• Skip to primary navigation
  • Skip to content

PSyLab

Digital, Mixed-signal and Integrated Power Circuits and Architectures
  • Home
  • Research
  • Publications
  • People
  • Chip Gallery
  • Positions

Research

Research Areas

ENERGY EFFICIENT COMPUTING
RUNTIME CONTROL AND OPTIMIZATION
NEXT-GENERATION CLOCKING ARCHITECTURES
BASEBAND PROCESSING
NEURAL INTERFACES
HARDWARE SECURITY
MACHINE LEARNING

Energy-Efficient Computing : From uW-Sensors to kW-Servers

Thermal Sensing for 3D SoC Technologies

Exploring modular temperature sensor architectures for runtime thermal management in 3DICs.

Computationally Enabled Minimum Energy Point Tracking

A digital technique to determine total energy consumed per computation, inclusive of regulator losses, and dynamically set supply-voltage for minimum energy dissipation subject to minimum performance requirements

Computationally-Controlled LDOs

A (significantly) simplified Model Predictive Control approach to optimizing regulator dynamics

Switched-Capacitor UniCAP

Demonstrating Unified Clock And Power architecture with a low-voltage switched-capacitor integrated voltage regulator.

Buck Converter UniCAP

A unified clock and power regulation architecture for digital SoC domains, virtually eliminating supply noise and temperature-related supply guardbands.

On-Chip Power Supply Measurement

Techniques for measuring on-chip supply noise with millivolt resolution and multi-GHz bandwidth.

Run-time Control and Optimization for Systems-in-Package

Computationally locked PLLs

Reducing PLL lock time for cold-start and relock by an order of magnitude.

Computationally Enabled Minimum Energy Point Tracking

A digital technique to determine total energy consumed per computation, inclusive of regulator losses, and dynamically set supply-voltage for minimum energy dissipation subject to minimum performance requirements

Computationally-Controlled LDOs

A (significantly) simplified Model Predictive Control approach to optimizing regulator dynamics

Next-Generation Clocking Architectures: Synthesis and Distribution

Computationally locked PLLs

Reducing PLL lock time for cold-start and relock by an order of magnitude.

Buck Converter UniCAP

A unified clock and power regulation architecture for digital SoC domains, virtually eliminating supply noise and temperature-related supply guardbands.

Quasi-Resonant Clocking

The first-ever voltage-scalable, frequency-independent and DVFS-enabled resonant clock architecture.

Baseband Processing

Computationally locked PLLs

Reducing PLL lock time for cold-start and relock by an order of magnitude.

Buck Converter UniCAP

A unified clock and power regulation architecture for digital SoC domains, virtually eliminating supply noise and temperature-related supply guardbands.

Quasi-Resonant Clocking

The first-ever voltage-scalable, frequency-independent and DVFS-enabled resonant clock architecture.

Neural Interfaces : Realizing Chronic Low-Latency Neuromodulation

High-Density Neural Signal Recording

A multiplexed neural-recording front-end that exploits neural-signal statistics to achieve a robust architecture with 10X improvement in neural recording density.

Machine Learning: Beyond Accelerators

MATIC

Learning around failures in low-voltage on-chip memories to enable energy-efficient neural network accelerators.

Hardware Security

Energy-efficient, Robust True Random Number Generators (TRNGs)

Digital true random-number generation with imperfect entropy sources.

  • Research
  • Publications
  • People
  • Chip Gallery
  • Positions

Copyright © 2025 · Executive Pro on Genesis Framework · WordPress · Log in