Focus Enhancements FS456 Data Sheet
FS453/4 AND FS455/6
DATA SHEET: HARDWARE REFERENCE
9.5 EMI Case Study
The following notes are from an engineer's experiences passing a new board through EMI testing.
1) You may need to analyze your board trace characteristic impedance, and adjust the optimal
termination. If you can slowdown the edges without disturbing your timing, you can exorcise the
(unnecessary) higher harmonics that can scoot across close trace clearances as if they were short
circuits. It wouldn't hurt if you increase the trace to trace separation as much as possible.
2) You may add a 100pF shunt capacitor directly at the video connector pin, and put a ferrite bead
between this cap and the output cap. This, in effect, creates a 5 pole low pass filter at EMI
frequencies, but reduces to the original 3-pole in the video bandwidth (the ferrite impedance
approaches zero). 100 pF was subtracted from the output cap. Therefore, instead of a Pi filter [270
pF, 1.8 uH, 330 pF], we had a double-Pi filter [270 pF, 1.8 uH, 220 pF, 40ohms @100 MHz, 100 pF].
Another possibility was to keep the 3-pole topology and simply add the ferrite in series with the
inductor. Ferrite and shunt capacitor placement is critical. If they are not both right on the pin, HF will
escape one way or another.
3) Use a near field EMI probe to identify hot spots before going to the lab. You can always use it
again in the lab, if necessary.
4) Don't waste time with poor quality cables. Use a braid + foil cable from a reputable company.
5) A leaky PC can cause EMI emission failure. Almost any PC will leak if it has been banged
around enough. Once you find a good PC, keep it under lock and key when not being used for EMI
testing. A Gateway Performance 2000 PC™ (ATXSTF-FED) with upgraded VX920 monitor was
found to be very quiet out of the box. Make sure your TV doesn't fail on its own as well.
6) Check that the card bracket is tightly coupled to the PC chassis frame and that all contacts are
clean.
7) Another trick is to use a different layer stack-up, putting the signal traces on the inside, and using
termination. If you can slowdown the edges without disturbing your timing, you can exorcise the
(unnecessary) higher harmonics that can scoot across close trace clearances as if they were short
circuits. It wouldn't hurt if you increase the trace to trace separation as much as possible.
2) You may add a 100pF shunt capacitor directly at the video connector pin, and put a ferrite bead
between this cap and the output cap. This, in effect, creates a 5 pole low pass filter at EMI
frequencies, but reduces to the original 3-pole in the video bandwidth (the ferrite impedance
approaches zero). 100 pF was subtracted from the output cap. Therefore, instead of a Pi filter [270
pF, 1.8 uH, 330 pF], we had a double-Pi filter [270 pF, 1.8 uH, 220 pF, 40ohms @100 MHz, 100 pF].
Another possibility was to keep the 3-pole topology and simply add the ferrite in series with the
inductor. Ferrite and shunt capacitor placement is critical. If they are not both right on the pin, HF will
escape one way or another.
3) Use a near field EMI probe to identify hot spots before going to the lab. You can always use it
again in the lab, if necessary.
4) Don't waste time with poor quality cables. Use a braid + foil cable from a reputable company.
5) A leaky PC can cause EMI emission failure. Almost any PC will leak if it has been banged
around enough. Once you find a good PC, keep it under lock and key when not being used for EMI
testing. A Gateway Performance 2000 PC™ (ATXSTF-FED) with upgraded VX920 monitor was
found to be very quiet out of the box. Make sure your TV doesn't fail on its own as well.
6) Check that the card bracket is tightly coupled to the PC chassis frame and that all contacts are
clean.
7) Another trick is to use a different layer stack-up, putting the signal traces on the inside, and using
the ground and power planes as Faraday shields. However, unless the traces were placed at
right angles on adjacent planes, this is not practical (Crosstalk becomes an issue).
right angles on adjacent planes, this is not practical (Crosstalk becomes an issue).
JANUARY, 2005, VERSION 3.0
37
COPYRIGHT
©2003-4 FOCUS ENHANCEMENTS, INC.
FOCUS Enhancements Semiconductor