Boost current for high-power tests

You can combine voltage and current sources to produce current that exceeds 100 A.

By David Wyban, Keithley Instruments -- Test & Measurement World, 6/1/2010 12:00:00 AM

SIDEBAR: Cabling and test-fixture considerations Read more articles from our June issue.

Characterizing solar cells, power-management devices, high-brightness LEDs, and RF power transistors often requires tens of amperes, but power MOSFETs and IGBTs (insulated gate bipolar transistors), can require currents in excess of 100 A. Most power supplies, however, can't provide that much current. Fortunately, you can combine voltage and current sources to achieve higher current than one source can provide.

You can use multiple DC current sources to achieve higher current, but high DC currents will cause the DUT (device under test) to heat, which will increase its resistance and possibly lead to measurement errors. To keep the device cool, you can provide pulsed power to the DUT. Pulsed-current-voltage (I-V) testing is often essential for testing power devices (MOSFETs or IGBTs) and high-power RF amplifiers. Although DC current sources typically don't let you pulse their outputs, you can build pulse circuits yourself or use an SMU (source-measure unit) to generate the necessary pulsed power.

Pulsed-power tests sidestep the problem created by high-power CW (continuous wave) testing. During high-power CW tests, semiconductor material will dissipate the applied power as heat. As the device heats, its conduction current decreases, because its carriers have more collisions with the vibrating lattice (phonon scattering). The self-heating will cause a low measured current. Given that power MOSFETs and IGBTs typically run in pulsed mode—intermittently rather than continuously—testing with low DC-measured currents won't accurately reflect their performance.

Pulsed sweeps, where power to a DUT may increase with each pulse, have little impact on many test results, yet they do have some limitations. Pulsed sweeps can affect DUTs such as capacitors, so the results of capacitor tests performed with pulses may not correlate adequately with DC sweeps. Large displacement currents, which can occur at the sharp edges of a voltage pulse, may change the device's electrical properties.

Figure 1. Two pulsed-current sources add to each other when connected to the same node.

When making a pulsed sweep, you must take pulse width into account. The current pulse must be wide enough to allow sufficient time for the device transients, cabling, and other interface circuits to settle so the system can make a stable, repeatable measurement. But the pulse can't be so wide that it exceeds the test instrument's maximum pulse width and duty cycle limits, which would violate the instrument's allowed power duty cycle. Pulses that are too wide can also create the same device self-heating problems that can occur with DC sweeps.

Combining channels

The most common way to combine SMU channels to achieve higher DC currents is to connect the current sources in parallel across the DUT. The test setup in Figure 1 takes advantage of Kirchhoff's current law, which states that when you connect two or more current sources in parallel to the same circuit node, their currents will add. Both SMUs in Figure 1 can source current and measure voltage. All of the low-impedance terminals (force and sense) of both SMUs are tied to ground. Table 1 provides an overview of the characteristics of this particular configuration.

To obtain maximum output with this setup, you should set the output currents for both SMUs to the same polarity. Whenever possible, configure one SMU as a fixed source and let the other SMU perform the sweep. During a sweep, the current source's output impedance changes. The DUT's output impedance may also change significantly, such as from a high-resistance off state to a low-resistance on state. Those changing impedances may cause a change in the circuit's overall settling time at each bias point. Although this transient effect dampens, fixing one SMU's source and sweeping the other's usually results in more stable and faster-settling transient measurements, which increase test throughput.

Merging pulse sweeps

Adding current sources when performing a DC power sweep is rather intuitive, but merging pulsed sources is not. Implementing this test method demands extraordinary caution to ensure the safety of test-system operators. You must, for example, provide insulation or install barriers to prevent user contact with live circuits.

Additional protection techniques will prevent damage to the test setup or the DUT. Furthermore, multiple pulses must be tightly synchronized (with nanosecond precision) so the power supply won't apply power to DUTs that are not yet turned on, as this could damage them.

Figure 2. Four SMUs, all producing 10 A across a precision load, add currents to produce a 40-A, 300-µs pulse.

Figure 2 shows the voltage across a high-power precision resistor (0.01 Ω, ±0.25%) that was produced by a single SMU at 10 A for 300 µs and by four SMUs at 40 A for 300 µs. The 10-A pulse, measured with an oscilloscope, shows a waveform of 0.1 V (10 A × 0.01 Ω). The 40-A pulse produced a waveform of 0.4 V magnitude. The 40-A pulse is delayed relative to the 10-A pulse because of the longer rise time needed to achieve that amplitude.

The curves in Figure 3—which were taken from a test of a P-N diode—show that DC and pulsed-current sweeps produce identical increases in current. The programmed pulse sweep resulted from a combined four SMUs. Note the correlation of the single-SMU DC sweep up to 3 A, the single-SMU pulse sweep up to 10 A, and the four-SMUs pulse sweep that produces 40 A. This experiment verified the validity of combining four SMU channels and pulsing to achieve 40 A on two-terminal devices (resistor and diode). With certain modifications, this technique is equally valid when applied to testing a three-terminal device, such as a high-power MOSFET.

Figure 3. Pulsed and DC current sources produce overlaying curves as the current rises.

To maximize accuracy and precision, you can apply several measurement techniques. You can use an SMU's measure functions to read the actual value of the applied voltage. The programmed value may not be the same as the voltage actually applied to the DUT. With multiple SMUs in parallel, the source offsets may add and become quite significant, so you need to measure the actual voltage output, not just the voltage that you've programmed.

By using four-wire (Kelvin) measurements for high-current testing, you can bypass the voltage drop in the test leads. The additional two wires bring high-impedance voltage sense leads to the DUT.

With very little current flowing into the sense leads, the voltage seen by the sense terminals is virtually the same as the voltage developed across the unknown resistance. At 40 A, even a small resistance such as 10 mΩ in the test cable can produce a 0.4-V drop over the length of each test lead. If an SMU forces 1 V at 40 A through a cable with a 10-mΩ resistance, two test leads will result in 20-mΩ resistance. With two-wire measurements, the total voltage drop will be 0.8 V.

Making four-wire measurements will result in significantly better accuracy on both the sourced and measured values. Four-wire measurements eliminate the voltage drop in the current-carrying wires that can affect the measurement. "Cabling and test-fixture considerations," explains how to minimize errors and maximize safety when making high-current measurements.

Connecting power sources to DUT nodes

When using multiple power sources, don't use more than one voltage source at each DUT node. Many test sequences use voltage sweeps that force a voltage across a DUT and measure current. In the case where more than one SMU is connected in parallel to a single terminal of the device, you might decide to set all sources to voltage-source mode and measure the DUT's current. Before you do that, consider these factors:

• A DUT can have a higher impedance than an SMU that is in voltage-source mode. The DUT's impedance can be static or dynamic, changing during the test sequence.
• Even when all SMUs in parallel are programmed to output the same voltage, small variations between SMUs related to the instruments' voltage source accuracy mean that one of the SMU channels will be at a slightly lower voltage (millivolt order of magnitude) than the others.

Figure 4. (a) Connecting two voltage sources together can result in damage to the lower-voltage source. (b) Resistors reduce the current, and (c) a diode prevents current from entering a voltage source.

If you connect three voltage sources in parallel to one terminal of a DUT with each operating at near-maximum current and the DUT is in a high-impedance state, then all current will go to the voltage source that produces the lowest voltage, which will most likely damage that voltage source. If you connect SMUs or other power sources in parallel to a single terminal of a DUT, configure only one to source voltage. The others should source current.

Figure 4 shows incorrect and correct configurations for connecting SMUs in parallel. An incorrect configuration (Figure 4a) could allow high currents to damage the SMU that sources a slightly lower voltage. The configuration in Figure 4b doesn't run the same risk of instrument damage, but it introduces additional error into the measurement that you must account for in the system's error budget. The "hybrid" approach in Figure 4c prevents SMU damage and lets you add current sources as needed to reach the application's current requirements.

When you connect two SMUs with the same output capacity in parallel to a single node in the circuit, one SMU must be able to sink all of the current produced by the other. This condition can occur, for example, when one of the leads breaks contact with the DUT (such as when the lead is accidentally disconnected or a contact isn't made properly). Thus, there will be a short period during which one SMU must sink all the current from the other. The SMU that will be forced to sink current is the one with the lowest voltage or lowest impedance, most likely the one sourcing voltage.

Adding a diode

To protect the signal input of the SMU forcing voltage, you can add a diode such as a 1N5820. A diode offers a much faster response than a fuse and has a much smaller forward voltage drop (typically around 1 V) than a resistor.

To be truly safe when using this method, use a diode to protect all the SMUs in the configuration. If the DUT goes into a high-impedance state, the current sources will try to force their current into the voltage-sourcing SMU, but the diode will prevent that from happening. Without the diode, the current-sourcing SMUs would increase their output voltage until they either forced all their current through the voltage source SMU or they reached their voltage limit. If they reached their voltage limit, the current sources would go into compliance and become voltage sources themselves.

Compliance (such as voltage compliance for a current source) occurs when the output voltage of the current source reaches its voltage limit. The current source "complies" with the voltage limit and becomes a voltage source whose voltage is programmed to the voltage limit. If this happened, you'd have multiple voltage sources in parallel. Even if their voltage limits were set to exactly the same value, their outputs would likely be slightly different and they would damage each other.

Putting a diode on each and every SMU in the configuration in Figure 4c has some consequences. First, adding the diode means that this test setup can only source power but not sink it, because the diodes prevent current from passing into the SMU. Second, to obtain maximum output voltage, you need to use four-wire connections on the current sources around the diode, because the voltage drop across the diode may cause the current sources to reach compliance prematurely. At these current levels, the typical voltage drop across a diode is about 1 V.

Cabling and test-fixture considerations When using an SMU for pulsed-power tests, you must configure your cabling and connections in a way that minimizes resistance, capacitance, and inductance between the DUT and SMU. To minimize resistance, use heavy-gauge wire wherever possible, particularly between the power sources and the test fixture. The gauge required will depend on the level of current. Cabling that must carry 40 A, for example, needs 12-gauge wire. For guidance on choosing cabling for higher current levels, refer to wire-gauge tables such as the one available at www.powerstream.com/Wire_Size.htm. Check the "Maximum amps for chassis wiring" column in the table to find the wire gauge needed to carry the level of current you need. Low-resistance cabling is critical to preventing instru­ment damage. Choose cables with resistances of 30 mΩ/m or less for 10-A pulses. Heavy, low-resistance cables will minimize voltage drops between current sources and DUTs. Keep cable lengths as short as possible, and always use low-inductance cables (such as twisted-pair or low-impedance coax types). Check the SMU's voltage output headroom specification for details on the maximum voltage drop allowed between the source and sense leads. Divide this spec by the desired output current level to get the maximum resistance allowed in your cabling. Although many believe guarding can minimize the effects of cable charging, this is typically more of a concern for high-voltage testing than for high-current testing. Keep four-wire Kelvin connections as close to the DUT as possible; every millimeter makes a difference. You should use an SMU's readback capability to measure voltage. For best results, always use the measurement function that corresponds to your source. The current-sourcing SMU voltage readings may vary because of the connections, and the measurement may differ from what reaches the DUT. Use high-quality jacks on your test fixture. For example, some red jacks contain high amounts of ferrous material to produce the red coloring, which can lead to unacceptably high levels of leakage due to conduction. The resistance between the plugs to the case should be as high as possible, always more than 1010 Ω. Many published test setups recommend adding a resistor between the SMU and the gate pin when testing a FET or an IGBT. These devices tend to oscillate when you apply large amounts of pulsed current through them. Inserting a resistor on the gate will dampen these oscillations and stabilize your measurements. Because the gate draws little current, the resistor won't cause a significant voltage drop. If your tests require voltages in excess of 40 V, then you need interlocks on the test fixture and SMUs. Install and operate the interlocks in accordance with normal safety procedures. Ensuring operator safety Many electrical test systems or instruments are capable of measuring or sourcing hazardous voltage and power levels. It's also possible, under single fault conditions (such as a programming error or an instrument failure), to output hazardous levels even when the system indicates no hazard is present. These high levels make it essential to protect operators from any of these hazards at all times. To ensure operator safety, take these steps: Verify the operation of the test setup carefully before it is put into service. Design test fixtures to prevent operator contact with any hazardous circuit. Make sure the DUT is fully enclosed to protect the operator from any flying debris. Double insulate all electrical connections that an operator could touch. Double insulation ensures the operator is still protected, even if one insulation layer fails. Use high-reliability, fail-safe interlock switches to disconnect power sources when a test fixture cover is opened. Where possible, use automated handlers so operators do not require access to the inside of the test fixture or have a need to open guards. Provide proper training to all users of the system so they understand all potential hazards and know how to protect themselves from injury. It's the responsibility of the test-system designers, integrators, and installers to make sure operator and maintenance personnel protection is in place and effective.—David Wyban