Page 2 of 2

(Sigh) What to Do, What to Do
So you have a pretty good idea what type of EMI demon you are dealing with. Now how do you get rid of it? Letıs start with conducted EMI. Can you leave the offending connection off? If possible, do it, but that is rarely the case. RC filter circuits may be used if the signal frequencies are different enough. (I wonıt go any deeper into RC filters, as they are a simple easy solution that most engineers already know how to implement. If you want to know more, write me or take stroll through my archives.

Killing Conducted EMI
You may try a choke (inductor) or even a 'juju' bead (also known as a ferrite bead). Sometimes ferrite beads seem almost magical when it comes to eliminating EMI. A few thoughts on the remarkable 'juju' bead. Ferrite beads dissipate common mode EMI. They do it be converting the magnetic field the surrounds electrical currently flow into heat. Common mode means the same signal is on all conductors. So a clamp just around the harness will not stop differential signals. Look at the back of your monitor, there is a ferrite bead there. In that application, it is stopping common mode emissions preventing the cable from broadcasting like an antenna. However, EMI is often a differential mode signal, so try this. Pass once through the ferrite, then loop the harness around and pass through again. Then your ferrite will do double duty and clamp both common mode and differential mode noise. Like this:

Figure 1: Common Mode Emission Clamp

Figure 2: Common and Differential Mode Emission Clamp

Works pretty well especially on higher frequency EMI. Now on lower frequency stuff you can try disconnecting the offending signal while still leaving it connected, how do you do that? In the example of audio interference above, we installed a transformer in the audio leg effectively disconnecting the ground loop. Sure now you ask what is a ground loop. Well, if you connect a lot of components together at various places on, letıs say a PCB. You might have them hooked up something like this:

Figure 3: Example of Several Connected Components

Note how each chip is connected to the ground plane and VCC plane at different points, so a more true representation would be this:

Figure 4: Connected Components Create Impedance

There is an impedance in all the lines. Please remember I said impedance, meaning combined elements of resistance, inductance and capacitance. So if you imagine these chips in operatation, sinking a sourcing current to and from the power supply, you could observe a different voltage at one point in the ground versus another. So if you hook up, let’s say, a sensitive audio circuit (digital circuits are pretty immune to this sorta thing) you could bounce the ground of that audio circuit around causing noise due to the EMI of the other chips. You can address this by a method called single point grounding, where ground signals connect at one common spot, usually the power source.

Taming Radiated EMI
First, if it is RF, try shielding, shielding, and more shielding. Oh and remember, shielding usually works. One point though, make sure there are no holes in your shield larger than 1 wavelength of the signal in any dimension. Otherwise the EMI will get in or out, whichever one you are trying to prevent.

Now if it is a near field effect, there are other alternatives. One easy trick is to lower the impedance of the affected circuit. If you can afford to waste the power, do it! Don’t put in a 1M pull-up resistor. Put in a 10K resistor. The more current flowing through given node, the harder it is for an external signal to disrupt it. If you think about it a little you will realize why. Another simple thing to do is to buffer any input signals with a series impedance. Even 100 ohms will take enough bite out of a fast transient signal to do huge amounts of good. Another easy thing to do is to move the EMI source away from the affected circuit. One thing I learned is that a circuit good at emitting EMI is also good at picking it up. So look at areas of your PCB where there might be antennas. RF needs an antenna right? If it is a magnetic or current induced problem, these antennas will be loops, if it is voltage based capacitive, electric field, the antennas are typically dipoles or plates. Find a place where your circuit emits it, and thatıs a good place to start working. If you canıt reduce the antenna, try shielding as a final resort. Ferrous materials work well on magnetic field signals, while anything conductive works on electric field signals.

Is it Dead Yet?
So have you learned all there is to know about EMI? Not by a long shot, but if I go on any longer, Iım afraid some of you will get "qwerty" face from falling asleep on your keyboard.

So next time you have an EMI problem, rattle your chicken bones, wave your oscilloscope leads around in ceremonial fashion, pull out your box of juju beads and amaze your coworkers as you exorcise the ghost in the machine.

<< Previous

Product Engineering Archive

		Click here to get your listing up.