Performance of image sensors for security cameras must be geared to low-light sensitivity, high signal to noise ratio, and high temperature performance. Security image sensors are typically mission-critical 24/7, often standing outside under sun, snow, wind, and other harsh environments. It's especially important that they can perform well at nighttime under dark conditions.
Although both CCD and CMOS sensors can be used for security and surveillance applications, CMOS is becoming the technology of choice. In the past CCD sensors have had superior performance in some areas, for example, good performance in low-light conditions, wide dynamic range, and low noise. However, in the last few years the performance of CMOS sensors has significantly improved to the point where they are matching or besting that of CCDs. A major factor in the rapid improvement in CMOS performance has been the research and development conducted by mass-market manufacturers. Their R&D has been motivated by the inherent benefits of CMOS, such as low power requirements, small size, simplified circuitry, and suitability for use in mobile telephone cameras. In addition, CMOS sensor chips are based on standard CMOS process technology for digital ICs, with some adaptation for imaging (e.g. pinned photodiodes) and are manufactured at foundries with the standard process technology typically used for all sorts of digital ICs. A CMOS imager converts charge to voltage at the pixel level, with voltage amplification integrated into the chip.
CMOS image sensors have historically been assembled with two different architectures: Frontside Illumination (FSI) — the older, and Backside Illumination (BSI), the more recent.
A frontside illuminated sensor is constructed with a lens at the front and a matrix of photodetectors at the back. This arrangement is relatively simple to manufacture; however, it places the matrix and its wiring in a position to reflect back some of the light. This reflection reduces the available information in the incoming image signal.
In a BSI sensor, the wiring is placed behind the photocathode layer, which is reversed during manufacturing. The reverse side is then thinned so that light can strike the photocathode layer without passing through the wiring layer. Because of this architecture, the chance of an input photon being captured and converted to an electron is increased from about 60% to over 90% — a significant advantage in low light security applications. The drawbacks to BSI have been problems such as cross-talk, which causes noise, dark current, and color mixing between adjacent pixels. Thinning also makes the silicon wafer more fragile.
A New Approach
While BSI outperforms FSI overall, it's more complex, and therefore more expensive, to manufacture. You build the photodiode first and then you turn the wafer around, grind it, and place the metal on top of it. In addition to the increased number of steps, there are more mask layers — more stacks — than FSI, therefore it's costlier to manufacture. In addition, the time-to-market is usually 1.5 – 2× longer than FSI.
SmartSens Technology (Shanghai, China) has developed an advanced CMOS vision sensor designed specifically for the security/surveillance market. It benefits from the ease of manufacture of FSI, but has performance that exceeds what would have resulted from either FSI or BSI type alone, said Chris Yiu, Chief Marketing Officer. To distinguish this new technology from the other sensors on the market, they have called theirs SmartSens DSI Technology. According to Yiu, its improved performance is the result of both pixel process and circuit design.
“Our DSI sensor uses FSI process technology, meaning that the photodiodes are laid underneath the metal layers in the process,” said Yiu. However, SmartSens DSI technology provides improvement in sensor performance based on several factors, she said. “We have improved the circuit design to reduce the noise generated by the system. We also use a pixel design architecture from the BSI world so that we can have the benefits of both technologies. The results exceed what we would have if we only used either type alone,” she added.
SNR1 was proposed by SONY in 2016 for the security market as an index to quantitatively evaluate picture quality at low illumination. Sensitivity to light is obviously important but it is not the only factor affecting performance. Noise has a major impact on the actual usable signal. Noise can be generated from background, which is a characteristic of CMOS, and also from pixel blooming or pattern noise generated by the system's circuitry, both analog and digital. So, there's a lot of random noise that can be generated. Therefore, in low light, where the light and signal levels stay low, the ratio between the actual meaningful signal and the random or pattern noise is critical.
SONYs SNR1 measure is an indicator of low light performance, but it is different from signal to noise ratio (SNR). It is a measure of the lowest illuminance level in lux, for which the signal is equal to the noise. Therefore, with SNR1, a lower number signifies better performance. The SNR1 index, however, is limited to CMOS sensors and security applications.
Another important measure of security sensor performance is dark current, which is the current generated by the sensor when there is no light impinging on it. This is an unavoidable characteristic of CMOS, and since it represents a signal error, it is a goal of CMOS sensor design to minimize dark current.
Global vs Rolling Shutter
One of the important choices in selecting a surveillance camera is whether to use a global or rolling shutter. A rolling shutter scans line by line, whereas a global shutter outputs all of the image pixels at the same time.
This is a difficult choice, as both technologies have their pros and cons. For traditional surveillance applications, real-time video clips are generally captured, displayed, and then stored in the system archive. These systems are usually implemented with rolling shutter sensors, which are widely available in commercial markets in different resolutions and different price points. While there are issues with rolling shutters when objects are moving quickly, this is generally not a concern. Everyday surveillance systems are mostly cost sensitive so the higher price tags that come with the global shutter sensors are an impediment. However, for security cameras used with AI or traffic control, moving objects are the main targets. In those cases, the motion artifacts caused by the slower moving rolling shutter are not desirable. These systems are mostly owned by authorized entities and inputting the data in the quickest and most accurate way will be more important than minimizing cost.
Selecting the Right Camera Sensor is Key to Optimizing Security Systems
The important thing is to evaluate the requirements for your particular application and then to understand the variables: for example, should you look to FSI, BSI, or SmartSens's new DSI architecture? To evaluate your needs, be careful to understand the different ratings, such as dark current and SNR1, as distinct from SNR. Reliability is a key concern, as is the ability to function over the anticipated temperature extremes. Also, decide whether you can minimize cost by using a rolling shutter or you need to optimize performance with a global shutter.
This article was written by Ed Brown, Associate Editor of Photonics & Imaging Technology.