Ground-based astronomy remains at the forefront of our quest to understand more about the cosmos. While many celestial bodies in the night sky can be seen at visible wavelengths, the signal often becomes very faint by the time it reaches us here on Earth. Optical telescopes must therefore be equipped with light detectors that combine high sensitivity – converting as many incoming photons as possible into electrical measurements – with minimal noise.

For decades the detector-of-choice has been the charge-coupled device (CCD), which combines a quantum efficiency of more than 95% in the visible with low-noise operation. An important advantage of CCDs is that they are back-illuminated, which means that the photodiodes used to convert the incoming photons into photoelectrons are positioned in front of the electrical circuitry. This maximizes the area available for photon capture, and also allows efficient cooling from the back of the sensor to reduce the build up of thermal noise during long exposures – which can last for several minutes or even longer.

However, CCDs have significant limitations at faster time scales, which are increasingly needed to observe dynamic processes or to enable quick analysis for applications such as adaptive optics. “A CCD usually only has a single read-out node,” explains Jason McClure, chief technology officer of Teledyne Princeton Instruments. “At the end of every exposure the photoelectrons generated in each pixel must be shifted across the sensor to reach the read-out node, and this slows down the read-out speed.”

Faster read-out times can be achieved by measuring the photoelectrons collected together at the end of each exposure with high-speed analogue-to-digital convertors (ADCs), but these can introduce high levels of read noise. Even at high ADC rates, long read-out times are a particular problem for large-area detectors with lots of pixels (4k x 4k or greater), ranging from seconds to dozens of seconds.

Another issue for the full-frame CCDs typically used in astronomy are the mechanical shutters that are used to block out incoming light during readout. These mechanical shutters have finite lifetimes and need frequent replacement when the camera is in heavy use – which can be problematic for remote observatories where maintenance can be challenging. Additionally, opening and closing a mechanical shutter is relatively slow, reducing the frame rate and generating quantitative errors for shorter exposure times.

Some of these issues have been addressed through the introduction of electron-multiplying CCDs (EMCCDs), which use on-chip amplification to boost the signal relative to the read noise. EMCCDs are able to detect much weaker signals than traditional CCDs, and can also operate at the higher frame rates needed to capture the evolution of dynamic events.

What’s more, many EMCCDs are equipped with electronic shutters that are faster and more precise than mechanical versions. “Electronic shutters also reduce the dead time during which the camera is unable to detect any incoming light,” adds McClure. “EMCCDs shift the detected photoelectrons into a frame storage area at the end of each exposure, which allows the next exposure to start while the signal is being read out from the storage area.”

Despite the advantages of EMCCDs, the random nature of the on-chip amplification process generates excess noise that limits the overall sensitivity when imaging brighter objects, typically when the signal is larger than a few photons per pixel. Their complex gain response can also make it more difficult to obtain reliable quantitative measurements, which has limited their use for some astronomical applications – as discussed in more detail in the article “Types of Camera Sensor“.

Alternative visions

While CCDs remain the most popular technology for ground-based astronomy, image sensors based on traditional CMOS technology now offer a viable alternative. “CMOS image sensors have not generally been suitable for astronomy because they are front illuminated, and are both less sensitive and more noisy than CCDs,” says McClure. “But more advanced CMOS imaging chips are now emerging that rival the performance of CCD and EMCCDs, while also overcoming the common limitations of these technologies.”

One of the big advances has been the introduction of back-illuminated CMOS sensors, which has increased the quantum efficiency to more than 90% in the visible range. Such CMOS designs are also more sensitive than a typical CCD at ultraviolet wavelengths, which makes them ideal for observations at multiple wavelengths.