The right filter topology depends on the input and out impedances of the system.
EMC engineers often design low-pass filters for client projects, which should pass radiated susceptibility tests. While the most common radiated-immunity standard (IEC/EN 61000-4-3) requires testing from 80MHz to 1,000MHz, this case study required immunity from a nearby 50W, 500kHz source.
In this case, the prototype product included a simple "PI" filter, comprising a parallel 330pF capacitor, a series ferrite bead and a parallel 100pF capacitor. The load was 4.02kΩ.
I thought this would be a good time to use Linear Technology's free LTSpice software to perform some quick simulations. LTSpice is a PC-based analysis software with a convenient schematic entry front-end. It's also very easy to add voltage and current "probes" throughput your circuit to monitor the performance.
But first, which of the several filter topologies would be most appropriate? Würth Electronik has a very nice filter design section in its book, Trilogy of Magnetics. It turns out the right filter topology depends on the input and out impedances of the system.
For the series filter impedance (the ferrite bead, in this case) to work, it must "see" a low impedance at each end. For example, placing a 200Ω ferrite bead into a system with 1,000Ω and 4kΩ input-and-output impedances wouldn't offer sufficient series impedance to affect interfering signals. On the other hand, that same 200Ω ferrite would work fine if the input and output impedances were near 1Ω. That's the purpose of the parallel capacitors.
In this case, the input impedance to the filter could vary, depending on the sensor impedance, and we knew the output impedance was 4.02kΩ. The designer had already used a PI filter topology, which should perform well for any impedance from low to high, per the chart.
By inspection, I could tell the current values of capacitors were way too small to provide a low impedance to ground for a frequency of 500kHz, so the first thing I tried was to increase them to 0.47μF, each. But first, let's simulate the current circuit to see where the impedance starts to roll off.
Figure 1: LTSpice's initial analysis of the current filter circuit.
I used an input impedance of 50Ω, but with the proper valued capacitor, it really doesn't matter much. Note that the 3dB roll off (solid line) occurs about 16MHz; way too high to help at 500kHz.
Figure 2: Final LTSpice simulation with 0.47μF capacitors and the ferrite replaced with a 0.47mH inductor.
Note the improved filter starts rolling off at 20kHz-well below the critical 500kHz interfering signal and provides about 90dB of attenuation at 500kHz. The client was able to use the same real estate on the PC board to mount the new components. While there were other circuit changes and better shielding required, due to the proximity of the source, the filter reduced the high level of interference.
First published by EDN.