In this thesis, based on the analysis of the energy efficiency of CMOS image
sensors, it discusses high-speed, high-efficiency CMOS image sensors and ultralow-
power CMOS image sensors for always-on applications.
The first study proposes a high-speed...
In this thesis, based on the analysis of the energy efficiency of CMOS image
sensors, it discusses high-speed, high-efficiency CMOS image sensors and ultralow-
power CMOS image sensors for always-on applications.
The first study proposes a high-speed CMOS image sensor utilizing a time-
Stretched (TS) Single-Slope (SS) ADC. In this image sensor, a two-step architecture
is implemented by using a time-stretcher with minimum column variation to extend
the time residue of the SS ADC for 16 times. The number of conversion cycles of
the SS ADC has been greatly reduced, and the power is significantly reduced by
using a lower clock frequency for the column counter. The sensor achieved a
horizontal conversion time of 4 μs with a random noise of 1.95 erms at 500 fps with
a 100 MHz counter clock and consumes only 76 mW. This corresponds to an energy
efficiency FoM of 0.96 e-∙nJ.
The second study proposes a 640 × 640 fully dynamic CMOS image sensor for
continuous operation. It proposes a dynamic pixel source follower that samples the
output signal into a column parasitic capacitor, and a dynamic single-slope ADC
based on a dynamic bias comparator and a two-step counter. The sensor achieved
0.3 % peak nonlinearity, 6.1 erms random noise and 67 dB dynamic range. Power
consumption is only 2.1 mW at 44 fps, and in sub-sampling mode it consumes only
140 μW at 5 fps. The sensor achieved an energy-efficient FOM of state-of-the-art
of 0.71 e-∙nJ.