Read laser_photonics_scmos_nov09.pdf text version


Scientific CMOS image sensors

Scientific CMOS image sensors


Scientific CMOS (or sCMOS for short), a new technology for image sensors, was presented at the >Laser. World of Photonics< trade fair in Munich recently. As is the norm when any new technology is launched, sCMOS was pitched as being far superior to anything that has gone before ­ but is this claim justified? And, what is the background to the joint development of this technology by three companies that are otherwise competitors?



he partners behind this development have all been in the business of making cameras for scientific applications for many years. A mere glance at the requirements for these camera systems is sufficient to show that the mass produced digital image sensors available on the consumer market simply don't fit the bill. The following example illustrates this point perfectly: high resolution is a standard requirement in many camera applications, so too in snapshot cameras and camera phones ­ today, resolutions in the range of 5 to 12 megapixels are considered par for the course. Pixel sizes of

just 1 to 2 µm2 have become the norm in order to maximize the number of image sensors that can be manufactured for a wafer surface, thus also minimizing the price. However, these tiny pixels have relatively small fill capacities for charge carriers. These charge carriers represent the amount of light at each pixel, and so the full-well capacity and readout noise determine the image sensor's dynamic range, that is, its ability to capture a wide brightness range in an image. As a result, camera phones have a dynamic range of just 6 bits, or 64 distinct levels of brightness. This may be adequate for color pictures, but for highend measurement applications it is vastly inadequate.

The requirements for measurement technology

Several examples can be found for the dramatic differences in the specified requirements between the different applications While high quality image sensors are common in high quality, digital single lens reflex cameras, these sensors are not available on the open market. This means that the only way for manufacturers of cameras for scientific applications to get their hands on significantly improved image sensors or new technologies in this area is to invest in development themselves. As such groundbreaking development is both expensive and risky, the three companies Fairchild Imaging, Andor



5 | 2009

Scientific CMOS image sensors


Technology and PCO joined forces in 2008 to invest cooperatively in the development and drive it forward.

Bigger, faster and more sensitive

Below is a wish list of the key improvements that are currently required in cameras used for scientific applications and in other fields: resolution greater than 1 megapixel frame rate faster than 10 frames/s (fps) high sensitivity dynamic range >12 bit (1:4096). One or more of these performance features may be important for any one particular application. Higher resolution, for example, allows samples to be investigated under the microscope with less magnification and using a large field of view, thereby reducing processing times and increasing throughput. In addition, recent years have seen an increase in microscopy-based research into dynamic processes, for example, the behavior of membrane calcium channels over time or the synchronization of frame sequences with the heartbeats of animals under study. As a rule, a highly sensitive image sensor is an absolute must in spectro-

scopic and physical applications, either because the measurement signal is very weak in the first place or because the costs of appropriate light sources increase almost exponentially with their power. In the pharmaceutical industry, for example, measurement volumes are getting smaller and smaller in order to improve throughput. This, in turn, reduces the number of light emitting (bio-)marker molecules and thus also the quantity of light that can be measured. Sensitive, low noise image sensors are therefore in high demand in this area also. Another aspect to consider is the requirement for imaging solutions with a high dynamic range in some applications. While logarithmic image sensors are available, these are often too imprecise for the measurement of luminance and so image sensors with a high dynamic range are far preferable.

When it comes to microscopic applications, optical dimensions limit the size of the image sensors that can be used. The optimal pixel size therefore has a pixel pitch in the region of 6 to 8 µm in order to limit the cost of the optical system and the required semiconductor surface.

So, what's new?

If we take each of the items on our wish list for improvement and compare these with what's currently on offer in cameras for scientific applications, it becomes apparent that cameras already exist that partially satisfy these requirements. For example, some CCD image sensors offer a resolution in excess of 4 megapixels (2048 x 2048 pixels and higher) with acceptable noise values of 9 electrons. However, even with a fast input signal (at 40 MHz), these are unable to achieve a frame rate of more than 15 fps, and then only at elevated noise values (14 electrons) and with correspondingly lower dynamic range (12 to 14 bit). CMOS image sensors for high speed imaging have also been on the market for a number of years and can achieve frame rates of 500 fps at full resolution. Unfortunately, these only offer a resolution V


PCO AG 93309 Kelheim, Germany Tel. +49 (0)9441 2005-0 Fax +49 (0)9441 2005-20 Vision 2009: Booth 6.B12

Performance comparisons

1 Performance comparison of different CMOS/CCD image sensors at different lighting levels (@ 20°C)

5 | 2009




Scientific CMOS image sensors

V of 1280 x 1024 pixels, 10-bit dynamic range, readout noise of 18 electrons and a quantum efficiency of less than 50 percent. emCCD image sensors with on-chip amplification were developed to detect the lowest levels of light. These also offer a respectable quantum efficiency of around 60 percent, and can even achieve frame rates of 12 to 30 fps. However, amplification, while reducing readout noise, comes at the expense of dynamic range. In addition, higher resolution will not be available in the foreseeable future for these expensive sensors. And that's not to mention the general long term stability issue that has been detected in these emCCD image sensors, which results in a continuously changing gain curve. In light of these previous developments, it could be argued, with some

justification, that there is nothing really new about sCMOS. However, while none of

2 Initial prototype of an image sensor in sCMOS technology

should be the component of choice. In between these extremes, both technologies find use in a wide range of other applications, where the choice is determined by the specific conditions that apply in each case (Figure 1).

the individual performance features are new in and of themselves, the restrictions outlined above show how, until now, the optimization of one parameter was always at the expense of another. These restrictions are typified by the issue of which sensor type to choose for any one camera application, as the answer in the past has always been something along the lines of: CMOS image sensors are best suited to high speed imaging with high frame rates due to their intensive use of parallel signal processing. If, on the other hand, high image quality and a low level of readout noise for low light signal values are a priority, a CCD image sensor

Scientific CMOS ­ sCMOS

What's truly innovative about sCMOS, however, is the way it ticks all of the boxes on the wish list of improvements, meaning that the time has perhaps come to rethink the answer to the question above. The image sensor prototype in the new sCMOS technology shown in Figure 2 has the following performance features: 2560 x 2160 pixel resolution quantum efficiency of 60 or 90 percent (front- or back-illuminated) readout noise < 2 electrons at a frame rate of 30 fps, or < 3 electrons at a maximum frame rate of 100 fps dynamic range of 1:16 000 (14 bit) at 30 fps pixel size of 6.5 x 6.5 µm2 Thanks to an innovative pixel structure and the merging of signal paths, all of these features are now combined in a single image sensor. The technology can, for example, simultaneously process a pixel signal with varying gain settings and convert it into a numerical value using 11bit analog/digital converters. The design avoids unnecessarily complex A/D converter structures and exploits the high dynamic range of the image sensor to make two data streams available. These can be merged into a single, extremely precise 16-bit data value or, if the objective is to reduce data volume, can be used for either high sensitivity or a high signal level. High resolution and a moderate chip size (21.8 mm diagonal) combine to facilitate a range of microscopic applications with a low magnification factor and large field of view. The low readout noise at frame rates of 30 to 100 fps also breaks new ground. Figure 3 shows an initial comparison of the imaging capabilities. The image on the left was captured with a cooled CCD camera designed for scientific applications. The camera has a 2048 x 2048 pixel image sensor (pixel size 7.4 x 7.4 µm2), operated at a clock speed of 40MHz to achieve a frame rate of 14.7 fps. The cooling temperature was set to +10°C and the image was captured with an exposure time of 1 ms. The image on the right shows the

3 A comparison of the first imaging results: image taken with a cooled CCD camera for scientific applications (left) and image taken with a test prototype camera with an sCMOS image sensor (right)



5 | 2009

Scientific CMOS image sensors


same scene captured with a pre-production sCMOS image sensor in a prototype camera. The sCMOS image sensor has a resolution of 2560 x 2160 pixels (pixel size of 6.5 x 6.5 µm2) and was operated at a clock speed of 150 MHz to achieve a 50 fps frame rate. The sCMOS image sensor was cooled with ambient air and the shot was exposed at 1.3 ms to compensate for the difference between this camera and the CCD in terms of pixel size. The two cameras captured the same scene of a mechanical stopwatch and a test chart in the background from an equal distance, using the same lenses at an f-stop of eight. The objects were placed in a light proof box with a low level of LED illumination and thus a correspondingly low signal level (<200 counts) at the sensor. The poor quality of the picture taken with the CCD camera illustrates an unfortunate feature of CCDs ­ >smear< ­, here observable in spite of the rapid readout. Because of their relative brightness, the reflections of the illuminating LEDs in the glass of the stopwatch generate additional signal during readout, appearing in

the picture as vertical stripes. Scientific CMOS image sensors do not exhibit smear. The example shown here also illustrates the higher dynamic range of the sCMOS image sensor, which is largely due to the lower level of readout noise.


Dr. GERHARD HOLST heads the Scientific & Research Dept. at PCO, one of the world's leading manufacturers of high-end digital camera systems, making use of image amplifiers, CCD and CMOS image sensors. He is also responsible for the scientific and technical information made available by PCO.

Summary: a new all-purpose solution?

It remains to be seen which camera and cameras with sCMOS image sensors will emerge in the future. However, based on the performance data gathered to date, it can be claimed, with justification, that sCMOS technology with its image sensors is already very close to being an all-purpose solution and that it is likely to open up a whole new range of future applications. More information can be found by visiting the sCMOS technology website ( or on the websites of the individual collaboration partners at:

You can find this article online by entering the document number eLP110038

5 | 2009



4 pages

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate


You might also be interested in

Microsoft Word - REO Article - Photonics West Paper 7912-40 Surface Quality
Layout 2 copy