A new built-in screening methodology to achieve zero defects in the automotive environment
Introduction
The ability to control the random defectivity of a product during its life cycle becomes more and more vital especially in the automotive field, where the drive towards “zero defects” is very stringent.
In particular, gate oxide defects and crystal defects are a major constraint in the development of reliable devices because of the extensive use of large gate and junctions areas for power applications.
According to the usual defective oxide classification, Class A includes such oxides, which are expected to fail (usually due to pinholes) for an applied field strength (EBD) lower than 1 MV/cm. Class B includes the so-called extrinsic oxides that fail for an EBD lower than the intrinsic value, and Class C includes those oxides, which fail at an EBD above 10 MV/cm (for the thickness range of interest in this paper) due to intrinsic dielectric breakdown. At present the required failure rate during the life cycle of the product is assured by properly designed screening procedures before and after packaging of the chip, which force oxides in the Class A and B to fail, without introducing any substantial pre-damaging of the robust subpopulation [1].
Crystal defects (e.g. dislocations and stacking faults) can be either inherently present in the bulk silicon, or generated during some critical process steps like epitaxial growth, implantation, and the formation of shallow trench isolations. During the lifetime of a device, crystal defects can coalesce or act as a gettering site for dopant atoms and contaminations. This can result into the formation of local highly conductive paths, which under some circumstances can heavily affect the performance of the integrated circuit [2], [3].
Nowadays, the most common procedure to screen gate oxide defects at chip level is still the so-called gate stress test (GST), where a high voltage pulse is applied through an automatic test equipment (ATE) to dedicated test pads of the LDMOS gate. Subsequently, the eventual occurrence of the oxide breakdown is detected by the measurement of the leakage current through the gate oxide.
The correlation between the presence of crystal defects and the increase of the transistor leakage current by various orders of magnitude is reported in the literature [2], [3]. For this reason, screening for crystal defects is usually accomplished by measuring the drain leakage current when the device is biased under subthreshold conditions after applying a high voltage pulse to the drain terminal. This procedure is usually defined as drain leakage test (DLT).
The main limitations of these traditional approaches are well-known. On one side they need expensive ATE, they require dedicated contact pads and overhead circuitry inside and outside the chip, they have very limited parallelization capabilities and finally they could be impossible after device packaging. On the other side, burn-in requires expensive equipment (load boards, control electronic, and thermal chambers), needs long testing times due to the limited acceleration factor, has limited accessibility to the points of interest, has to be carried out on finished devices and finally it still requires expensive parametric and functional tests that often exhibit a limited testing coverage. In addition, the numerous manipulations of the devices increase the risk of pre-damaging due to electrostatic discharges.
In order to solve these limitations, a new approach to the screening of defective gate oxides and junctions of LDMOS is presented, which is based on dedicated embedded circuitry to perform on chip the voltage stress and the measurement of the leakage current through the stressed device. Due to the fact that it relies on built-in circuit, the proposed solution can be applied both at chip level and to packaged devices, targeting directly the point of interest. Furthermore, since the whole process is managed by an internal circuitry, it does not require any additional testing equipment and can be run in parallel on a very large number of devices.
In this paper, the traditional GST and DLT approach in use nowadays for several automotive products are presented in Section 2, to point out requirements and limitations. In Section 3.1, the concept behind the novel built-in GST and DLT is defined together with its main constituting blocks. Section 3.2 describes in very detail the circuit solution used. In Section 3.3 the circuit layout and the results of the validation of the proposed design by transient SPICE simulations are proposed.
Section snippets
Traditional gate stress test method
Traditionally, GST is performed accessing the gates of power FET-switches by contacting needles of an ATE probe-card on one or more test pads. A typical configuration of such a circuitry is shown in Fig. 1.
Several HV-LDMOS transistors might share the same gate stress test pad by using decoupling diodes (D4 and D5 in Fig. 1). The zener diode Z is required to protect the gate driver low voltage N-MOS and P-MOS transistors from high voltages. The series resistor R is needed between the gate driver
Working principle
The proposed solution consists of the integration of all the functionality required to perform a GST and DLT into each device. Following this approach, every chip should become responsible for its own stress test. The principles of the implementation of this solution are presented in Fig. 3. In order to save additional pads/pins for stressing, the high voltage, which has to be applied to the LDMOS, is forced via the battery pin VS.
The sections responsible for the built-in GST (BI-GST) and
Conclusions
A novel built-in approach has been proposed to screen out defective gate oxides (BI-GST) and crystals defects (BI-DLT) in Lateral Diffused MOS transistors in integrated circuits for automotive applications. This technique is based on a programmable embedded circuitry for built-in testing, which provides high voltage pulse-stressing and accurate quantitative measurement of the leakage current through the gate oxide or through the drain. The test sequence is controlled by internal logic and the
Acknowledgement
This work was partially supported by the Austrian project ATEMET (Advanced Test Methodology) 813856 for MEDEA+ SPOT-2 (label 2T205).
References (9)
- Ciappa M. Some reliability aspects of IGBT modules for high-power applications. In: Hartung-Gorre, editor. Modeling the...
- et al.
Crystal defects and junction properties in the evolution of device fabrication technology
J. Phys: Condens Matter
(2002) - Siegelin F, Stuffer A. Dislocation related leakage in advanced CMOS devices. In: Proceedings of the 31 international...
- Malandruccolo V, Ciappa M, Rothleitner H, Fichtner W. New on-chip screening of gate oxides in smart power devices for...
Cited by (8)
Investigation on gate oxide reliability under gate bias screening for commercial SiC planar and trench MOSFETs
2024, Materials Science in Semiconductor ProcessingInvestigation of different screening methods on threshold voltage and gate oxide lifetime of SiC Power MOSFETs
2023, IEEE International Reliability Physics Symposium ProceedingsReliability and robustness of SiC power devices - how to ensure the quality level established in the silicon world
2022, 2022 International Power Electronics Conference, IPEC-Himeji 2022-ECCE AsiaGate-oxide reliability and failure-rate reduction of industrial SiC MOSFETs
2020, IEEE International Reliability Physics Symposium ProceedingsDesign and experimental characterization of a new built-in defect-based testing technique to achieve zero defects in the automotive environment
2011, IEEE Transactions on Device and Materials ReliabilityA new built-in defect-based testing technique to achieve zero defects in the automotive environment
2011, Journal of Electronic Testing: Theory and Applications (JETTA)