(Note: Part 1 of this article examined the typical lab setup and powering the circuit under test. Click here to read Part 1.)

Different concerns with different topologies
Most dc/dc power-converter circuits have input capacitors and output capacitors. The impedance Z2 sees, however, is quite different.

There are converters, such as the boost converter, which have an inductor on the input side. A typical example is National Semiconductors LM3481 boost controller. In such a topology, the input current only changes very slowly, due to the inductor present on the input side. With a low input-current ripple, not much input capacitance is needed and the supply cable sees relatively stable dc currents.

However, boost-type converters require a lot of input current to start up. By definition, the input current is larger than the output current in a boost topology. During startup, the output capacitors have to get charged up, which requires a large duty-cycle during startup and thus large input currents.

In step-down switch-mode power supplies, the input current is usually smaller than the output current. Charging up the output capacitors of a step-down power supply requires less input current than for a boost converter. An example of a step-down power supply is the National Semiconductor LM5575 Simple Switcher. It even offers an adjustable soft-start functionality which can be used to reduce the startup load of the lab power supply.

Loading the circuit under test
A very convenient way of loading the circuit under test is to use an electronic load box. The simplicity comes from only having to adjust the load current at the load box, and not having to find appropriate resistors, and having to add and remove resistors when the load current needs to be varied.

For many power-supply tests, however, electric load boxes behave very much differently than a realistic load in a real system. The load box has an active control loop with the sole purpose of keeping a constant current. When the circuit under test is starting up and is increasing the output voltage slowly, the load box is changing its impedance (Z3 in Figure 2) so that the set current continues to be drawn. Sometimes, this impedance can be so low that the circuit under test can not establish the full output voltage, due to the high load on the output. Most load boxes can also be set to a fixed resistance value rather than a fixed current value. Often, however, the fixed current value is used for simplicity.

Figure 4 shows the oscilloscope plot of the startup of a step-down regulator with an active load attached to the output.

Figure 4: Typical startup when using an active load set to fixed current
(Click on image to enlarge)

Figure 5 shows the same startup, but with a purely resistive and passive load at the output.

Figure 5: Typical startup when using a passive resistive load
(Click on image to enlarge)

In these startup tests, the input voltage is slowly increased by using the enable button of a lab power supply (channel 2, purple). Channel 4 (green) shows the set output current of 1 A. Channel 1 (yellow) is the output voltage of the circuit under test, which is set to 5 V.

Figure 4 shows the big current spike at the point when the active load is starting to regulate the adjusted 1 A of current. It actively pulls large amounts of current, to get to the adjusted 1 A. By doing so, the output voltage of the evaluation board (channel 1) is forced low.

Figure 5 shows the same startup, but instead of the electronic load, a 5 Ω resistor is used. As we expect, the output current rises linearly with the output voltage.

Figure 6 shows yet another possible startup.

Figure 6: Typical startup when using an active load set to fixed resistance
(Click on image to enlarge)

An electronic load is used, but set to resistive mode where 1 A is set by setting 5 Ω as a fixed resistance. Notice that the time axis has zoomed out. It shows that during the actual startup, there is no load applied to the circuit. Only after 60 ms is the load suddenly applied.

This behavior of some electronic loads is very tricky, and is probably not detected if the output current is not measured with a probe. After a startup, the 1 A of current shows up on the electronic load display, and when looking at first few milliseconds, the startup looks nice and clean. In reality, the load is not applied during startup, but is activated after a longer delay time.

In principle, the startup, load-transient, and intermittent short-circuit behavior of a power-management circuit under test should be tested with a passive load in order to really see how the circuit under test behaves. We do not want to evaluate the behavior of the active load-box regulation loop. For steady-state testing, however, electronic loads offer a lot of convenience when changing the output current and usually cause no harm.

In general, a passive load can be attached to a circuit under test before powering it up, as long as the lab power supply has enough peak-current capability to power up the device under test and supply the load at the same time. Some circuits under test even require a minimum load for the output voltage to be regulated well.

Plugging a load in or out after the circuit under test has been powered up will result in a big load transient to the circuit under test. Such transients usually cause some output voltage overshoot or undershoot. Evaluation boards of dc/dc converters may not always be optimized for heavy load transients. This may make the output voltage of the circuit under test overshoot and possibly break the output capacitors of the circuit under test depending on their voltage rating and the way the evaluation board has been optimized for heavy load transients.

Evaluation boards are often a good compromise between size and cost of the external components, output ripple voltage, efficiency, line and load transients as well as wide input voltage and output current ranges. These design goals are sometimes contradicting. This is why power design engineers are on the save side when monitoring the output voltage of the circuit under test during smaller load transients before doing short circuit tests on the output voltage or when plugging in and removing loads.

The energy charged in the inductor may be enough to cause excessive output-voltage overshoot without the circuit under test's regulation loop being able to do anything about it. This is especially the case when the input voltage of the circuit under test is very high and the output voltage is set very low, after a sudden removal of the load.

How does lab testing represent reality?
Many real systems have the power source coming from a tight local supply such as a battery right next to the dc/dc converter. In such cases the supply cable becomes very short and less of an issue. The load is also usually very close to the power supply which makes the load cable short and low inductance. However, there are many examples of systems where the power source is further away form the dc/dc converter such as systems with ac/dc wall transformers. Other systems such as LED drivers often have the load further away from the dc/dc converter making the load cable longer and of higher impedance.

One can conclude, however, that all the rules described in this article for dealing with the dc/dc converter in evaluation also hold true for the final application.

About the author
Frederik Dostal is an Application Engineer for Power Management at the National Semiconductor Corp. Design Center in Phoenix, Arizona. He joined the company in 2001, supporting Europe. After two years, he became a Field Application Engineer for Central Europe, covering many automotive accounts. His current position involves product development as well as support for switching regulators, controllers, and linear regulators. Frederik holds an Electrical Engineering Diploma (Dipl.-Ing.) from the Friedrich-Alexander-Universität in Erlangen, and is a member of the IEEE.