Power Semi Devices Demand Different Characterization Solutions at Each Stage of Development
The need for improved energy efficiency is driving demand for better high power semiconductor devices. However, as their performance has improved, characterizing these devices has become increasingly complex. This article uses a typical design and development workflow to highlight important measurement challenges associated with each stage.
The need for improved energy efficiency is driving demand for better high power semiconductor devices. However, as their performance has improved, characterizing these devices has become increasingly complex. This article uses a typical design and development workflow to highlight important measurement challenges associated with each stage.
Figure 1. Typical power semiconductor design and development lab-to-fab workflow.
The Design Stage
Today, the fabrication process and the devices themselves are often being developed simultaneously. Often, the device designer will direct the fab to create simple sub-circuits (simple Schottky and bipolar diodes, even resistors and capacitors) to test some element of the process. This usually involves I-V (current-voltage) characterization. Given that exceptional measurement accuracy and sensitivity are typically not critical to such applications, curve tracers are suitable for this task.
The Experimental Tests and Fab Device Stages
Next comes fabricating a real device and taking simple I-V (current-voltage) curves to determine basic device characteristics. Again, a curve tracer can be a simple and efficient choice for this task. It allows the designer to apply a controlled voltage carefully, read back the current to determine if the junctions and interconnects designed actually exist, and discover the typical breakdown voltage or ON current the device can handle. However, connecting a curve tracer safely to a probe station can be complicated. Curve tracers were developed to test packaged parts, but testing devices on the wafer directly eliminates the cost and time of packaging the part. High power source measure unit (SMU) instruments, including Keithley’s Models 2651A and 2657A High Power System SourceMeter® instruments, provide safer and more accurate connections, and better control over the voltage and current than traditional curve tracers. The designer can control the instrument using the front panel knob and display, via the LXI-compliant web page embedded in the instrument, or via an ACS Basic Edition application running on a host controller.
Drain breakdown voltage (BVDSS) is one commonly checked parameter during fabrication of experimental devices. When checking BVDSS manually, the voltage is slowly increased until a target current is reached. A faster, more efficient way to make this measurement is to source the target current and read the voltage at which the device settles. This requires an instrument that provides both an accurate, fast-acting voltage meter and a voltage clamp. Because this test is often performed using pulsed test signals, the user must make certain that the voltage has settled to a final value before taking the reading. The newest SMU instruments combine fast settling with fast digitizers to allow the designer to verify these breakdown measurements are valid.
The threshold voltage of the transistor has some interesting test challenges associated with it. For example, one way of measuring threshold voltage is to hold the drain at a fixed bias while sweeping the gate and measuring the resultant drain current. However, the drain actually needs to be in a pulse mode to prevent the device from heating up, which would cause a shift in the device’s characteristics. In any pulse test, in order to make accurate, repeatable measurements, making certain the entire test system has reached a settled state is crucial.
The Sample Test Stage
In some cases, designers use the initial data they obtain to provide feedback to the fab, requesting a tweak of some process parameter, the better to establish the particular device characteristic for which they are looking. Once the designer is satisfied the device is ready, he or she moves on to a much more detailed sample testing phase. In addition to static DC I-V tests, this stage of characterization includes AC impedance parameters, such as various device capacitances (such as CISS, COSS, CRSS). Designers also typically measure several other AC and transient characteristics, including rise/fall time (tr, tf), switching delay (tdon, tdoff) and reverse recovery time (trr).
For power FETs and IGBTs, numerous charge measurements are performed, including total gate charge (QG), pre- and post-threshold gate charge (QGS1, QGS2), drain charge (QGD) and output charge (QOSS). Because of curve tracers’ limited applicability, they have been largely replaced with semiconductor parameter analyzers (SPAs) for these types of measurements. Until recently, even SPAs were limited by insufficient power and capability for such applications. For example, many lacked the speed or power necessary to capture many of the charge characteristics. A difficult test called Dynamic On Resistance (sometimes referred to as “current collapse” in a gallium nitride power HEMT) is one good example. The device must be biased to full breakdown voltage, typically 600V, then switched to full ON current, often as high as 50A, all while monitoring the ON resistance for rapid changes in RDS(ON). These changes can occur as rapidly as microseconds and can continue for milliseconds. Keithley’s high power SMU instruments incorporate numerous digital enhancements, such as high speed digitizers, intended to make it easier for designers to capture important data.
Figure 2. High voltage pulse showing unsettled current transients.
Other equipment used at the sample testing stage includes pulse generators and oscilloscopes, IR cameras, EMF Test, and wafer probers. In some cases, the devices will be packaged to allow for easier testing in a specially designed test fixture rather than at the wafer level.
In statistical sample testing, the number of devices tested and the volume generated usually requires a more automated process. However, automated power device testers tend to be extremely costly and the scope of the tests they can perform is often limited. For example, typical drain leakage currents on a silicon carbide power FET might be less than one nano-amp, but an automated power device tester is often limited to hundreds of nano-amps or even micro-amp capability. One solution is for the power device manufacturer to configure their own automated solutions using higher performance instrumentation, such as high power SMU instruments with automation features like embedded scripting capability and a virtual backplane capability for synchronizing multiple instruments.
The Long-Term (Life) Testing Stages
The next step is a production qualification run. From these wafers, a Long Term Life Test can be set up to validate the reliability of the devices under different conditions designed to emulate the real world. Many hundreds of devices are tested for hundreds or thousands of hours under specific electrical and environmental stress conditions, which makes it essential that the test equipment be flexible and readily reconfigurable. Creating and maintaining parallel test resources and monitoring and managing the masses of data acquired during this phase can easily become overwhelming. Typical parameters tested here include leakage current (IDSS and IGSS), ON resistance (RDS(ON)), and threshold voltage drift (delta VTH).
Correlating the results of tests performed in various test environments is always been complicated due in no small part to the very different test systems (each with its own unique set of capabilities and sources of error) often used in different test environments. Today, a growing number of device manufacturers are striving to consolidate on a uniform set of instrumentation, so that, for example, the leakage current measurements made in the early stages of the device design process can be correlated with those from devices being produced the fab.
Characterizing power devices has always been challenging. The combination of new, higher performance power devices and instruments that lack the capabilities required to test them makes it even more challenging. Fortunately, test equipment companies are starting to develop instruments optimized for power device characterization and test.
