Electrostatic discharge can damage sensitive semiconductor components, so understanding and standardizing ESD test procedures is a significant focus in the industry. Among the widely adopted models for ESD testing, the Human Body Model (HBM) is especially relevant, as it simulates the effect of human contact, a common source of ESD. However, RF GaN HEMTs present unique challenges in HBM testing, as their failure voltages can vary widely due to differences in testing methodology, sample sizes, device configuration, and test setup.

This study compares two primary testing methods - Step Stress Testing and One Level Testing - and evaluates how each affects the failure classification of GaN HEMTs. By examining ESD test results across multiple samples, batches, and testing setups, researchers aim to identify the most reliable approaches for accurately classifying GaN HEMT devices' ESD robustness.

GaN HEMTs are widely used in high-power applications due to their efficiency and high breakdown voltages. However, when subjected to ESD testing - especially the Human Body Model (HBM) test, which simulates the electrostatic effects of human handling - GaN HEMTs can show wide-ranging failure voltages. This variance complicates their classification, potentially leading to inaccurate assessments of their reliability in real-world conditions.

Two main approaches are typically used for ESD testing:

1. Step Stress Testing: voltage is gradually increased until the device fails. This method gives insight into the overall robustness but may harden the device against failure or cause early failures.
2. One Level Testing: devices are tested at a single, high-voltage level, reducing cumulative stress. However, this method can miss some nuances in failure behavior due to limited testing points.

This study focused on applying both methods to different GaN HEMTs from varying batches, assessing how test setup and sample variation impact ESD classification.

1. Lognormal distribution: ESD failure voltages generally follow a lognormal distribution, which helps in quantifying the probability of failure across different voltage levels.
2. Variation across testing methods: One level testing often resulted in lower failure rates than predicted by step stress testing. This discrepancy suggests that cumulative stress in step stress tests may skew results by "hardening" devices.
3. Batch and setup impact: Variations in failure distributions were observed not only between devices but also between batches and testing locations. Minor differences in test setups or device configurations can shift failure thresholds, highlighting the importance of standardized procedures.
4. Sample size matters: Small sample sizes (e.g., 3 per voltage level, as recommended in JS-001 standards) could lead to inconsistent classifications. Using more samples, as suggested in automotive and industrial standards like AEC-Q101, provides more reliable insights.

To improve the accuracy of ESD testing and ensure consistent classifications:

• Combine testing approaches: using a mix of one level testing at select voltage points above a known failure baseline can give a more comprehensive view.
• Increase sample size: testing with at least 10–30 samples at various voltage levels, especially during technology development, helps capture the variability inherent in GaN HEMTs.
• Tailor testing to device type: since different GaN HEMTs exhibit different responses, testing protocols should account for specific device characteristics.