Power supply testing is taking on a new meaning these days. It is not business as usual when a power supply’s performance was rated acceptable if it passed tests at some static conditions.
Testing a power supply at nominal input voltage, full load current, and with balanced loads does not adequately check performance under real-world situations. What happens when the AC line voltage drops from 115 VAC to 105 VAC or soars to 127 VAC? Or when the battery drops from 48 V to 36 V or starts increasing if a charger turns on?
What happens when your application only uses a small percentage of an output’s current capability? Is this a problem? And what is the impact if other power supply outputs go into a protection mode?
If your power supply is tested at just the nominal input voltage and full load, the rationale may be that:
The power supply was fully characterized by engineering.
The high cost of dedicated test equipment dictated only minimal production test times.
The figure-of-merit tests confirmed that the power supply met only typical specifications.
However, the market forces at work in the industry are superseding any traditional test philosophy. Today, you no longer can bus a single large power supply throughout your product to power the latest processor and application ICs. The physics of the conductor parasites alone will prevent your high-speed application from working.
The power supply tests in this article are performed on a low-cost, Windows-based PC power supply tester. Low-power semiconductors to kilowatt DC/DC and AC/DC power supplies can take advantage of this tester’s ease-of-use, fast test times, and test libraries. The Windows-based software accommodates a variety of power supply tests and provides graphs of power supply specifications as a function of multiple parameters.
Ripple Amplitude vs Input Line Voltage
The power supply rejection ratio (PSSR) of amplifiers and digital-to-analog and analog-to-digital converters really has improved over the years. What you do not see is that PSSR is a DC specification. If these devices are subjected to even 10 kHz of frequency, the 100+ dB PSSR quickly degrades to the 40-dB range (approximately 6 bits). So switching spikes and ripple frequencies quickly get most everybody’s attention. Could a faulty or marginal component have ripple amplitude acceptable at nominal line but not at high line?
Figure 1 shows the Ripple Amplitude vs Input Line Voltage test screen. The test runs on the +5-V, 4-A output from low line (12 V) to high line (36 V) in five steps (Num Points). As shown, the ripple amplitude experiences its largest values at high line and maximum load current. It takes only 7 s to perform and graph this multiple-step test.
Ripple Frequency vs Load Current
The steady-state switching frequency of a power supply may not be static. In fact, concerns about the switching frequencies for some applications often lead to the use of a special power supply control pin to synchronize the switching frequencies of distributed power supplies. This ensures that there will be no coherent spectral artifacts surfacing in sensitive bands of the frequency-domain spectrum of the power supplies.
Conversely, multiple converters may randomly self-synchronize. Now peak currents occur simultaneously and greatly exceed the average input currents, resulting in unwanted harmonic interactions. The moral: investigate the switching frequency of your power supply.
The Ripple Frequency vs Load Current test screen provides insight into the frequency change (Figure 2). The +5-V, 4-A power supply has its load current varied from a high value of 4 A to a minimal loading current of 1 A.
The number of load-current steps to be tested is user selectable (Num Points = 5). Here, the ripple frequency is highest with lower load currents. A similar test, Ripple Frequency vs Input Line Voltage, shows that the device experiences its highest frequency at high line.
If your master clock for synchronization needs to operate at a higher switching frequency than any of your converters, then ensure your test conditions when ascertaining your converter’s highest switching frequencies. Alternatively, low-line and high-load currents have this device switching at its slowest frequency.
Efficiency vs Load Current
What power supply data sheet today does not show 80% to 90% efficiency? But do you always use the power supply at full load currents? Also, how many watts or kilowatts of power do you essentially lose if you only use 40% of an output’s capability? Do you think you are still at 80% efficiency?
While much of the industry provides standard, off-the-shelf power supplies, many power supply vendors offer high-volume users special models that optimize efficiency for multiple outputs at various load currents. Beware: ± 15 V @ 10 A for each output can be an efficiency killer if you only are using the -15-V output @ 4 A.
Figure 3 shows the Efficiency vs Load Current test screen. A +5-V, 4-A power supply has its load current varied from a high value of 4 A to a minimal loading current of 10% (0.4 A). The Number of Points (5) specifies the number of incremental steps in load current. The input line voltage is set at high line (36 V) for this test, yielding a 62% efficiency measurement at 10% loading. But, at the nominal 24-V input, the minimum efficiency is 72% (minimal load) and improves to a maximum of 83% efficiency (maximum load).
Current Limit Balanced
The term “balanced” in this test is actually a misnomer¾ its true value lies in the capability to imbalance the loads of a multiple-output system. Today’s multiple-output power supplies can yield some interesting surprises when the load currents are imbalanced, whether you are measuring the input current, efficiency, load regulation, or transient response of a multiple output supply. Many production tests only use a balanced load condition which may not be representative of actual application.
Master-slave architectures for ± 15-V or ± 12-V dual-output supplies offer various schemes for limiting excessive output currents. An output current sharing topology also is a very useful option in a supply. If you need to draw more output current from the +12-V supply and are not drawing that much from the -12-V supply, then a current-sharing architecture will let you steal from Peter to pay Paul.
All is fine until your next production lot of power semiconductor’s betas runs low (vs your prototypes and first production lot data used to release the design), and your protection circuits take that little bit extra of current before turning on. Now what happens to your power supply when the -12 V is only at minimal loading and your +12 V is in current limiting, but not enough to fully foldback.
The Current Limit Balanced test screen allows you to create an imbalanced state and to investigate where the current limit now takes place (Figure 4). A bipolar ± 12-V supply is shown with nominal full-load output currents of 415 mA.
Dynamic Load Balanced Test
Much as Transient L-H and H-L tests and other imbalanced versions are available, sometimes you may want to exercise a power supply dynamically for seconds or minutes, or even perform power-cycling burn-in conditions for days or weeks. The Dynamic Load Balanced test not only allows you to dictate the duty cycle and load conditions to alternate between, but also lets the loads to be imbalanced, perhaps creating a more stressful test condition (Figure 5).
Summary
Scaling from power semiconductors to kW DC/DC and AC/DC power supplies, a PC and Windows-based test system provides powerful combinations of test capabilities. It offers the necessary insight to ensure your power products meet the requirements of the application.
About the Author
Bob Leonard is the director of marketing and sales at ELTEST. He has a B.S.E.E. from Northeastern University. ELTEST, 26 Oxford Rd., Mansfield, MA 02048-1127, (508) 339-8210.
Copyright 1997 Nelson Publishing Inc.
June 1997