I'm looking for information on designing EMI filters.  In my case,
used on the input of switchmode power supplies.

One I've seen reference I've seen is "Power Line Filter Design for
Switched Mode Power Supplies" by Mark Nave.  Unfortunately, I can't
find a place to purchase it. 


At the moment I have common-mode conducted emissions problems in the
10-30MHz range, and I'm taking a close look at the design of
common-mode inductors, including ferrite material characteristics and
winding capacitance.

TIA
--
Patrick Lawler
[email protected]

Reply via email to