The last post examined CM radiation. Wow, what a problem! Over 100,000 microvolts/meter from a 3.5 ma signal. If the DM is normally the intended signal, where does the big common mode signal come from? Let’s take a look at that. The whole world seems to be enamored with high speeds. I know I could hardly wait to get a 1 GHz notebook. Did it help me any? Probably not. I couldn’t type any faster on it than I could on the 300 MHz system I had previously. Now, I have a 2.4 GHz quad core system and it still doesn’t type any faster than I do.
In order to process the data at high speeds without jamming the local TV stations, most designers are using high speed balanced differential signaling. This technique uses complementary signals (ie +signal / -signal) where the transmitted pulses are equal but opposite in polarity. These pulses run on a wire/trace pair and terminate into a signal receiver that responds to the difference in the transmitted signals. It takes more time to change a signal from a low voltage level to a high level; hence, the voltage swing is kept to a minimum. By maintaining the balance and keeping the signal level and loop area small, the electromagnetic fields from the complementary signals cancel each other out. This keeps the radiated emissions levels down. Plus, by using a balanced configuration, any external fields will couple equally into both signal leads and will appear as common mode and thereby be rejected by the receivers. Unfortunately nothing is perfect (just ask Murphy) and thus, signal rise and fall time variations, timing skew, differences in amplitude and pulse width all result in creating common mode currents in the ground and power systems.
Although the effects of dissimilar rise and fall times are not included in Figure 1, it does show some of the CM issues. From left to right: (1) perfect with no CM created, (2) dissimilar pulse width, (3) dissimilar amplitudes, (4) skew, (5) dissimilar amplitude and skew and (6) dissimilar amplitude and pulse width.
 |
| Figure 1 – Creation of CM noise. |
To paraphrase Thomas Paine: these are the things that try men’s souls. At least the ones doing intra-systems EMC design, also known as signal integrity (by the digital crowd), where timing, timing, and timing are the three most important elements.
Low voltage differential signaling (LVDS) is good . . . way better than singled ended where everything, especially the noise, is referenced to ground. However, it’s not perfect. The great thing about LVDS signaling is that it reduces both the radiated emission as well as radiated susceptibility. Much of the radiated RF coupling into a system from an external source affects all conductors simultaneously, creating a CM signal with respect to ground. A lightning strike is a good example. An LVDS receiver is non-responsive to CM noise because it is expecting a complementary differential signal pair; with CM, the signals are not complementary.
If your common mode emission is bad, check balance, check termination impedance (more on that later), and check complementary pulse symmetry. They all go hand in hand.
