A radhard CMOS image sensor for high-energy physics must solve two demanding problems at once: maintaining function in a radiation environment and moving large volumes of pixel data quickly. A camera-on-a-chip architecture described in a University of Nevada, Las Vegas dissertation addresses these challenges with a megapixel-oriented design, parallel readout and circuit-level radiation-hardening techniques. Its prototype testing demonstrates how sensor architecture, control logic and output interfaces need to work together to support high-speed imaging.
Why combine speed and radiation tolerance?
CMOS image sensors can integrate photodetectors with analog and digital processing, memory, control circuitry and high-speed I/O. This integration enables a compact camera system with digital output. But increasing resolution and frame rate also increases the amount of data that must be converted and transferred. In radiation-intensive applications, the sensor must additionally withstand effects that can alter transistor behavior or increase leakage.
Radiation damage is not a single phenomenon. Total ionizing dose (TID) can create charge in silicon dioxide and contribute to leakage paths, while displacement damage can disturb the silicon crystal lattice. The described design focuses its hardening approach on TID tolerance. Its target applications include high-energy physics and other environments where radiation exposure is a design consideration.
Quadruple column-parallel readout for high pixel rates
The central speed innovation is a quadruple column-parallel readout (QCPRO) architecture. Rather than relying on one converter for a large array, column-parallel designs distribute conversion across the sensor’s columns. QCPRO extends this idea: four readout blocks serve each column, allowing four rows to be processed within one line time. The architecture uses a rolling shutter, in which rows are exposed and read in sequence.
Each readout block includes analog time-interleaving (ATI) sampling, a switched-capacitor programmable gain amplifier (SC-PGA), a 10-bit successive-approximation-register (SAR) ADC and two memory banks. ATI sampling uses two sample-and-hold circuits, each able to perform correlated double sampling (CDS). CDS compares signal and reset-related levels to help suppress reset noise and pixel-level fixed-pattern noise.
The two memory banks support ping-pong operation: while one bank receives converted data from the ADC, the other can deliver data to the output interface. This overlap helps keep conversion and data transfer moving in parallel. High-speed differential Scalable Low-Voltage Signaling (SLVS) transmitters carry the output data toward an external FPGA. The design specifies 1 Gbps per channel for these transmitters.
The dissertation reports a maximum pixel rate of 10.485 gigapixels per second for the proposed architecture and describes a 1,024 × 1,024 array operating at up to 10,000 frames per second as an architectural capability. These figures should be distinguished from the reported measurements of the fabricated small-format prototype, which demonstrated programmable operation from 20 to 4,000 frames per second.
Managing rolling-shutter effects and frame rate
Rolling shutter can introduce image artifacts when a target moves during sequential row exposure. The architecture reads multiple lines in parallel and uses oversampling to help relax these artifacts. This does not make rolling shutter equivalent to global shutter: the rows are still captured in sequence. Rather, the design uses parallelism and timing to support rapid acquisition while managing the consequences of sequential readout.
Frame rate is programmable by adjusting three parameters: the input clock frequency, the region of interest and the sequencer counter size. Region-of-interest selection can reduce the portion of the array being read, while clock and sequencer settings provide additional control over timing. For system designers, these controls illustrate how imaging requirements can be balanced against readout workload and available data paths.
Radiation-hardening techniques in the pixel and periphery
The design was implemented in a 0.18 µm CMOS process. Radiation tolerance combines enclosed-layout transistors in pixel cells with P-type guard rings in peripheral circuitry. Enclosed geometry changes the transistor layout to reduce sensitivity to radiation-induced leakage at device edges. Guard rings are used to interrupt leakage paths between devices in the peripheral circuits.
The pixel uses a pinned photodiode, selected to support low dark current, together with CDS for noise suppression. The reported pixel-cell layout is 20 × 20 µm². Two array formats were designed: a 128 × 1,024-pixel small prototype and a 768 × 1,024-pixel larger version. The system-level hardware and FPGA firmware implementation and testing described in the dissertation focus on the small-format prototype.
What prototype testing established
Measurements of the small-format camera system validated key functions in the readout path, sequencer, on-chip phase-locked loops and SLVS transmitters. The programmable frame-rate feature was demonstrated by checking sequencer control outputs at different rates. TID testing showed that the pixels continued to operate after exposure up to 125 krads.
These results provide evidence for the tested prototype and its specified radiation exposure; they should not be read as a guarantee for every operating condition or for the larger array. In a deployed system, engineers still need to consider the intended radiation environment, image requirements and the interface to downstream acquisition hardware.
Practical implications for imaging system design
The design shows why high-speed radiation-tolerant imaging depends on more than the pixel array. Parallel analog sampling and ADC conversion address the readout workload; memory banking and SLVS links support data movement; programmable timing offers flexibility; and layout techniques target radiation-induced leakage. For system integrators and technical decision-makers, the key lesson is to assess these elements as one coordinated signal chain.
A radhard CMOS image sensor can extend imaging capability in demanding environments, but performance figures need to be interpreted in context: architectural targets, fabricated formats and measured prototype results are not interchangeable. Clear requirements for frame rate, resolution, radiation exposure and data handling are essential when translating a sensor design into an application.