Behavioral Modeling of CMOS SPADs Based on TCAD Simulations

• Autores: Juan Manuel López Martínez
• Directores de la Tesis: Ángel Benito Rodríguez Vázquez (dir. tes.), Ricardo Carmona Galán (dir. tes.)
• Lectura: En la Universidad de Sevilla ( España ) en 2022
• Idioma: inglés
• Número de páginas: 150
• Enlaces
  • Tesis en acceso abierto en: Idus
• Resumen

  • SPAD stands for Single Photon Avalanche Detectors. SPADs are photodiodes structurally similar to those used in conventional image sensors. However, while the conventional ones are biased in the low voltage zone within the inverse region, where there is no gain, the SPADs are biased in the avalanche region. In this region, specifically, when operating in Geiger mode, a single photon is enough to start a chain reaction that manifests as a current pulse. This pulse’s sharp, active edge encodes the time instant when the photon was detected, which supports the use of SPADs to measure Time-Of-Flight and, from these, distances to the objects focused on the sensor. SPADs are therefore suitable for capturing 3D images, that is, for estimating the depth of objects in an image, without resorting to binocular vision or interferometry techniques. They can also estimate light intensities and therefore capture 2D images based on counting pulses. For its many applications, there has been an interest in developing models from these devices in recent years. During this time their capabilities have been better understood, and that knowledge has refined their design. Still, the structure of the SPAD models developed in the literature has been simplistic until now. There is no mention of a guard ring or a second junction contributing to their capacitance or other complex SPAD structures. In other words, all previous models only consider a simple SPAD composed of a sole junction. Our methodology overcomes this drawback by pursuing physically-consistent models, i.e., models that take into account the SPAR inner structure and map the underlying physics on a VERILOG-A description. In the quest of this challenge, this Thesis proposes a workflow including the following points: Select an actual SPAD structure, described at physical level, which model is targeted. Simulate its fabrication process with Athena from the Silvaco tools suite to feed a TCAD simulator with an accurate model structure that can include fabrication defects. Extract its key parameters with TCAD simulation of the device, performed with ATLAS, from the Silvaco TCAD tools suite. Combine physical parameters and analytical descriptions and the data extracted from the TCAD simulations to build an accurate VERILOG-A model. Besides using this physically-consistent methodology, our models embed the following new features: Inclusion of the contributions to the dark count rate from the TAT processes. New approximation to the inclusion of the Band-To-Band Tunneling (BTBT) contribution to dark counts. Inclusion of defect data from spectroscopy results about traps and deep-level traps. Simulation of the SPAD self-heating and the SPAD dynamic behavior with the temperature. Simulation of the time response and photon-timing jitter. Crosstalk analysis of the models. This Thesis explains the proposed method and the approach followed to include the above-mentioned features.