Early testing avoids compliance woes
Several tips can help you catch interference problems at the design stage.
Martin Rowe, Senior Technical Editor, and Richard A. Quinnell, Contributing Technical Editor -- Test & Measurement World, 9/1/2006
| Download the September 2006 "Compliance Test Report" |
Q: What are the most important concepts to understand about EMC and EMI?
A: Probably the single most important item is to think about EMC early in the design and use a little bit of testing in your own lab. There are a number of simple techniques that you can use in the lab that don’t require a lot of expensive equipment. They won’t cover all the bases, but they will give you a 90% confidence factor that you’re not going to attract the attention of EMI guys when your design goes out for test.
One simple test is measuring common-mode current on system cables. If you’re measuring more than 15 mA at some frequency, there’s a good chance you’ll be in trouble. All you need for this test is a current probe, which will pay for itself the first time you use it, and a cheap spectrum analyzer that covers the frequency range of interest.
 |
| A paper clip bent into a square shape and covered with insulation forms an unshielded magnetic-loop probe. Courtesy of D.C. Smith Consultants. |
A: The single most useful piece of equipment that I use in my own troubleshooting operations is a square wire loop. I make them out of paper clips because that’s usually the most handy stiff wire around, but it’s amazing what you can do. I can take a 1-in.2 loop made out of 16 gauge brass wire stuck on the end of a BNC connector, and I can tell you whether a particular chip package has low enough parasitics to support the speed of the chip in the package. I can spot signal-integrity weaknesses in the package, and I can spot EMC problems. I can also induce problems with that loop, with various types of signal generators. (Editor’s note: You can learn more about the paper-clip method on Smith’s Web site at www.emcesd.com/tt080699.htm.)
Q: Do you mean that you’re hooking up your 1-in.2 loop to a signal generator and using that as a transmitting antenna?
A: Yes. You can take a hefty noise generator or an electrical fast-transient generator, feed that into the square loop, and use that to inject controlled amounts of noise. Maybe 99.9% of the energy is reflected back to the generator, but you don’t care. I can induce 10 V of noise into circuits and determine things like noise thresholds. I can actually determine if there’s a layout defect on the board by knowing that areas of the board are more sensitive than others.
Q: Any common mistakes that engineers make?
A: Using an active probe that’s older than four years. A good differential probe these days will have a damping resistor to kill resonance from the inductance of the wires connecting the probe to the target. Older probes don’t have that.
It’s one of my pet peeves in probes, especially active differential probes, that they come with a kit of wires for hooking up the probe, and those wires are as long as 6 in. With a high-frequency probe, you really need connections that are very short. Otherwise, it’s easy to kill the probe’s bandwidth, and the wires may cause crazy resonance. For a probe of 4 GHz, I would say something less than 1/4 in. is best. Don’t go longer than 1/2 in., or about 1 cm.
