by Ryo Takeda and Alan Wadsworth, Keysight Technologies
Investments in power semiconductor devices continue to grow. Performance and reliability also improve each year through the employment of new structures and new wide bandwidth materials such as SiC and GaN.
Particularly with respect to these new power devices, it is important to evaluate operating properties. Sometimes device makers don’t provide figures for all the parameters necessary for designing power circuits. It’s the responsibility of the device user to characterize power devices for use in designs that employ them. Similarly, product assurance engineers and incoming inspection technicians must fully characterize the power devices used in their products.
Energy consumption is a factor in the use of any power device. Thus, it is important to understand the sources of energy loss in power devices. These losses can be divided into three main components: conduction loss, switching loss and driving loss. The interaction between input stimuli and basic device properties (on-resistance, gate resistance and junction capacitances) determines the level of loss. Thus, the evaluation of device characteristics generally involves examining these parameters. In addition, as power device performance improves, on-resistances drop and operational currents rise.
The typical way of evaluating semiconductor operational qualities has been with a curve tracer. As a brief review, a curve tracer basically applies a swept (automatically continuously varying with time) voltage to two terminals of the device under test and measures the amount of current the device passes at each voltage. This voltage-versus-current graph displays on an oscilloscope screen. The main terminal voltage can often be swept up to several thousand volts, with load currents of tens of amps available at lower voltages.
When testing three-terminal devices, such as transistors and FETs, the curve tracer connects to the control terminal of the device, such as the base or gate terminal. For transistors and other current based devices, the curve tracer steps the current to the base or other control terminal. By sweeping the input voltage or current through a range of main terminal voltages, the curve tracer generates a group of V-I curves for each step of the control signal. This group of curves makes it very easy to determine, for example, the gain of a transistor.
One difficulty with curve tracers arises because semiconductor devices have internal sources of energy storage. These internal sources can degrade the accuracy of the V-I curves the curve tracer generates. In addition, there can be thermal effects within the semiconductor device that arise because of internal heating when the device handles current. It can be difficult to compensate for both the heating effects and internal storage when measuring device parameters. These effects can be particularly substantial at the VHF switching frequencies and high power levels at which wide bandgap semiconductors sometimes operate.
However, the use of VHF and higher switching frequencies brings many benefits in modern electronics. The circuits can be physically smaller than those operating at conventional power supply frequencies on the order of 1 MHz. For example, smaller power electronics make it possible for mobile devices to squeeze into more compact packages and help lighten up hybrid/electric vehicles for better fuel economy.
As frequencies rise, switching loss eventually exceeds conduction loss. Switching loss strongly depends on the power device’s junction capacitances (parasitic capacitive elements between gate, drain and source), which in-turn depend strongly on the drain-to-collector voltage.
Conventional test equipment cannot measure capacitance at dc biases greater than 100 V. For example, power loss from the charging and discharging of parasitic capacitance from the drain-to-source and from the source-to-the-gate comprise the primary components of switching loss in the case of a resonant converter employing FETs. The parasitic capacitances have voltage dependency in the nano-farad range because the power device’s depletion region modulates the applied varying operational voltages. These capacitances are conventionally measured by an LCR meter. But the maximum voltage of an integrated LCR voltage source is limited to around ±40 V. So, most device data sheets do not include capacitance measurement data with more than ±40 V bias.
Circuit designers traditionally use curve fitting to estimate capacitance in their design work for voltages above 40 V. However, curve fitting is impractical when the semiconductor device is built with trench or super-junction structures. In addition, complicated manufacturing processes induce additional variations in device performance, for example, a capacitance gap between the high side and low side of a FET. Designers have to identify performance differences when selecting a power device and when analyzing failures. Accordingly, it has become essential to characterize device capacitance from actual chip and module-level measurements.
The difficulty of getting accurate high-voltage measurements has prompted some device manufacturers to create their own in-house test systems. Unfortunately, the process of creating the connections needed to make correct input, output or reverse-transfer capacitance at high drain/collector voltage biases is complicated. This often causes connection mistakes and reduces productivity.
Gate charge (Qg) is another important parameter relevant to power loss at high switching frequencies. Gate charge is related to the parasitic junction capacitances. Its measurement can be complicated because, like junction capacitances, it is dynamic. It is also inversely proportional to drain-source resistance. Again, there are no commercially available bench-top solutions to measure Qg for high-power devices. Some device manufacturers have devised in-house Qg test systems. However, these systems require ultra-high-power sources that can simultaneously supply high voltage and high current.
Besides being dangerous, Qg test systems constructed using ultra-high-power sources aren’t particularly accurate. The poor accuracy arises because it is difficult to supply large, constant currents to the gate terminal without reducing both the device turn-on time and the measurement accuracy. In addition, there can be measurement disruptions because of unpredictable influences from stray inductance in the test circuit or voltages. These voltages can get superimposed on the gate due to the gate resistance Rg. Furthermore, if the constant gate current is set too low, then the turn-on time rises. This results in more power applied to the DUT and creates the possibility of device damage.
Finally, it can be difficult to design a circuit that can supply the appropriate constant current for the measurement. For all of these reasons, Qg test systems designed in house have trouble making reproducible measurements. If deployed to multiple sites across, the world they can often have correlation issues.
Reliability is important for all types of power electronics applications, but especially so for applications in automotive, aerospace and defense, rail or medical equipment. Power devices used in these applications must be reliable because their failure could cause the loss of human life.
Device breakdown voltages and leakage currents are obvious reliability concerns. In addition, designers must understand how devices behave over temperature for two reasons: Power devices innately operate at high temperatures because they self-heat, and power electronic products often operate in harsh environments of both extreme heat and cold. Thermostatic chambers are often used for characterizing operation over temperature, but they have drawbacks. Most problematic is that they usually require long cable extensions run from the measurement resources to the chamber. This arrangement introduces both residual resistance and inductances that can cause oscillations during the voltage stepping involved in making measurements. It also takes a lot of time and effort to monitor the test system for long time periods. In the real world, many people just abandon device characterization across temperature, though it is important for product reliability.
Although not strictly related to reliability, counterfeiting is another important concern for power devices. The accidental use of counterfeit semiconductors can reduce credibility and cause financial loss. The possibility of counterfeiting forces power-device users to evaluate inversely related parameters (such as on-resistance and capacitance) to ensure they meet all specifications.
Of course, most power circuit designers are not device physics experts and do not regularly evaluate power devices. The learning curve on power device test equipment typically entails some wheel spinning on the part of the circuit designer.
Circuit designers typically follow a three-step process when it comes to working with power devices: 1) Measure device characteristics to select an appropriate device, 2) Use simulators to design and verify circuit behavior, and 3) Build the circuit and evaluate it. There are iterations involved in the circuit design, assembly and evaluation process until the circuit meets its specifications. Unfortunately, it can take many cycles to complete the hardware verification loop if the simulation is off because the device data is inaccurate or the device models lack capacitance parameters that are critical at high switching frequencies.
Consequently, instrumentation suppliers have developed specialized equipment optimized for characterizing modern power semiconductors. An example is the Keysight Technologies B1506A.
The B1506A can evaluate all key power device parameters. It displays device tests in data-sheet format and lets users measure key device parameters without specialized training. Pass/Fail capabilities, as well as data-sheet-format report generation, are also available. Device evaluation work that previously took several days to a week can now take place in less than an hour. The B1506A has a wide current/voltage range (up to 1,500 A/3 kV) to handle a broad range of power devices. Similarly, the B1505A Power Device Analyzer/Curve Tracer is available when even higher voltages (up to 10 kV) are involved.
The B1506A is more accurate than conventional test equipment. It has traceability to international standards for both low current and voltage measurements and also for high current (>1,000 A) and high voltage (up to 3 kV) measurements. These capabilities are especially valuable for SiC or GaN devices, which have small leakage currents in the picoamp range. The B1506A can provide data that will be consistent, not only within a given company, but also across multiple companies in different locations. Moreover, the B1506A has a simple DUT connection scheme designed to help eliminate connection errors. It employs TO-type device connections that don’t require cables. IGBT modules connect easily with furnished cables and connectors. For other types of packages, a universal-type fixture is available.
The B1506A also supports the characterization of device parameters across temperature. If devices don’t require characterization at cold temperatures, a thermal plate can sit in the test fixture.
An optional adapter that interfaces with an inTEST Thermostream system allows a wider temperature range (from -50° C to more than 200° C). It minimizes the test cable length which maximizes current carrying capacity and reduces the risk of oscillation.
Designers must measure device junction capacitances at thousands of volts of dc bias to fully characterize modern power electronics. In addition, it is essential to evaluate both IV and CV characteristics to to detect counterfeit or substandard devices. However, the connection scheme for making these measurements can be complicated because it involves complex connections utilizing dc-blocking capacitors and ac-blocking resistors. The B1506A handles this job using test fixture internal circuitry that configures itself correctly for each measurement. This eliminates the need for the external RC network. Typical parameters measured this way include input and output capacitance, gate-drain and gate-source capacitance, and reverse-transfer capacitance.
Gate charge (Qg) is notorious for being hard to measure. Often measurement equipment situated at different locations will record different values for Qg from the same device. The B1506A can perform gate charge measurements at up to 3 kV of drain-to-source bias with extremely high repeatability. It uses an innovative technique that measures gate charge in two passes. The first measurement takes place at high current with low voltage; the second at high voltage with low current. This technique reduces the power needed to measure Qg and boosts measurement safety.
Because the B1506A uses SMU (Source Monitor Unit) technology with internal feedback, it can easily control the gate current. There is no need to select an optimal gate resistor or to create a special driving circuit for each device. The B1506A can make accurate and reproducible Qg measurements merely by adjusting gate turn-on time.
The B1506A can measure all basic device parameters (IV, CV, Rg and Qg), so it can also easily calculate all contributors to power loss (such as conduction and switching loss).
Quantifying power loss is extremely important for selecting the optimal device in a power electronics circuit. Moreover, understanding junction capacitance as a function of dc bias, in addition to IV device behavior, greatly improves the accuracy of circuit simulation results. The net result is a drastically improved power electronics circuit design process.
Power device manufacturers often have broad measurement requirements (such as on-wafer test, ultra-high-voltage test up to 10 kV, GaN current collapse characterization and so on) beyond those of circuit designers. The B1506A isn’t the best choice for such demands. In this case, the B1505A is better because it has a modular architecture scalable up to 10 kV/1,500 A with leakage current measurement capabilities down to the sub-picoamp level. Thus, the choice between the B1505A and the B1506A depends on a user’s particular measurement application needs.