QUANTUM EFFICIENCY ADVANCES SENSORS FOR NIGHT VISION
By Frederic DeVriere, E2V Technologies

In recent years, low-noise CMOS imaging sensors have triggered many new opportunities and applications for the night-vision market. Although low noise is key, it is not the only parameter to consider when dealing with low-light vision. Sensitivity also is essential for the overall performance of the sensor.

Recent developments in CMOS image sensor (CIS) technology have garnered significant attention from designers of low-light devices. Although not yet able to directly compete with traditional night-vision devices (such as image intensifiers), CMOS image sensors exhibit several interesting features that make them the best solution for some parts of the market.

The technology is developing quickly and promises exciting performance in the coming years. Key to night-vision applications is noise level, which has been widely advertised as the major performance indicator – but it is not the only factor to take into consideration. Recent developments combine high-sensitivity pixel architecture with a lower-noise approach and extended wavelength response.

For many years, image intensifiers have dominated the night-vision market. They offer good performance, thanks to high amplification gain and low power consumption, but are purely analog. Digital versions involve coupling a CCD or CIS to the image intensifier.

Electron-bombarded CMOS (EB-CMOS) is now becoming a serious competitor to the image intensifier for portable digital low-light-level solutions.

Electron-multiplying CCD (EMCCD) is more focused on non-portable applications because of its need for cooling; low-light CMOS is a new solution for an application that intrinsically does not have a multiplication mechanism. This represents a drawback for photon-starved environments because it needs very low noise to be able to discriminate a relevant signal, but it makes it the perfect device for dual day and night use.

Opposite the image intensifier or EB-CMOS – which cannot operate without an amplification gain – CMOS image sensors can operate under bright illumination conditions, and features such as high-dynamic-range pixels make them even more suitable for these conditions. They also have unique advantages such as low cost, a much longer life span than other solutions, no drift of the key parameters with time, reduced pixel size (leading to lower cost integration) and high resolution.

MTF and SNR

Two parameters are driving the performances of low-light sensors: modulation transfer function (MTF) and signal-to-noise ratio (SNR).

MTF is a common way to characterize the ability of a sensor or system to provide good contrast of a scene. It is this parameter that makes the low-light CIS perform better than an image intensifier for night levels (down to night level 3), because it provides more contrasted images.

The advantage of a CIS over an EB-CMOS sensor on this parameter is also proved, but between two CMOS image sensors, the difference is not generally large.

SNR is the other key factor in the ability of a sensor to deliver a useful image. In this figure, the eye can discriminate contrast differences down to 1/64, but noise significantly affects the perceived image. At first glance, the sensors produce acceptable quality images with SNR = 10, and excellent images are produced with SNR = 40. Above that level, the improvement is no longer perceptible to the human eye.

For low-light environments, there are two obvious ways to increase SNR:
Decrease noise, which is the approach of all low-noise sensors, and increase signal, optimizing the response of the sensor in terms of quantum efficiency, or QE – the ratio of conversion between incoming photons and measured electrons – and spectral response. Considering the noise factor for a CMOS image sensor along with the intrinsic response of the pixel leads to the concept of noise-equivalent illumina