Agile Loads Exercise Fast Supplies

The ideal power supply would provide perfect regulation regardless of load or line changes, have a very low cost, and produce zero losses and EMI emissions. It would be easy to use and never fail. A real power supply can only approach the ideal, and well-planned testing is required to ensure that design specifications and EMI emissions limits are being met.

Because environmental, safety, economic, and technical issues are involved, power supply testing can become complicated. Dividing testing into two stages—design verification test (DVT) and production test—simplifies things, especially because each stage has different requirements.

Two Flavors of Test

DVT determines how well a design performs and complies with required standards. It must meet or exceed the design specification, even under extreme combinations of operating conditions.

For example, unless the manufacturer states otherwise, a fair test is simultaneously applying minimum operating temperature, minimum line voltage, maximum load transient, and maximum line voltage transient dip. Running this test at the DVT stage helps designers understand how robust the supply is, even if few customers would likely experience this particular combination of conditions.

Standards such as IEC/EN61000-4-8 for magnetic field susceptibility or EN61000-3-2 for AC input harmonic distortion specify additional performance requirements beyond power supply functionality. The design must meet the required standards with an adequate margin to allow for manufacturing tolerances. Production sample or lot testing may be performed to prove that the products being shipped continue to meet the standards, but ideally, standards compliance is dealt with during the DVT stage.

Production testing is performed to determine that a power supply has been built correctly and meets its published specification. There usually is a deliberate distinction between the specification that is published and that which the engineering design must meet. Unless the engineering specification is more stringent than the published specification, a manufacturer is relying on only good design practices to account for component batch and other process variations.

An engineering specification may differ from the published specification because a performance margin has been added. For example, the thermal design of the power device heat sinking may be specified conservatively. This approach could improve reliability because the device temperature rise would be reduced under normal operating conditions and remain within safe limits even if cooling airflow became impeded.

Additional parameters can be included in an engineering specification. In the published specification, it may be stated that the supply meets the radiated emissions limits of EN61326. To make certain that the design actually supports this performance, the engineering specification may limit signal slew rates to reduce high-frequency emissions.

A secondary benefit of production testing is identification of design problems not found on the limited number of power supplies tested during the DVT phase. At the DVT stage, all marginal performance issues should have been determined and corrected. Any production test failures should only be caused by errors introduced through the production process itself.

However, for a number of reasons, there may be minor design faults exposed after the DVT stage. For example, several component batches simultaneously may be at their specification limits. Or substituting a different manufacturer’s part for an obsolete component can have caused an unforeseen problem. A thorough DVT program should result in very few design problems being found during production testing.

Further testing may be required to confirm the correct engineering changes. Limited redesign could be required because the customer’s application environment was not completely specified initially or may have changed subtly.

Both DVT and production testing can be categorized as static or dynamic. Static tests include the following:

Separate or combined load and line regulation.
Cross regulation.
Noise.
Efficiency.
Current limit.
AC input power.
Voltage adjustment range.

Dynamic or transient testing can be further subdivided into slow and fast events. Slow events are related to the AC line frequency and include half- or full-cycle dropout tests and AC line sag/surge tests. In contrast, AC inrush current at switch-on, the related turn-on overshoot at the output, and the output load transient response are fast events.

Load Transient Response

Because of the recent trend in semiconductor logic circuitry toward lower supply voltages and higher currents, load transient response testing is receiving much attention. According to Jim Pennington, applications engineering manager at AUTOTEST, “Probably the greatest challenge in dynamic load response testing of power supplies is not in the electronics. It’s in the interfacing. As the load step increases, so does the problem caused by any cabling which the product might require. If the product, by its own design, requires a cable for connection, the problem is almost insurmountable.

“In 1983, when AUTOTEST released its first dynamic load,” he continued, “we found the fastest practical load step where cables were involved was 1 A/µs. The industry really wanted 10 A/µs, but it was impractical in almost all cases to build interfacing fixturing that would allow this.

“Now the problem is more difficult. Some of the supplies, especially DC/DC supplies, are very fast in reacting to a current demand. Also, the demand of a single DC/DC product to provide power to a number of high-speed processors running asynchronously is one of the newer problems. This random demand can really tax a supply. I’ve actually seen a demand for a 100-A/µs step of 30 A,” Mr. Pennington concluded.

Practical Solutions

So, what approaches are power supply test manufacturers using to provide very high current slew rate load switching? AUTOTEST is addressing these demands with a load where the interface literally is part of the power component heat sink. There also are cases where the only solution is to use a fast switching device. The load is hand-selected for the individual product, and the electronic switch simply produces the step value.

A base load is provided by an electronic load, and the interface again is part of the switching module. Even using these approaches, AUTOTEST is not comfortable with steps faster than 10 A/µs unless the UUT and its interface technology are compatible.

Bob Leonard, director of marketing and sales at ELTEST, explained, “ELTEST makes its own FET-based loads, and it encompasses a variety of design disciplines: power engineering, thermal-mechanical engineering, and analog/parasitic layout issues. The ELTEST loads achieve their precision with feedback control while obtaining 10-A/µs slew rates. We consider the actual circuit techniques to be proprietary, but we work closely with customers having advanced test issues when speed improvements are desired and precision is not such a critical issue.”

The Dynaload Division of TDI produces water-cooled type WCL488 electronic loads that provide a maximum slew rate of 0 to 100% of full-scale output in 100 µs. For the 10,000-A systems, this could mean as much as a 100-A/µs slew rate. However, according to the company’s data sheet, the practical upper limit of the maximum slew rate is highly dependent on the operating mode, source characteristics, and source-to-load wiring.

High slew rate load transient testing is practical only if the power supply and its load are physically close to each other. If they are interconnected by a cable, the wiring inductance will limit the highest current slew rate that can be used.

Most commercially available loads have maximum slew rates of 10 A/µs or less, although faster rates may be available with lower control accuracy. There is a practical limit to the transient load current slew rate, but it is possible to perform a qualitative evaluation at slower edge speeds.

Power Supply Test Products High Power Load

The water-cooled, 12-kW Model WCL488 Electronic Load is approximately seven to 10 times smaller than a comparable air-cooled load. Intended for automotive, battery, fuel cell, and power supply testing, the load features constant current, constant resistance, constant voltage, constant power, and dynamic pulse modes of operation. A master unit controlling nine slave units provides up to 120 kW with 0 to 100-V or 0 to 400-V and 0 to 1,000-A ratings in a single, 48″-high 19″ rack. Voltage and current limiting and thermal overload protection are included. From $1/W. Dynaload Division of TDI, (973) 361-2922.

Rack-Mount Test Systems

Three versions of the EL-SYSR family of rack-mount power supply test systems handle up to 12 outputs and kilowatt power levels. The Model EL-DCSYSR tests DC/DC and VRM power supplies, Model EL-ACSYSR deals with AC supplies, and Model EL-ADSYSR is suitable for either AC or DC supplies. Test manifolds integrate a number of electronic loads rated at 25 A/50 V, 10 A/100 V, or 1 A/500 V and provide cooling up to 1.6 kW. Multiple manifolds are accommodated. POWERWIN^®,a Windows 95-based software program, controls the test hardware, runs standard tests, and displays captured waveforms. EL-DCSYSR: from $24,825; EL-ACSYSR: from $36,175; EL-ADSYSR: from $39,575. ELTEST, (800) 701-9347.

Power Source/Analyzer

The HP 6813B AC/DC Source and Power Analyzer has maximum AC output of 1,750-VA power, 300-Vrms voltage, 13-A rms current, 80-A peak current, and a crest factor of 6. The output frequency ranges from 45 to 1,000 Hz. DC output power up to 1,350 W, voltage to ±425 V, and 10-A maximum current are provided. In addition to rms V and I, power, VA, VAR, frequency, power factor, and crest factor measurements, harmonic analysis of I and V and triggered acquisition are featured. Programmable waveshape, voltage and frequency slew rates, and output impedance complement the pulse, step, and list output modes. From $9,690. Agilent Technologies, (800) 452-4844.

Modular AC Power Source

The SIRIUS range of AC/DC sources provides from one to three 750-VA outputs. You can program multiple outputs to act in parallel for increased current and power capacity or as a 3-phase AC source. Module outputs are independently programmable for DC to 1 kHz frequency, 0 to 280 VAC or 0 to ±400 VDC voltage, up to 15-A rms or DC current, line disturbances, and phase separation. Three amplitude levels and four time intervals offer transient, sag, and surge simulation. AC power also can start at any point in the waveform. Call company for price. AUTOTEST, (210) 661-8661.

Fast Load Switching Limitations

Transient load testing determines the speed with which a power supply can increase or decrease its output current in response to a change in current demand. The characteristics of the supply’s internal feedback control loop and its output filter determine the dynamic performance when the load changes. Figure 1 shows typical positive and negative voltage transients corresponding to a 1-kHz on/off switching load.

Assuming that the load switches very quickly—the rise and fall times have been set to 0 in this example—the power supply defines the amount and shape of the output voltage ripple. In this case, the supply requires less than 50 µs to make an initial recovery, but a much longer time is needed to settle to a final value. In addition to the electrical design of a supply, thermal design also may affect the total settling time and could account for the slow response shown in Figure 1.

According to an application note from Texas Instruments, “Forthcoming workstations, driven by next-generation high-performance microprocessors, may require from 40 to 80 W of power for the CPU alone….Parasitic interconnect impedances between the power supply and the processor must be kept to a minimum since maximum current could be anywhere from 20 A to 40 A. Load current must be supplied with up to 30 A/µs slew rate while keeping the output voltage within tight regulation and response-time tolerances.”

For the 667-ns transient that results from a 20-A load change at a slew rate of 30 A/µs, “the output filter alone controls the initial output voltage deviation. The output capacitor’s equivalent series resistance (ESR) and equivalent series inductance (ERL) are the parameters that are most critical.”¹

Transient load testing becomes much more difficult if the power supply is connected to the load by a cable because the effect of the cable inductance adds to the output filter parasitics. As Mr. Pennington of AUTOTEST commented, “The real issue is the inductive kick of the wires when the current-change magnitude and speed increase. This can totally obscure the response of the product.”

Figure 2 shows a series of voltage waveforms across an RLC network that has been driven by current steps with different rising-edge slew rates. The waveforms are drawn to different time scales to make comparison of the ringing easier.

A slew rate of 1/t0 A/s corresponds to the lowest waveform. The other three responses result from slower slew rates: 1/(1.8 t0), 1/(3 t0), and 1/(10 t0). The length of any input rising edge = t1, shown by the vertical line t1 = t0, t1 = 1.8 t0, t1 = 3 t0, and t1 = 10 t0. At time t = t0, the dashed curved line crosses each response to show their relative alignment had they been drawn to the same time scale.

As the speed of the input ramp approached the resonant frequency of the network, ringing was excited at the beginning and end of the ramp. This is similar to the effect produced by power supply output cable inductance when the load changes quickly.