Alex Lidow Ph.D., Robert Strittmatter Ph.D., Shengke Zhang Ph.D., and Alejandro Pozo Ph.D. Efficient Power Conversion
An automotive application using GaN power devices in high volume is lidar (light detection and ranging) for autonomous vehicles. Lidar technology provides information about a vehicle’s surroundings, thus requiring high accuracy and reliability to ensure safety and performance. This article will discuss a novel testing mechanism developed by Efficient Power Conversion (EPC) to test eGaN devices beyond the qualification requirements of the Automotive Electronics Council (AEC) for the specific use case of lidar.
Methods for Lidar
Two significant forms of lidar dominate the lidar industry today: direct time-of-flight (DToF) and indirect time-of-flight (IToF). Typical DToF lidar sends individual pulses, and times the reflection to compute the distance to the target. The more photons that are emitted, the farther away objects can be detected. The narrower the pulse, the higher the resolution making these systems ideal for long-range liar systems. IToF lidar works by comparing the phase of transmitted and reflected pulse trains (figure 1). This type of system is primarily used in the medium range, from less than 1 meter to tens of meters.
GaN for Lidar
eGaN FETs and ICs are widely implemented in both DToF and IToF lidar circuits for autonomous vehicles, where they offer several key benefits:
· Faster switching for shorter pulses and better range resolution
· Smaller footprint which enables high power density, low inductance, and compact solutions
· Higher efficiency at higher pulse repetition rates
In a typical DToF lidar application, the GaN device delivers short, high-current pulses, on the order of 1−5 ns, which drive a laser diode to generate narrow optical pulses. The peak currents are usually substantially greater than 50% of the FET pulse-current rating. The pulse duty cycle is typically low, and the pulse repetition frequency is in the range of 10 to 100 kHz. When not being pulsed, the GaN device is in the OFF state, exposed to a certain drain bias.
This stress condition is unprecedented for a power device, making it difficult to predict lifetime in operation by relying on conventional DC reliability tests such as high temperature gate bias (HTGB) or high temperature reverse bias (HTRB). The simultaneous high current and voltage during a pulse raises concerns about hot-carrier effects, potentially leading to threshold voltage (VTH) or on-resistance (RDS(on)) shifting within the device. In addition, the cumulative effect of repetitive high current pulses raises the specter of electro-migration leading to degradation of the solder joints.
Even GaN-specific tests, like the hard-switching reliability testing employed by EPC, do not effectively emulate the stress conditions in a lidar circuit. To address these concerns EPC initiated a novel test method in collaboration with key lidar customers. This lidar reliability testing is part of EPC’s “Beyond AEC” initiative, a series of GaN specific stress tests that go beyond the conventional reliability tests developed for MOSFETs as part of AEC-Q101 standard.
Long-Term Stability Under High Current Pulses
The concept of this test method is to stress parts in an actual lidar circuit for a total number of pulses well beyond their ultimate mission profile. The mission profiles for automotive lidar vary from customer to customer. A typical automotive profile would call for a 15-year life, with two hours operation per day, at 100 kHz pulse repetition frequency (PRF). This corresponds to approximately four trillion total lidar pulses. Some worst-case scenarios might call for as many as 10−12 trillion pulses in service life.
By testing a population of devices well beyond the end of their full mission profile while verifying the stability of the system performance and the device characteristics, this test method directly demonstrates the lifetime of eGaN devices in a lidar mission.
Click image to enlarge
Figure 2: Long-term stability of pulse width (bottom right) and pulse height (top right) over 13 trillion lidar pulses
Test Methodology and Results
To achieve the large number of pulses, parts are stressed continuously at a pulse repetition frequency (PRF) much higher than in typical lidar circuits. The test circuit is based on EPC’s popular EPC9126 lidar application board. Experimental details of the testing procedures are provided in Appendix B of EPC’s Phase 11 Reliability Report.
For this study, two popular AEC grade parts were put under test: EPC2202 (80V) and EPC2212 (100 V). Four parts of each type were tested simultaneously. During the stress, two key parameters were continuously monitored on every device: (i) peak pulse current and (ii) pulse width. These parameters are both critical to the range and resolution of a lidar system.
Figure 2 shows the results of this test over the first 13 trillion pulses. Note that there is no observed degradation or drift in either the pulse width or height. The cumulative number of pulses corresponds to a typical automotive lifetime. While this is an indirect monitor of the health of the eGaN device, it indicates that no degradation mechanisms have occurred that would adversely impact circuit performance.
Click image to enlarge
Figure 3: Long-term stability of RDS(on) and VTHduring lidar reliability testing. These parameters are measured at six-hour intervals on every part by briefly interrupting the lidar stress
To gain better visibility into the eGaN device parametric stability over time, the test system interrupts lidar stress every six hours to measure the RDS(on) and device threshold VTH. After this brief parametric measurement, the parts are returned quickly to lidar stress mode. The results are shown in Figure 3. Both parameters show excellent stability over the duration of the test. The stability indicates that lidar stress is relatively benign to eGaN devices.
Short, high-current pulse (lidar) testing of eGaN devices shows they are very reliable in a lidar application over a typical automotive lifetime. To date, no failures mode or parametric degradation has been observed. GaN power devices, already in volume production in lidar applications, provide the accuracy and robustness needed to ensure the performance and safety required for autonomous navigation.