Analog, digital, and RF CMOS ASICs that my students have designed and built. The chips are for molecular measurement, electronic AI computation (bioinformatics and path planning), and communication.
NIST Nanotechnology Accelerator project.
DNA molecules in / electronic bits out. Lots of analog & digital in between. 130-nm CMOS.
As above. But different timing scheme. 130-nm CMOS
As above. Different circuitry. The granddaddy of 'em all. 130-nm CMOS.
Design for IEEE SSCS supported Chipathon project.
Beefier AI for DNA sequencing. 22-nm CMOS.
Deep-learning-enchanced bioinformatics (some analog iniside). 22-nm CMOS.
Electronic bits in / DNA basecalls out (e.g., ...CCGTTAAATTGG...). RISC-V + AI bioinformatics hardware acceleration in between. Linux capable. 22-nm CMOS.
The above, just snug to its package frame.
Robot brain. Same idea as below. Has data converters for sensors (ADCs) & actuators (DACs). 90-nm CMOS.
Micro-bot planner. Implements a nonlinear Lyapunov-based controller processor. Optimized for a trimmed down ISA (MIPS-based) to save power, but still allows algorithm to scale. Intended to operate in subthreshold. 130-nm CMOS.
Fast sampling of clock and random data as expected in high-speed wireline comms. 65-nm CMOS.
3X frequency without the need for an idler (a good thing if you want to keep the cost down). 130-nm CMOS.
Upconversion using a nonlinear cap. No DC power required. Should work at much higher frequencies. 130-nm CMOS.
I don't think people tried these since WWII. The operating frequency was probably too low for lumped components in this case. 130-nm CMOS.
Humble doubler utilizing a nonlinear capacitance. I was surprised these things didn't appear sooner, but GaAs-based multipliers have achieved great performance. This approach gives some hope for tighter integration with other system components. 130-nm CMOS.
Wireless signals that bounce off this thing get phase modulated. It makes otherwise stationary communicators look like they are moving. This prevents wireless communicators from sitting in deep fades for too long and thus enables correction codes to do their job. Measurement results implied 4X improvement in WLAN coverage area. 130-nm CMOS.