The High Speed Electronics Laboratory (HSEL) was established in 1997 by Professor M.C. Frank Chang.  Over the past 2.5 decades, we have demonstrated worldwide leadership in realizing real-time, reconfigurable ultra-high-frequency (up to terahertz) CMOS Systems-on-chip for broad/multi-band radio, radar, interconnect and AIoT computations by inventing:

• Tunable Artificial Dielectric (AD) that can be synthesized lithographically within CMOS interconnects, to form diffrential-mode transmission-line (DTL) high-Q resonators up to terahertz. By inserting CMOS on/off switches to engage/disengage periodic floating metal strips underneath the DTL, one can even digitally vary DTL’s permittivity (up to x22 times in practice) and phase on-the-fly to create a “Digital Controlled Artificial Dielectric” (DiCAD) for realizing reconfigurable (software-defined) multi-band/mode radios and radars for extended operation bands and resolutions

• A new class of Radio with embedded self-diagnosis/healing and band-selection capabilities with on-chip sensors/controller and DiCAD-DTL to achieve high “performance yield”, and against temperature/ process variations & radiation/aging effects for long-term sustainability of broadband 57-64GHz 4Gb/s radio-on-a-chip

• With DiCAD-DTL VCO & injection locked dividers to realize first fully integrated frequency synthesizer (PLL) at 560GHz (Ref.4) and enable single-chip Spectrometer at 183/540-600GHz for NASA’s Continuity-MLS and ASTHROS missions with reduced size (X100), mass (6.5X) and power (5.5X) to meet stringent payload/power requirements

• Multi(N)-band RF-Interconnect, with DiCAD-DTL harmonic  oscillators, beyond traditional baseband-only interconnect, to achieve multi-channel & bi-directional communications within shared TL or waveguide  (especially, plastic waveguide for terahertz signaling) with unprecedented (N-times) bandwidth, efficiency, with real-time reconfigurability & multi-cast capabilities

• Reconfigurable Streaming Convolution Neuron Network (RCNN) Accelerator that can be reconfigured on-the-fly to compute respective RNN/CNN/ANN network inference with super-high energy efficiency for AIoT and ADAS applications