There are many reasons for needing to filter power supply or signal lines. Power supply lines from a switching regulator may be too noisy for sensitive analog circuitry. Signal or power supply lines may pick up interference either from the rest of the system circuitry or from external devices. With signal lines the problem of filtering can particularly tricky because you want to remove any interference while preserving the integrity of the desired signal. That can be a problem if the desired signal and noise have similar characteristics and frequency bands. In that case you would probably need to resort to screened cables but that is often a last resort due to cost. Even high speed network cables (e.g. for 100BASE-TX or 1000BASE-T) use twisted pairs rather than screened cable like 10BASE5. With a network cable you have a low impedance and high drive current capability but with analog signals you may have a high impedance and poor drive capability so filtering becomes important.
The basic principle of a lot of filtering is to add a high series impedance and low parallel impedance to remove the unwanted frequencies or noise spikes. In the simplest terms that will be a series resistor and parallel capacitor which you will know is a low pass filter. While changing the resistor for an inductor may seem like a good idea to improve the attenuation, it is not necessarily a good idea. You will have a steeper attenuation slope due to the inductor but you will now have a resonant circuit which could give you some undesirable results. Take an example of a low impedance signal source and 1M ohm load as shown below:
Using a 1k/10nF combination results in a 16kHz 3dB point and 20dB per decade roll-off (green trace). Replacing the resistor with an inductor should improve the roll-off to 40dB/decade, which it does but it also adds an undesirable effect (red trace):
So, you can see the increased attenuation with the inductor but the peaking due to resonance as well. The exact response you will get depends on the source and load impedances. If you are familiar with passive LC filter design then you will know that the source and load impedances are important factors in designing a filter. Another way to look at it is that the lower impedances damp the resonance. So, a similar filter in a power supply line would have a lower load impedance and so a different response. Changing the load to 1k ohm instead of 1M ohms gives the following:
You still have the improvement in the roll-off with the inductor but now the peaking has been damped and is fairly small. The other advantage of the inductor compared to adding a resistor is that the series power loss and voltage drop is a lot lower with an inductor. With a power supply you probably wouldn’t want to be adding 1k ohms of series resistance but a 10mH inductor will have a series resistance of a lot less than 1k ohm. For very high frequency interference, ferrite bead type products are useful. These are specified by the attenuation at a certain frequency such as 600 ohms at 100MHz. Here you will probably have a resistance of less than 1 ohm but attenuation will often be almost non-existent at low frequencies (e.g. 100kHz) – they are mainly for very high frequency attenuation.
One device worth knowing about for reducing pickup in differential signal lines is the common mode choke which is usually a torroid with two identical windings. The mutual coupling between the windings makes the currents from common mode signals add and so the inductance attenuates the common mode signal. For differential signals the inductor currents cancel out so the signals are not attenuated. The schematic below illustrates the effect. L1 and L2 form the common mode choke which has mutual coupling between the two inductances (not shown on the diagram). The inductors are shown with slightly different values because perfectly matched inductors would perform rather better than practical ones.
The attenuation of common mode (green) and differential (red) signals is shown below. The coupling coefficient is 0.9 in the example simulated. A higher coupling coefficient would give better results at higher frequencies.
The 100pF capacitors are not an essential part of the filtering but might represent the input circuitry you are driving. With higher load resistances you need to watch out for unwanted resonances even with small, stray capacitances. Keeping the load resistance as low as possible is a good idea provided it doesn’t introduce excessive measurement errors.
Filed Under: Analog IC Tips, Capacitors