The battery symbol represents an ideal constant voltage source with perfect regulation and zero output impedance at all frequencies, but every regulated power supply has some small output impedance at high frequencies. Thus a more exact circuit model for a power supply includes an equivalent source resistance and inductance as shown in Figure 37. RS is the power supply output impedance at dc, and is found by dividing the load regulation by the current rating; for example, a power supply which has a load regulation of 10mV for a full load change of 10 amps has an equivalent RS of 1 milliohm, a typical value. Similarly, a power supply with an output impedance of 0.2 ohms at 100KHz and 2 ohms at 1MHz has an equivalent high frequency output inductance LS of 0.3µ H - again a value typical of high performance power supplies.
The connecting lines on a schematic represent ideal connection between two points, but the physical wires used to connect any two terminals (such as power supply and load) are characterized by distributed resistance, inductance, and capacitance. For determining necessary load wire sizes, it is usually sufficient to consider only the equivalent lumped constant series resistance and inductance (L0, L1, L2 . . . and R0, R1, R2. . .). Given the wire size and length, these lumped equivalents can be determined from wire tables and charts.
In general, the power supply performance degradation seen at the load terminals becomes significant whenever the wire size and length result in a load wire impedance comparable to or greater than the equivalent power supply output impedance. With one load, this degradation can be evaluated by comparing 2Ro with Rs, and 2Lo with Ls. The total impedance seen by the load is ZT = (RS + 2R0) + jω (Ls + 2Lo), and the variation of the dc load voltage caused by a sinusoidal variation of load current is Eac = Iac ZT. If load current variations are more
If these calculations indicate that the resulting variations in dc voltage provided to the load are greater than desired, then shorter and/or larger load leads are required.
With multiple loads (Figure 37B), it is necessary to consider separately the common or mutual impedance seen by the loads, RS + 2R0) + jω (LS + L0), and the added impedance seen by each load individually, 2(R1 + jω L1), 2(R2 + jω L2), etc. Remember that the mutual impedance presents an opportunity for a variation of one load current to cause a dc voltage variation at another load. If the loads are pulse or digital circuits, false triggering may result. Similarly, if one load is the output stage of a high gain amplifier, and another load contains low level stages feeding the same signal path, unintentional feedback may occur via this mutual impedance, with resulting amplifier oscillation.
Connecting remote sensing to the load terminals of Figure 37A or the DT's of Figure 37B has the effect of reducing R0 by a factor equal to the loop gain of the power supply regulator, usually of the order of 103, 104, or 105. However, remote sensing does not in general alter the effective value of L0 seen by the load, since L0 predominates at frequencies above the bandwidth of the power supply regulator.
Since remote sensing affords little or no reduction in the effective load wiring impedance at high frequencies, some amount of capacitive load decoupling is sometimes desirable, particularly when multiple loads are connected to a power supply.
Load Decoupling
A local decoupling capacitor, if required, should be connected across each pair of load and distribution terminals. This reduces the high frequency impedance seen by any individual load looking back toward the power supply, and reduces high frequency mutual coupling effects between loads fed from the same supply. The use of load decoupling capacitors is most often employed with multiple loads drawing pulse currents with short rise times. Without local decoupling these current changes can cause spikes which
64
