Low-light scientific cameras are behind a number of ground-breaking discoveries, ranging from quantum imaging to astronomy. For more than five decades, CCD cameras and their variants —intensified CCD (ICCD) cameras and electron-multiplying CCD (EMCCD) — have provided the single-photon sensitivity and moderate frame rates needed for spectroscopy and scientific imaging applications.
More recently, scientific CMOS (sCMOS) cameras capable of achieving low read noise and higher frame rates have turned out to be an alternative to CCD cameras in a number of applications. However, the first generations of these sCMOS devices seem to fall short on sensitivity because of their front-illuminated architecture, which imposes a basic limit on their quantum efficiency (that is, the fraction of incident photons detected in each pixel).
Aided by the latest CMOS fabrication technology, it is now possible to develop sCMOS devices with a back-illuminated sensor architecture. This results in sCMOS sensors now being capable of CCD like quantum efficiency (>95%) and dynamic range without compromising the high frame rates and low read noise for which they are known. The new generation of sCMOS cameras, such as the KURO™ from Princeton Instruments, take complete advantage of this back-illuminated sensor technology in order to provide a major improvement over previous-generation, front-illuminated sCMOS cameras.
Back-illuminated sCMOS camera technology is considered to be a serious contender as an optical detector for myriad applications, including astronomy, cold-atom imaging, quantum imaging, fluorescence spectroscopy, hyperspectral imaging and high-speed spectroscopy. This article presents the salient features and performance characteristics of the new technology.
For several years, back-illuminated technology has been available for scientific CCD detectors. Back-illuminated detectors are chosen over front-illuminated detectors for ultra-low-light applications ranging from astronomy to Raman spectroscopy to biological imaging because of their higher sensitivity over a broader spectral region (deep-UV to near-IR).
The difference between back-illuminated and front-illuminated sensors is apparent at a glance. Front-illuminated sensors are reflective in appearance since most of the incident light is reflected back to the viewer, whereas back-illuminated sensors are visibly darker because of their absorption of most of the incident light (see Figure 1).
Figure 1. Typical front-illuminated CCD / sCMOS sensors (left) are reflective in appearance whereas back-illuminated sensors (right) appear dark.
The KURO is available with a back-illuminated sensor architecture similar to that of the most sensitive CCD detectors available. Back-illuminated technology allows this new sCMOS camera system to deliver >95% quantum efficiency and 100% fill factor (see Figure 2).
Figure 2. Back-illuminated sCMOS technology provides higher quantum efficiency than front-illuminated sCMOS sensors across a broad spectral range, including the UV.
Since front-illuminated sCMOS sensors have readout/conversion circuitry within each pixel, just a portion of each pixel is sensitive to light. This portion is known as the “fill factor” of the pixel. Most front-illuminated sCMOS sensors contain microlenses on top of each pixel in order to refocus the incoming light into the photosensitive part of the pixel and increase the effective fill factor (see Figure 3).
Even though microlenses help enhance the light-collection efficiency of a front-illuminated sensor, they also carry specific drawbacks that restrict the performance of most of these sCMOS cameras. It should be noted that unlike front-illuminated sCMOS sensors, which claim ~80% peak quantum efficiency, back-illuminated sCMOS sensors do not use any microlenses.
Figure 3. The typical front-illuminated sCMOS sensor architecture (left) relies on the use of microlenses. Back-illuminated sCMOS sensors (right) do not utilize microlenses.
Microlenses unfortunately are most efficient only when the incident angle of light is normal to the sensor surface (see Figure 4). If light enters the sensor at any other angle, like in the case for most scientific imaging and spectroscopy applications, the efficiency of microlenses degrades significantly — particularly at wider entrance angles (“high NA” in microscopy parlance).
While this quantum efficiency vs. incident angle relationship is not extensively published in camera or sensor manufacturers’ literature, the degradation is a real cause for concern when ultra-low-light performance is needed. Even though the layout of microlenses has been enhanced by different CMOS manufacturers, the angular dependency of photo-response leads to non-uniformity, particularly at the edges of the sensor.
Figure 4. Front-illuminated sCMOS sensors often rely on microlenses, which significantly degrade quantum efficiency when light is incident at any angle other than normal to the sensor surface. New back-illuminated sCMOS sensors do not exhibit this performance limitation.
Additionally, these microlenses are usually made of plastic-like materials capable of transmitting very poorly, or not at all, in the UV range (below 400 nm). The lack of microlenses in a back-illuminated sCMOS sensor’s architecture translates to outstanding response in the UV range (see Figures 2 and 4).
High Frame Rates and Low Read Noise
Back-illuminated sCMOS cameras, such as the KURO, offer extremely high frame rates, up to 82 fps at full 1200 x 1200 resolution, with an remarkably low 1.3 e- rms (median) read noise. The KURO camera can deliver hundreds of frames per second with decreased resolution (see Table 1). sCMOS sensors typically do not support on-chip binning and yet they do allow “off-chip” software binning after frame acquisition.
Table 1. Low read noise and high frame rates make the new back-illuminated sCMOS camera ideal for high-speed spectroscopy applications.
||Frame rate: fps (12 bit / 16 bit)
|1200 x 1200
||82 / 41
|1200 x 512
||192 / 96
|1200 x 256
||384 / 192
|1200 x 128
||768 / 384
|1200 x 64
||1536 / 768
|1200 x 32
||3072 / 1536
It should be noted that the 11 µm2 pixel pitch of the new back-illuminated sCMOS sensor captures 2.8x more photons than other sCMOS sensors. Each pixel is capable of handling a large full well of 80,000 electrons, allowing exceptional dynamic range (61,500:1 or 95 dB).
Reduced Fixed-Pattern Noise
The KURO back-illuminated camera uses the latest sCMOS fabrication technology along with optimized electronics. As a result, it has a considerably better noise profile than any front illuminated sCMOS camera (see Figure 5).
Figure 5. Fixed-pattern noise: front-illuminated sCMOS sensor (left) vs. backilluminated sCMOS sensor (right).
Table 2 provides a convenient summary of a number of key specifications and performance capabilities linked with front-illuminated sCMOS and recently launched back-illuminated sCMOS sensors.
Table 2. Comparison of front-illuminated sCMOS and back-illuminated sCMOS sensors.
||~65% – 80% (at normal incidence)
||>95% (at all incident angles)
|Pixel fill factor
||60% – 70% typical
||400 – 1000 nm
||<200 – 1100 nm
|UV (<400 nm) sensitivity
||Yes (up to 75% QE in 200 – 400 nm region)
Readout Modes: Rolling and Global Shutter
One of the key features of CMOS sensors refers to the availability of rolling electronic shutter mode. This is distinct from global shutter or “snapshot” mode, which exposes all pixels simultaneously. Global shutter is selected when the object needs to be “frozen” in time; however, this mode usually causes the read noise to increase by as much as 1.5 to 2x compared to rolling shutter mode while decreasing the frame rate by 2x.
In rolling shutter mode, the first row of the sensor is exposed first, followed by the second row being exposed after one line read time, and so forth. In other words, the last row of the sensor is exposed at “(N-1) x line time” after the first row.
This may cause image artifacts when looking at high-speed events that happen significantly faster than the frame rate, but most scientific applications can work with rolling shutter mode as long as the frame rate is adequately high. Global shutter mode, on the other hand, is chosen for industrial imaging applications in which the objects under inspection are moving at a rapid pace.
Improved camera designs, such as that of the KURO, offer a way to activate external light sources/shutters so as to develop a “pseudo” global shutter mode. To do so, the camera just outputs a TTL signal that goes high when “all” the pixels are exposing. As a result, this will cause the pixels to take a snapshot of the event when it is illuminated. Figure 6 provides timing diagrams for the KURO camera’s rolling shutter and “pseudo” global shutter modes.
||Exposure Start time
||Exposure End time
||TO+EXP TIME (user entered value)
||TO+(1xLINE TIME)+EXP TIME
||TO+(N-1 * LINE TIME)
||TO+(N-1xLINE TIME)+EXP TIME
Figure 6. Timing diagrams for rolling shutter and “pseudo” global shutter modes.
Which Sensor Technology: CCD, EMCCD, ICCD or Back-Illuminated sCMOS?
Engineers and Scientists should carefully consider which sensor technology is well-matched to their application. In general, for imaging or spectroscopy applications that need extended integration times (seconds to hours), EMCCD or CCD cameras are still preferred. This is also true for spectroscopy applications that require on-chip binning. In the meantime, for time-resolved applications that need ultrafast gating, intensified cameras (emICCD or ICCD) are the best choice.
Back-illuminated sCMOS cameras provide the frame rates and sensitivity needed for all other applications with comparatively short integration times (less than 10 seconds). Table 3 summarizes several vital features of these sensor technologies and offers some general recommendations for a wide range of applications.
Table 3. Comparison of sensor features with application recommendations.
||~50% (photocathode QE)
||~2 – 4 e- rms
||<1 e- rms (with EM gain)
||<1 e- rms (with intensifier gain)
||<2 e- rms (<1.5 e- rms median)
|Frame rate at full resolution
||~30 fps – 60 fps
||~10 fps – 30 fps
||~40 fps – 80 fps
|Frame rate at reduced resolution
||>5,000 fps (in kinetics mode)
||>10,000 fps (in kinetics/crop modes)
||Yes (<500 psec)
||Astronomy, Raman spectroscopy with requirement for long integration times (msec to hours)
||High-framerate, single-photon-sensitive applications with relatively short integration times (µsec to msec)
||Gated (psec to µsec exposures), time-resolved imaging and spectroscopy
||High frame rates with moderate integration times (<10 sec) in lowlight applications such as adaptive optics, ion imaging, hyperspectral imaging, etc.
The advantages of back-illuminated sCMOS technology are important and so is the potential to fully leverage the inherent benefits of this new sensor type. Designed for operation within the Princeton Instruments LightField® software ecosystem, the KURO is easy to control and can be combined quickly in spectroscopy and myriad imaging experiments. Camera integration for use with MathWorks’ MATLAB® and National Instruments’ LabVIEW® is also simple and fast. A full suite of input-output TTL signals is also provided, making it easy to synchronize camera operation with light sources or external events.
This information has been sourced, reviewed and adapted from materials provided by Teledyne Princeton Instruments.
For more information on this source, please visit Teledyne Princeton Instruments.