Access Official Technical PDF

Share Summary:LinkedInEmail

Ultraviolet imaging can help detect chemical substances, monitor processes and identify flames, but separating UV signals from visible-light backgrounds can require additional optical components or capture steps. A filterless UV imaging sensor developed at Tohoku University takes a different approach: it combines pixels with different UV sensitivities to capture UV-selective and visible-light images simultaneously in a single exposure.

The design is a CMOS image sensor (CIS) that extracts UV information from the difference in spectral response between neighboring pixels. It also incorporates a lateral overflow integration capacitor (LOFIC) to combine high sensitivity with a wide dynamic range. For industrial sensing, the approach illustrates how pixel structure and signal processing can perform spectral separation within an image sensor.

How a filterless UV imaging sensor separates light

The sensor uses two pinned photodiode designs: one with high UV sensitivity and one with low UV sensitivity. These pixel types are arranged alternately in a checker pattern. Both respond to visible light over a similar spectral range, while their responses to UV light differ.

That difference provides the basis for separating the signals. The UV-selective image is produced by extracting the differential spectral response between the two pixel types. The visible-light image is obtained from the low-UV-sensitivity pixels. Because both signals are captured at the same time, the system does not depend on switching illumination or taking separate images under changing conditions.

In the fabricated sensor, the high-UV-sensitivity pixels respond over 200–700 nm, while the low-UV-sensitivity pixels respond over 370–700 nm. Subtracting their responses produces a UV-selective range of 200–480 nm. These ranges describe the sensor’s measured spectral sensitivity; they do not mean that each output is limited to a perfectly isolated wavelength band.

Photodiode structure controls UV sensitivity

The pixel designs use the way light penetrates silicon at different wavelengths. UV light is absorbed close to the silicon surface, while longer-wavelength light penetrates farther. By adjusting the photodiode’s doped layers and internal potential barriers, the designers shaped which generated photoelectrons are collected.

In the high-UV-sensitivity pixel, the photodiode structure allows UV-generated charge to be detected while a p-well layer helps limit sensitivity to visible and near-infrared light. In the low-UV-sensitivity pixel, an additional surface N+ layer creates a barrier near the surface. UV-generated electrons in that region can recombine or be drained, reducing the pixel’s UV response. The p-well structure is also used to align the two pixel types’ responses at longer wavelengths, supporting differential extraction.

This is a device-level method of spectral discrimination rather than a conventional optical filter placed in front of the sensor. It requires carefully designed photodiodes and pixel matching, as well as processing that compares the neighboring pixel signals.

LOFIC supports sensitivity and dynamic range

Capturing weak UV signals while accommodating brighter illumination is a design consideration in imaging systems. The CIS uses LOFIC in both pixel types to support high sensitivity and a broad range of measurable signal levels.

For each pixel, the sensor provides a high-conversion-gain signal, S1, using the floating diffusion (FD), and a high-saturation signal, S2, using the FD together with the LOFIC. The reported conversion gain for S1 is 172 µV/e-, and the full-well capacity for S2 is 131 ke-. Combining the high-sensitivity and high-saturation signals yielded a reported dynamic range of 92.3 dB.

Demonstrated performance and image capture

The prototype was fabricated using a 0.18 µm CMOS process with pinned photodiodes. Its effective image array is 640 × 480 pixels, with a 5.6 µm pixel pitch. The reported frame rate is 30 frames per second.

To demonstrate simultaneous UV and visible imaging, the sensor captured male and female white cabbage butterflies under UV and white LED illumination. The three outputs—high-UV-pixel response, visible image from the low-UV-sensitivity pixels, and UV-selective image from differential extraction—were captured in a single exposure without band-pass filters. The UV image distinguished a difference in UV absorption between the two butterflies, while the visible image showed their visible-light appearance.

For applications such as flame detection, food freshness inspection or semiconductor process control, capturing both spectral views together may be useful when visible background conditions vary. A visible image can also provide scene context for selecting regions in the UV image. The reported work presents the sensing capability; application-specific performance would still need to be evaluated in the intended operating environment.

Practical implications for automation systems

For engineers and system integrators, the central trade-off is between optical filtering and sensor-level spectral separation. This design avoids a separate UV band-pass filter and captures two image channels in one exposure, but it relies on a checkerboard of different pixel types and differential signal extraction. The pixel arrangement and resulting image outputs should therefore be considered when evaluating image processing and integration requirements.

The prototype demonstrates a CMOS approach to simultaneous UV-selective and visible imaging, with a measured 200–480 nm differential response and a 92.3 dB dynamic range. For industrial inspection and monitoring, the practical value lies in obtaining UV information alongside visible context without sequential captures. Whether that simplifies a particular system depends on its illumination, sensing target and image-analysis needs.

Access Official Technical PDF

About The Author