•  
• EDN Europe MAGAZINE
•  
• Free SUBSCRIPTION
• Subscription Renewal
• View Media Kit

• EDN
• EDN Australia
• EDN Asia
• EDN China
• EDN Japan

E-loads lighten PSU test burden

DESPITE THE ARRIVAL OF DIGITAL TECHNIQUES, OPTIMISING POWER-SUPPLY PERFORMANCE MOST OFTEN INVOLVES MULTIPLE ITERATIONS OF COMPONENT SELECTION OR OPERATING-POINT ADJUSTMENTS THAT REQUIRE A STANDARD TEST SET-UP. ELECTRONIC LOADS PROVIDE A UNIFORM PLATFORM TO SIMPLIFY YOUR TEST PROCESSES.

Barely two decades ago, power-supply design formed the least glamorous task within the equipment-creation process. Routinely the responsibility of the engineering team’s most junior members, the most demanding aspect of configuring the linear supplies of the time saw aspiring designers grappling with heatsink selection and packaging. The advent of switch-mode supplies changed this situation at a stroke—starting a trend toward’s today’s situation where issues ranging from ac-line quality to the pursuit of optimal energy efficiency make power-supply design a specialist subject. In the latest evolution of power design, we have digital gurus applying DSP techniques to replace the handcraftedanalogue control loops that form the heart of any traditional supply.

But regardless of the method of finetuning a converter of any topology, the question remains: When you get your prototypes back, how do you ensure that they work to their optimalpotential?

The circuitry that you need to test largely determines the best approach for evaluation and qualification. For instance, if you’re designing for automotive powertrain applications, realworld loads such as the fans and solenoids that the circuit must drive are indispensable. Here too, incandescent bulbs are often preferable to fixedresistor loads as bulbs can dissipate high power levels in small spaces and don’t require heatsinks. Bulbs also make a cheap-and-cheerful dynamic load as their cold-to-warm current at power-on is some 7 to 15 times the steady-state value. But when you need to characterise equipment to a standard operational template, such as a circuit’s ability to function correctly during cold-start engine-cranking conditions, you need a more scientific test set-up that’s capable of reproducing the same loading conditions time after time. And when the circuit’s operating voltage falls to today’s CPU core-logic levels, the load’s ability to sink high currents at low terminal voltagesbecomes crucial.

AT A GLANCE

Multi-mode loads simplify component and system tests. Analyse test requirements before selecting equipment. Analyse test requirements before selecting equipment. Analyse test requirements before selecting equipment.

Enter the electronic load or “e-load”, a test instrument that’s basically an active current sink, albeit offering more functionality than a conventional load-resistor array. Long the test equipment of choice for evaluating components such as capacitors and rectifiers, e-loads are indispensable in powersupply development—and find uses in less obvious applications, too. Wolfgang Horrig, sales manager at e-load maker Elektro-Automatik, cites the example of a fork-lift-truck manufacturer using an e-load to absorb excess energy when testing the truck’s electric motors: “Firstly, they replace the 48V battery system with a power supply to speed motor testing over range of voltages. But then there is a problem when braking the motor, when a big back emf appears that will destroy the power supply’s output capacitors. Connecting an e-load set to around 49V in constant-voltage mode absorbs the excess energy and permits safe testing.” This approach is widely applicable for testing many other inductive loads, so if you plan similar uses for your equipment, ensure that it supports constant-voltage mode as well as the conventional common constantcurrentmode.

WHAT’S INSIDE AN E-LOAD?

So, what’s inside an e-load and how does it maintain CC (constant-current) and CV (constant-voltage) operation? Analogue enthusiasts will immediately visualise a power op-amp and may even consider building one—possibly following the outlines of a Design Idea that appeared in EDN some time ago (Reference 1). This Design Idea illustrates the principles that lie at the heart of an e-load that offers currentand voltage-regulation modes. As the outline circuit shows (Figure 1), these modes are fundamentally different in operation. In constant-current mode, feedback action around IC1A forces the voltage across RSHUNT to equal VREF. The current that flows through Q1 is then VREF/RSHUNT. This mode suits the testing of voltage sources such as a typical power supply. By varying the current load, you can quickly assess a power supply’s regulation and confirm its overload characteristics. Adding a digitally programmable voltage reference in the shape of a DAC transforms the constant-current circuit into a dynamic load tester that’s particularly suitable for assessing a power supply’stransient response.

In constant-voltage mode, the e-load forces a current source to deliver a voltage that the load value determines. Applications for this mode include the testing of current sources, for example photovoltaic arrays. Such arrays appear as current sources at points below their voltage at the maximum power point for a given load; above this point, they appear as voltage sources. In our outline circuit, IC1B forces the voltage at the junction between R1 and R2 to equal VREF. This makes Q1 act as a high-power shunt regulator, dumping current into RSHUNT to maintain a fixed voltage across the load. The transfer characteristic is VLOAD
VREF(R1R2)/R2, allowing suitable resistor values to enable voltage operation far above the op-amp’s supply level. It’s worth noting that this mode best suits true current sources, where the presence of external shunt capacitances can create stability problems. Because power supplies thathave a constant-current mode only appear as high impedance within theirown loop bandwidth and because anyshunt capacitance reduces impedancesubstantially beyond this point, it canbe difficult to maintain stability whentesting this configuration.

CONSIDER LOAD DEMANDS

A surprisingly large selection of vendors offers off-the-shelf test equipment that ranges from simple benchtop loads for a few hundred euros to rack-mounted, remotely controlled, water-cooled devices that suit ATE systems in power-supply productiontest applications (see sidebar “For more information” for a representative listing). So what should you look for in selecting an e-load? Unsurprisingly, power-handling ability is the prime selection factor, with vendors most often classifying equipment by its maximum dissipation. Just as important are the maximum voltages and currents that the equipment can handle, while the number of operating modes and facilities such as system connectivity are often decisive factors, too. Separate load and sense-input connections increase accuracy by compensating forseries resistance in the current-sinking wiring, while galvanic isolation allowsthe e-load’s inputs to float by as muchas several hundred volts on either sideof ac-line ground.

Rick Parizot, sales director for Transistor Devices’ Dynaload products, notes that engineers increasingly specify just one e-load to cover a range of applications that previously required multiple e-loads. A prime reason is that today’s power FETs are capable of operating over a much wider area than yesteryear’s bipolartransistor designs. For example, a traditional bipolar e-load capable of highvoltage operation requires multiple, relatively low-voltage power transistors in cascode, which intrinsically limits the lowest operating voltage to tens of volts. Today’s high-voltage FETs extend the operating point toward zero volts, are relatively easy to operate in parallel in order to sink more current, and are both faster and more rugged than their bipolar counterparts. As with most e-loads, it’s also possible to connect multiple Dynaloads in parallel to accommodatearbitrarily large currents.

Any e-load’s power-handling abilityfollows a voltage/current curve: for instance, the RBL488 100-600-6000from Dynaload’s RBL488 range of6-kW e-loads handles up to 100Vand 600A.This series of rack-mountinstruments targets system applicationsand includes Ethernet, RS-232,and IEEE-488.2 interfaces. One ofthree versions that cater for levels ofup to 600V/200A, the RBL488 100-600-6000’s maximum power envelopeextends from 100V at 60A to 600Aat 10V (Figure 2). Notice that as thevoltage drops below the 1V level, theunit’s ability to sink current graduallyfolds back to around 80A at 100mV asthe current sink’s minimum compliancevoltage approaches. The abilityto sink high currents below the 1Vlevel is one differentiating factorbetween e-loads, some of which can’tsustain full conduction below 2 or3V: at some point in that range, theoperating point of their power devicesshifts to limit conductance. In somee-loads, this behaviour can lead toslower response and even stabilityissues under unfavourable conditionsas loop gain drops sharply. Parizotnotes that careful FET selection andmaintaining tight regulation aroundeach device is critical to extractingmaximum performance at the lowvoltages that the testing of sourcessuch as fuel cells and very low-voltagelogic supplies demands: “MOSFETsdon’t like sharing current any too wellat low operating voltages, so it’s importantto individually close-loop-controleach device in the parallel array.”

The MOSFET’s behaviour within its linear region is an essential selection parameter; devices that target audio-amplifier applications typically are more suitable than those that suit switch-mode power supplies. Other crucial circuit elements include the choice of series shunt resistor, which must be a precision, low inductance device. Parizot says that it’s important to balance regulation complexity between stability and loop response to avoid the “hairline-trigger” conditions of instability that some early-generation loads exhibited. Programming lower slew rates has a limited effect, leading some manufacturers to add damping capacitance to help alleviatethis problem—whereas the lowest possible amount of shunt capacitancemaximises loop response. As an additionalmeasure, Dynaload productsship with low inductance cables thatminimise inductances in the wiringsystem. Such parasitics create phaseshifts that the regulation loop will tryto compensate, again compromisingresponse and loop stability.

If your e-load can’t conduct at sufficiently low voltages, it’s possible to use a separate power supply to extend the effective operating point to 0V. This strategy consists of inserting a supply of opposite polarity in series between the e-load and the source. In effect, this additional supply adds an offset voltage between the e-load’s current-sink terminals and the source to raise the aggregate voltage to a point that the e-load can handle. It’s then important to sense the source voltage directly across the source by using the e-load’s remote-sense connections, leaving the current-sink connections wired to the e-load side of the offset supply. The offset supply must be capable of supporting the full load current, and its dynamic performance must not compromise that of the source under test. You must also guard against applying reverse voltages to the e-load—many of which have protection diodes permanently connected across their input terminals— which potentially leads to uncontrolled current flow under faultconditions.

SIDEBAR—CASE STUDY: TESTING IC DC/DC CONVERTERS

For many engineers, today’s impressive selection of controller ICs and power FETs has done little to ease the task of designing high-power dc/dc converters—especially when the circuit must accommodate a wide range of input voltages and occupy as little real estate as possible. Recognising this opportunity, Linear Technology released its LTM4600 Module, a 10A dc/dc converter in a 15152.8-mm LGA package that’s the “power system and modules” section winner of EDN’s 2005 Innovation Awards. The device only requires bulk input and output capacitance and a resistor to set the output level. It accommodates input voltages from 4.5 to 20V, with a 28V version also available; output voltages span 0.6 to 5V. Because they are constant on-time currentmode converters, it’s possible to connect multiple LTM4600s in parallel to obtain more power. The recently released and broadly similar LTM4601 improves on this scheme by adding a synchronisation input that allows you to build a multiple-phase converter— for instance, by clocking two devices at 0 and 180. In this way, the 12A-rated outputs again sum to share current while doubling the output-ripple frequency, allowing reductions in bulk output capacitance and improving response times. The default operating frequency is 800 kHz while the external synchronisation range spans 560 kHz to 1 MHz. Other refi nements include two voltage-margining pins that cause the converter to swing its output to 5 and 5% of its nominal 0.6 to 5V range. For $45, the LTM4601EV evaluation board demonstrates the 20V converter; the otherwise identical LTM4601HV substitutes the 28V device and costs $55. I tested an LTM4601EV to gauge how well the device might perform in a 5V confi guration that—if successful— will power the intermediate bus in a battery- powered application that has a nominal 14.4V input level. The test load is a model LD300 from Thurlby Thandar Instruments (TTi), which retails for around €1043. This versatile 80A/80V e-load has a 300W rating and includes CC, CV, CP, CG, and CR modes (see main article) together with dynamic test abilities that can use the built-in variable slewrate transient generator or an external analogue input for connection to, say, an arbitrary waveform generator. To establish a test regime, I started testing the evaluation board in CC mode, using a 10A linear bench supply and no additional capacitance. Parameters of immediate interest comprise the module’s conversion effi ciency and its power dissipation. Applying 14.4V to the module generated a 4.955V output level with the no-load current measuring almost exactly 100 mA (if you’re planning to use input voltages above 16V with a 5V output, add a 100k resistor between the module’s frequency-setting input and ground). Under these conditions, the casetemperature rise was 18C for a 20C ambient environment. Allowing a 15-min settling period at 12A output following a test run of 1A load increments raised the case temperature by 55°C, while the output-voltage level fell to 4.862V. Operating effi ciency appears essentially fl at at around 90% from about 3A upwards (Figure A). Without any output capacitance, slewing the load between 8 and 10A at 100 Hz, 50% duty cycle, and 10A/msec unsurprisingly imposed a square wave on the output that the addition of a low-ESR 470-F capacitor reduced to between 50 and 100 mV in a low-frequency bandwidth. There was about fi ve times this level of high-frequency noise on the output that sometimes broke into the digital radio channel that adds some life to my lab, but I attribute this to the large inductances and aerial loops that the test set-up’s interconnecting leads created—signifi cantly, this noise was present regardless of transient or steady-state loading. For my space-constrained target application, these results suggest that it will be worthwhile producing a test board to compare the performance of a pair of LTM4601 devices with that of a discrete multiphase converter design, as the space savings that the modules offer are likely to more than compensate the increase in bill-of-material costs.

 

OPERATING MODES EXTEND USE

Operating modes that expand upon constant current/voltage operation comprise CP (constant power), CG (constant conductance), and CR (constant resistance). Within the source’s limits, constant-power operation removes the dependency on supply voltage, which is useful in applications such as the testing of batteries in equipment that employs switch-mode converters. Within their operating envelope, most such converters draw largely the same amount of current irrespective of input voltage. Other common applications include checking the output characteristics of photovoltaic cells under varying levels of irradiation. Constant- power circuits implement the expression I
W/V by measuring the load-current and voltage levels and by applying feedback to maintain the preset level. As a result, terminal voltage falls as the current level rises—effectively creating a negative resistance, where the output impedance of the source may interact to form a negative- resistance oscillator. For batteries, it’s much more likely that a point will come when the source can no longersustain the preset load level. Beyond this point, the e-load latches into hardconduction, absorbing maximum currentat close-to-zero voltage. To avoidthis condition, many e-loads include avoltage preset threshold, below whichthe device ceases conduction.

Constant conductance mode is similar in that the e-load measures the source voltage and adjusts the current level to simulate the conductance level that you set. But because the e-load now acts to express I
VG, load current rises in direct proportion to the source’s terminal voltage. This mode is generally most useful for simulating very low-load-resistance settings. Conversely, constant resistance mode implements I
V/R and is better for setting high-resistance values. Both modes use multiplication/ division techniques to derive the current level from the terminal voltage, and stability isn’t normally a problem under static-load conditions. However, this situation can change under dynamic conditions. While constant-conductance mode is a linear function, constant-resistance mode demands that the current be inversely proportional to linear changes in resistance value. Applying a transient generator to the constant-resistance circuit thus creates highly non-linear waveforms that rise very rapidly from the low-resistance setting, potentially creating ringing and overshoot due to interconnection inductances. In general, constant-resistance mode best suits static operation at higher voltagesand relatively low currents.

DYNAMIC TESTS MADE EASY

All but the simplest e-loads offer some form of transient generator and/or an external input that you can use to modulate the load’s response (see sidebar “Case Study: testing IC dc/dc converters”). For example, the EL3000A series of benchtop e-loads from Elektro-Automatik implements CC, CV, CP, and CR modes that you can use in conjunction with two preset values. In automatic operation, these e-loads switch between the preset values with adjustable pulse widths and slew rates that you can set manually or via the optional CAN, USB, or RS-232 interfaces. The pulse time ranges from 50 sec to 100 sec to cover the frequency range of 10 kHz to 0.005 Hz. The slew-rate setting is the same for rising and falling edges and is continuously adjustable from 30 sec to 200 msec with 10% worst-case accuracy. The EL3000A expresses the change in terms of A/t; consequently, a change from 20 to 40A in 100 msec equates to 20A/100 msec. The 15-pin analogue interface that’s a standard feature includes four input channels that convert a 0 to 10V control signal to 0 to 100% of set value for voltage, current, power, and resistance. The unit requires at least the first three of these signals to operate in this mode. For instance, to control its constantcurrent mode requires 0V into the voltage channel, 10V into the power channel, and 0 to 10V into the current channel. Using a function generator or a programmable-logic controller’s analogue outputs to modulate thecontrol voltage provides a method of level and slew-rate control with thesame range as the internal settingspermit.

The EL3000A also features a batterytest mode that allows you to discharge a battery and measure its ampere/hour capacity until the terminal voltage falls below an adjustable value, when the unit will disconnect the battery to prevent deep discharges. Costing €875 each, two models are available that cover the ranges up to 160V/60A and 400V/25A, respectively, at 400W continuously, with a 1%/C derating in dissipation above 30C ambient up to an environmental maximum of 70C. Elektro-Automatik’s Horrig observes that thermal issues determine any e-load’s dissipation limit, noting that the functionally similar EL9000A series that handles 1200, 2400, and 3600W continuously will dissipate as much as twice these values during the instrument’s warm-up time or when operating in a cold environment. These rack-mount instruments include a proprietary two-wire control bus that assures accurate current sharing between multiple units in parallel, when one instrument acts as the master and all others as slaves. Pricing for the EL9000A series spans €1695 to €3795; digital-interface options comprise RS-232 (€119), USB (€159), CAN (€298), andIEEE-488.2 (€698).

Common features among e-loads include a soft-start mode whose ramp-up time you can program plus, of course, comprehensive protection mechanisms for the e-load’s internals. These mechanisms typically include overcurrent and overtemperature monitors that disconnect the load under fault conditions. It’s more difficult to comprehensively protect against overvoltage conditions. Don’t forget that many e-loads implement reverse-input-voltage protection using a diode that’s permanently connected across the input terminals—uncontrolled current will then flow if, for instance, you reverse-connect a 12V car battery. Horrig advises simply to avoid reverse-polarity connections as there’s a substantial possibility of damaging the e-load as well as the source. Also remember that malfunctions within power circuits can easily threaten your health, so be sure to observe safe working practices—asexperience in my own lab attests, components can and do explode, soalways wear eye protection!

Elektro-Automatik also offers its BFC 2000, a 2-kW “green” load that recovers the source’s input energy and feeds it back into the 230V/50-Hz line via a line-locked sinewave inverter (Figure 3). As the unit requires around 40W to operate, its efficiency in performing this transformation depends on the input-power level. Horrig estimates that under best-case conditions, the device can recover power with 88 to 90% efficiency, which not only minimises energy costs but also greatly reduces the unit’s cooling requirements. It’s possible to connect multiple units in parallel to extend the device’s 60V/80A capability, which—given the reduction in power dissipation—can eliminate the need for air conditioning in applications such as telecoms test stations. The BFC 2000 is available for €3495, with the optional RS-232/IEEE488.2interface adding €498.

SPEED TRANSIENT PSU TESTS

Long a maker of e-loads, Agilent recently introduced an unusual instrument that integrates an oscilloscope, DMM, arbitrary waveform generator, datalogger, and up to four dc supplies to build what the company describes as a dc power analyser. Here, analysis refers to the ability to generate and record phenomena such as transient voltages or to simulate ripple to assess how well a power supply or systemwithstands the interference. In effect, you can consider its model N6705A as a power arbitrarywaveform generator with built-in instrumentation anda control interface that dispenses with the need for PCsor programming. For systems use, you can program theN6705A over Ethernet, GPIB, and USB; the instrumentalso complies with LXI Class C. Bob Zollo, product managerfor power products, explains that three types of plug-indc modules are available—basic, high-performance, andprecision—each of which carries instrumentation featuringvarying capabilities. The choice of 21 modules for the fourslotmainframe allows you to configure the instrument tosource as much as 600W, with all but one module occupyinga single slot (Figure 4). Voltage and current levels areas high as 100V and 20A. Because each module includes itsown measurement circuits, it’s possible to simultaneouslystream events into the mainframe’s 64-Mbyte file store forlater recall. If you need more data storage, plug in a USBmemory stick.

Zollo says that the key to the instrument’s power-sourcing performance lies within its fast switch-mode power supplies. Now in its seventh generation, Agilent’s SMPS substitutes a DSP heart for analogue control loops to equal the performance of the best linear supplies. Zollo quotes maximum peak-to-peak noise levels of around 3.5 mV with the rms-equivalent level in the microvolt region. Moreover, the high-performance and precision modules support 10V step changes in either direction within 160 sec. The DSP plays a key role here as it is reprogrammable on-the-fly to, for instance, change the slew rate while the signal is slewing to reduce overshoot. Another critical element is the “output down programmer”, which is a small e-load that rapidly discharges the output capacitors to speed the negative-going transition. This circuit doesn’t sink current continuously but enables the instrument to charge and discharge as much as 1000 F of external capacitance fourtimes per second.

Other useful features include overvoltage, overcurrent, and overtemperature protection along with circuitry that guards against accidental sense-wire disconnection: “Instead of drifting to possibly dangerous levels, the output will float to within about 1% of what you’re expecting,” Zollo says. There’s also a “red panic button” that’s a master circuit breaker to disconnect all supplies in the event of the user’s circuitry malfunctioning. Zollo notes that the instrument continues recording through panic events to document the sequence of events leading up to and following the problem. The N6705A mainframe is available nowand costs $6500; module prices span $450 to $2500.

CONTACT DETAILS You can reach Contributing Technical Editor David Marsh atforncett@btinternet.com.

REFERENCES

Garrigos, A and Jose M Blanes, “Power MOSFET is core of regulated dc electronic load,” EDN, March 17, 2005, pg 92, www.edn.com.