A simple fix that allows the use of these higher permeability cores while avoiding the saturation is to form common mode inductors where the sum of the currents through the wires are fairly well balanced as only the net current through the core gives rise to the magnetic field in the core. The coupled inductor that results from passing sets of signals through a ferrite bead can be used in multiple ways to reduce noise issues.

The first example using the bead as a BALUN – a contraction of Balanced-Unbalanced. A Balun configuration can be used to convert balanced signal lines into a unbalanced line or vis versa.

A good example of an unbalanced signals would be the drivers powering a motor: these are usually switched one-at-a-time rather than differentially (as would be the case for a RS-485 line).

We will draw this for a 2 wire system, but the same concept applies to a 3 phase system. The schematic shows the phase that is not switching as being grounded, while the phase that is switching at this time is shown as a pair of transistors. Not shown is the capacitance between the wires nor from the wires to the surrounding environment.

When the signal at IN1 takes a step, and we will describe for the positive transition here the step in voltage a the transistor causes a current to flow into IN1 wire that is needed to charge the capacitances associated with IN1. This current causes a positive to negative voltage drop from IN1 to Out1 due to the rapid change in current interacting with the inductance of the ferrite bead. This reduces the size of the step at OUT1 as compared to the step size at IN1. This same negative voltage step (from the transformer function of the bead) appears OUT2. If the step is sufficiently fast compared to the bead inductance, the output will be (instantaneously) nearly balanced. That is, OUT1 will step to half of V++ while OUT2 will go to minus one half of V++. The voltage between the two conductors at the input will also be essentially the voltage between the two conductors at the output. However, at the input the average of the two wires jumped to one half the supply voltage, while at the output the average remained almost at zero volts (for the high frequency components of the waveform). This greatly reduces the coupling to nearby signals. The current is also forced to be essentially equal (at high frequencies) reducing the net H-field from the wires. I earlier noted that this was true for an instant, but over time, the rate of change of the current will decrease, causing the voltage drop across the bead to decrease, and eventually OUT1 will reach V++ and OUT2 will return to zero. The frequency at which this happens can be made significantly lower than the switching edge frequency, which significantly reduces the emissions.The use of a twisted pair (or triple) causes the polarity of the coupling from each wire to reverse at the wire twist rate, which (for balanced lines) reduces both e-field and h-field coupling to the adjacent wires to mostly cancel out. This can be improved further both by the use of shielding (For the electric fields), and by using a different twist (turns per length) for the encoder versus the motor signals in a cable so that the magnetic loops of the two sets of conductors do not align. Both the shielding and the different twist counts are used in CAT-7 to reduce coupling between pairs.

But how about the Shield?

A shield is often added around the driven wires between the driver and the motor to reduce the electric field from the drive wires from radiating. The question always comes up as to where and how to ground the power cable shield. 

We have found that grounding the shield – either directly or through a significant capacitor – at the drive, and also grounding the driver signals shield at the motor significantly reduces the radiated noise from the drive. Note that the power supply is often not directly connected to the chassis ground, but generally EMI filters add a capacitance from the DC rails to the chassis. This is not shown for simplification, but the return paths for high frequency signals are still present.

The first thought is that this configuration is forming a ground loop through the chassis as the driver and the motor are  both connected to the chassis, and thus it should be avoided! For feedback signals that do not inject (significant) current into the chassis, this is certainly the case – ground one end of the shield. For the motor, the extra capacitance between the motor and the stator changes this. A ferrite bead over the whole cable also changes the operation.

Lets first look at what would happen if the shield connection to the motor were not present. A positive step at IN1 will cause OUT 1 to go positive and OUT2 to go negative, but the extra capacitance of the shield will give an additional current path through the ferrite bead and it will not work as nicely balanced as the balun example above. The motor windings have a capacitance between them and the stator of the motor. The step in the average voltage to the winding will induce a current through this capacitance into chassis ground when the driver takes a step. The only return path for this current is through the chassis, and so there will be a loop through the cable to the motor and back through the chassis. The noise will be radiated by the cable and through the chassis ground.