4.1 Introduction

It is interesting to consider the way energy is supplied to a switched electrical load in a building. For our discussion, pick a moment in time when the voltage between the “hot” lead and the grounded power conductor is at its peak of 170 V. Note that this peak of voltage appears at all points in the building at the same time.

The events we will consider will last no longer than 1 ms. At t = 0, a switch closes placing a 17-ohm resistive load across the line. Assume that the power wiring looks like 100-ohm transmission lines. At the moment of switch closure, the voltage drops to 24.7 V. This means that a step wave of −145 V starts to travel down the power wiring. When this wave reaches a branch point in the wiring, the wave divides. A part of the wave reflects and travels back to the load. At the load this wave divides again and a fraction of the wave adds voltage to the load. The reflected part of this wave then travels back to the first branch point. A part of the first wave that reached the first branch point continued outward toward other branch points. In a short period of time, waves will have traveled and reflected all over the facility. Each time one of these reflected waves returns to the load the voltage is increased.

As the waves travel and reflect, the voltage at each branch point first drops and then begins to recover. What is happening is the energy stored in the electric field in the building wiring is being moved around to supply energy to the load and to even out the voltage in the facility. In the first microsecond, waves have traveled throughout the building, and there have been hundreds of transmissions and reflections. Some of the waves will have actually reached the service entrance and will have moved out onto the distribution wiring. If the power wiring is in conduit, very little of this wave energy will radiate.

A fast oscilloscope placed at the load will show the initial voltage sag. Within the first few microseconds the voltage at the load will recover to near 170 V. If the voltage is viewed at nearby branch points, the voltage will sag but not to the same degree. If the line voltage is observed at the service entrance, the first sag may only be a few volts. Beyond the service entrance, the sag in voltage is hardly noticeable. The transmission lines in the building are not crafted for high frequency performance, so there are all sorts of discontinuities. Very soon, the waves are smeared, and there is no way to identify them as specific transmissions or reflections.

The energy demanded by the load must eventually come from a power generator. The only way a generator can sense that it must put out more power is for there to be a sag in voltage at the generator. This sag is made up of a large number of wave reflections that travel from the building to the generator. The request is a slow drop in voltage and not a sharp spike as measured at the switch contact.

It takes a few milliseconds before a voltage demand reaches the generator. Several seconds later, the generator can respond and raise the voltage. Even if this response is carried as waves to the load, the wave action cannot be identified.

After the switch closed, the energy was supplied to the load from the local electric field. The energy stored in the electric field in a building can be easily calculated. If there is 600 m of wiring and the capacitance per meter is 333 pF, the total capacitance is 0.2 μF. The energy stored in the capacitance at 170 V is 5.4 mJ. If this energy were supplied in 1 μs, the power level would be 5400 W. The peak power demanded by the load in this example is 1700 W. This calculation shows that the electric field energy that is stored in the building is capable of supplying the initial energy to the load.

There are many parallels between the switching action on a circuit board and the switching action in a building. Assume that a fast logic switch closes and connects a logic trace to the power supply. At the moment of switch closure, the voltage will sag based on the characteristic impedances of the immediate connected traces. Figure 4.1 shows traces that might be involved.

Figure 4.1 The traces involved when a logic switch closes. Note: All traces and connections are transmission lines. This is not obvious in a schematic representation.

The connections include logic traces, traces to a decoupling capacitor and vias that connect to the ground/power plane. Those logic traces that are already connected to the power supply voltage store electric field energy in their transmission lines. The voltage sag is based on the characteristic impedances of these parallel lines.¹ Waves with a negative voltage will travel out on all the power connections at the IC to the decoupling capacitor and to the ground/power plane. The reduced voltage will also progress forward on the newly connected logic. When these outgoing waves reflect and return to the load, they bring back energy that increases the voltage at the IC. Many reflections must take place before the voltage rises to near its final value. In a facility, a voltage sag lasting 1 μs is acceptable. In a logic circuit, a voltage sag of 30% lasting nanoseconds can cause trouble. It is this problem that we address in this chapter.

When the power supply sags, the traces that carry logic signals to the IC can be at a voltage higher than the power supply. If protecting diodes become conductive, then some of this voltage sag will be transmitted on these traces back to logic sources. If these logic lines are series terminated, this wave energy will be absorbed in these resistors. The point to be made is that the power supply sags can propagate on both the incoming and outgoing logic lines and must be considered a part of the noise budget.