Some time ago I introduced the equation for DM radiation from a loop. I also let Howard Johnson make fun of the single ended crowd that do not believe in loops by quoting from one of his articles. (Refer to “Why Digital Engineers Don’t Believe in EMC”, IEEE EMC Society Newsletter, Spring 1998.) The whole article is available on the internet.
Anyway, to continue:
E = 1.32 A I F2 / r
E(max) = µV/meter
A = radiating loop area (sq cm), 0.1<(l/w)<10 br="">I = drive current (amps)
F = emissions frequency (MHz)
r = measurement distance (meters)
All of the parameters in this equation directly affect the emission levels. However, the loop area is one that the designers have a lot of control over. It affects PCB layout, component placement, sub assembly location in a rack, rack configuration and cable design. Since cables tend to be the biggest emission/pickup antennas, that’s what gets discussed next.
Loop antennas are low impedance coupling transducers that are sensitive to magnetic fields. Magnetic fields are created by sources with impedances less than 377 ohms. Most semiconductor devices have low impedances, which are significatly lower than 377 ohms. This means they generate magnetic fields and will be a good match to a loop antenna for both magnetic field emission and pickup. It is very difficult to shield for magnetic fields, so the most effective hardning techniques are to reduce the drive and/or the coupling loop area.
The DM loop area for a cable is the product of the length (L) of the cable multiplied by the separation (S) between the signal and return wires (i.e. A = L * S). Because it is probably impossible to bring the wires closer together, the area is reduced by making the cable shorter. I can hear the response now – What’s a matter with you? I wouldn’t have made it that long if I could have made it shorter! How about field cancellation instead?
Okay, we’re going to do the twist and it goes like this: Take the signal and return lead and twist them together so that the lay is uniform and the twisted pair is homogeneous, and make sure that there is an even number of half-twists between the ends of the cable. A twist is one complete 360 degree rotation, and there needs to be a minimum of one twist per half wavelength (λ/2) of the highest offending frequency. More is better.
If each half-twist is considered as an individual current loop, the loop current reverses in the adjacent half-twist and the magnetic fields are partially cancelled. The reason for requiring an even number of half-twists is to make sure each half-twist is provided cancellation by its neighbor. The tighter the twist, the greater the reduction in magnetic field coupling and the higher in frequency the technique works. Balanced twisted pair transmission line may have earned its reputation in the analog telephone industry at 300 – 3000 Hz, but today we are using the Cat 7A cable over short distances (10-15 meters) at 100 Gbps data rates.
Twisted pair reduces the magnetic coupling, so it works best in low impedance circuits. The lower the impedance, the better the technique works. To help as a memory hook, twisted pair provides cancellation for DM currents because they are out of phase. Common mode (CM) currents just get dizzy as they spiral around the twists, but they won’t be cancelled.
– Ron Brewer