08 Jun 2007
Charge-coupled devices have been developing at a rapid pace over the past few decades. Antoinette O'Grady gives an overview of what's on offer today and provides useful guidelines on choosing the right detector format and the right architecture for your application.
Charge-coupled device (CCD) cameras have been around since the 1960s and have been constantly improving in terms of performance, design and ease of use. Today, they are the detectors of choice for a wide range of scientific applications (see table). Scientific CCDs can be divided into three main formats – full-frame, frame-transfer and interline devices.
A full-frame CCD (figure 1) is basically a 2D array of photo sensors, which comes in a silicon-based semiconductor-integrated circuit package. Each individual photo-sensor is called a pixel. Photons falling on the CCD's pixels can either be reflected, pass straight through the sensitive silicon layer, or be absorbed in the sensitive silicon layer, and then converted into electrons.
The potential for a photon of a given wavelength to be converted to an electron is known as the quantum efficiency (QE). The electrons can be stored in each pixel. In order for the corresponding charge to be read out, it is shifted vertically from pixel to pixel sequentially into a read-out register. The electrons in each pixel in the read-out register are then shifted horizontally into an output charge amplifier.
One of the fundamental characteristics of CCDs is that they have low read-out noise. It is the read-out noise that determines the detection limit or smallest detectable signal, as a signal has to be larger than the noise floor in order to be seen. As the bandwidth of the charge amplifier increases, so does the noise level of the device. However, keeping the bandwidth small to minimize the noise level of the device will limit the read-out speed of the CCD and result in a slow frame rate.
The 2D array is usually in a square or rectangular format and pixel sizes generally range from 6.8 to 26 µm. Typically, imaging applications use square arrays with smaller pixels, as resolution is important. A mechanical shutter is usually used with full-frame devices for imaging applications in order to prevent light falling on the pixels while they are being read out, which would result in a smearing of the image, unless the exposure time is significantly longer than the read-out time. However, mechanical shutters can have lifetime issues and can be relatively slow.
Rectangular arrays with larger pixels are typically used for spectroscopy applications as here the dynamic range (the ability to see small signals adjacent to large signals) is more important than the resolution. The bigger the pixel, the larger the well depth and the better the dynamic range. Mechanical shutters are often not required for these applications as the sensor is usually read out in a series of columns, rather than individual pixels. In this case, it does not matter that light is still falling on the detector as it is being read out.
Frame-transfer devices (figure 1) negate the need for a mechanical shutter by having a sensor that is divided into two parts – an image area and a storage area. The image and storage areas are usually identical in size, and the storage area is covered with an opaque mask, usually made of aluminium. The image can be transferred very quickly into the storage area. However, to achieve continuous frame rates, the minimum exposure time has to be equal to the read-out time of the storage area. It is possible to reduce the minimum exposure time to a lot less than the read-out time of the storage area if a continuous frame rate is not required – for example, if only two extremely fast successive frames were required. Typically, frame-transfer devices can provide frame rates of up to 3–4 frames/s. Full-frame and frame-transfer CCDs are ideal for static low-light applications with medium to long exposures and for moderate time resolution on a millisecond timescale.
Interline sensors also consist of photosensitive and masked storage areas. Although in this case, the photosensitive and masked areas extend along the CCD's vertical axis so that each masked area is adjacent to a photosensitive area. The charge collected in the sensitive region can be shifted extremely quickly into the storage area and is subsequently read out. Interline devices provide much faster frame rates but have much lower sensitivity as they only have approximately 25% fill factor, which means that only 25% of the image area is sensitive to incoming photons. In order to try to increase the fill factor, microlenses are often used to focus the light onto the sensitive regions. Interline sensors have limited dynamic range and are typically 12 bit as opposed to the 16 bit performance of full-frame and frame-transfer CCDs. This limits their use for the more demanding scientific applications, but they are very popular for applications such as microscopy that do not have very low light levels.
While the format and the type of CCD selected is dependant on the area, resolution, dynamic range and frame rate/spectral rate required, it is the QE or the probability of detecting a photon that will determine the ultimate sensitivity of the camera. The QE depends on the structure of the CCD itself with the most common structures being front illuminated (FI), back illuminated (BI), open electrode (OE), indium tin oxide (ITO) and deep depletion.
• FI structures give a moderate QE (about 45%) over the visible region. Phosphors can be coated on these devices to extend their sensitivity into the ultraviolet (UV) region.
• BI CCDs give the ultimate in QE (95%) and can be coated with an antireflection coating to further improve their sensitivity in specific wavelength regions. These devices can also be coated with phosphors to extend their sensitivity into the UV but suffer from fringing effects in the near-infrared (NIR) above 750 nm.
• OE structures provide a good QE response (55–60%) over a broad wavelength range. They are not as expensive as BI CCDs and do not suffer from fringing effects in the NIR.
• Deep-depletion CCDs with fringe suppression offer the best QE in the NIR region, but have approximately ten times higher dark current then normal BI CCDs. They have to be operated at much cooler temperatures to get the benefit of their sensitivity.
• ITO devices are not very common and are only currently available in imaging sensors with small pixels. These devices have a better QE compared with OE devices, but tend not to perform as well as detectors with BI structures.
Intensified CCDs (ICCDs) consist of an intensifier that is usually fibre-optically coupled to a FI CCD (figure 2). The intensifier consists of a photocathode that converts the incoming light into electrons. The high electron gain of a multichannel plate amplifies these electrons (gains of 10,000 are easily achieved) and the resulting signal is converted back into photons by a phosphor on the exit window of the intensifier. The photons then travel through the fibre optic and are detected by the CCD in the usual way.
The high gain of the multichannel plate enables it to amplify the signal far above the noise floor, essentially bestowing these devices with single-photon sensitivity and making them highly sensitive detectors for low-light applications. However, the key benefit of ICCDs for scientific applications is their gating capability. It allows the intensifier to act as a very fast shutter and provide time resolution on a sub-nanosecond timescale. This is useful in applications that feature a short time event of interest within a broadband continuum with a longer lifetime. These events are usually repetitive, and the intensifier can capture them and accumulate their signal while ignoring the broadband continuum.
It is the intensifier rather than the CCD that determines the sensitivity of an ICCD system and there are many different types of intensifiers available depending on the time resolution, wavelength range and sensitivity required. However, the QE of intensifier tubes is significantly less than the QE of conventional CCDs. ICCDs also suffer from the disadvantages of additional noise due to the amplification process (known as the noise factor), crosstalk, "chicken wire" and halo effects.
Electron-multiplying CCDs (EMCCDs) are manufactured using standard fabrication techniques. They differ from conventional CCDs in that they have an additional, or gain, register inserted between the end of the usual shift register and the amplifier. Higher-voltage amplitudes than normal are used in the gain register to generate electrons (the gain) by impact ionization. Because this is done before the amplifier, a signal that is lower than the read-out noise can be amplified by up to 1000 times and can be easily identified provided that it is greater than the electron-multiplying CCD's generated noise factor. The key benefit of EMCCDs is that they overcome the limitations of conventional CCDs by providing excellent detection limits with fast frame rates. They are ideal for extremely low-light-imaging applications that require high frame rates. Their key limitation is that their detection limit is set by their noise factor, which is a noise source due to the amplification process. For applications where the light levels are above the CCD read-out noise, EMCCDs offer no benefit and a conventional CCD is often a better choice in this instance.
Best of both worlds
Hybrid-sensor devices utilizing CCD and CMOS devices represent the next generation of detectors. They combine the sensitivity benefits of scientific CCDs with the read-out capabilities of CMOS devices without having the limitation of increased read-out noise or noise factors. Therefore, they will be able to offer high dynamic range, high sensitivity and extreme detection limits independent of read-out speed.
• This article originally appeared in the May 2007 issue of Optics & Laser Europe Magazine.Optics & Laser Europe magazine – subscribe here