CCD Sensor Architectures
Common CCD Architectures
Four types currently make up almost all CCD area sensors sold – full frame (FF), frame transfer (FT), interline transfer (IT) and frame interline transfer (FIT). Almost all of these are frontside illuminated, that is, built so that the light enters on the same surface of the device that holds the circuitry, and they are almost all single-tap, having only one output point. There are some exceptions and a few additional variations on the structures that will be discussed in detail below. In all of the figures in this section, the light gray areas are sensitive to light and the dark gray areas are covered with aluminum. The arrows show the direction of charge motion when the electrodes are operating.
Full Frame Architecture
The Full Frame CCD image sensor (Figure 1) is the simplest type. It consists of a set of photosensitive registers arranged next to each other in columns with a transport register at the bottom configured so that each well can receive charge from a different column. In a typical operating cycle, light is prevented from reaching the sensor by a shutter and then the charge is cleared from the photosensitive registers while the image sensor is in the dark. The shutter is opened for the desired exposure interval and then closed. In the dark, the charge in all of the columns is shifted down by one row so that the last row moves into the horizontal transport register. A faster clock then moves the charge packets in the horizontal transport register to the sampling node to generate the output voltage. When the row is complete, the next row is moved down for readout. This process is repeated until all rows have been read. The sensor is the ready for another exposure.
Figure 1 - Full-frame CCD architecture
Frame Transfer Architecture
The Frame Transfer (FT) architecture (Figure 2) was developed as the first CCD type suitable for continuous (video) imaging. In an FT sensor, the exposure and readout functions of the sensor are physically separated with the exposure section built essentially identical the an FF sensor and a storage section almost a copy of the exposure section but covered with an opaque layer (usually aluminum) as shown in Figure 5. In an FT sensor, the exposure and readout can occur almost simultaneously because while the exposure section collects light for a new image, the previous image, held in the storage section, is shifted out.
Figure 2 - Frame transfer CCD architecture
The FT sensor is slightly more complicated to operate that an FF sensor because the exposure and storage section need separate shift drivers. In a typical cycle for a simple 30 frame-per-second progressive scan video camera, the exposure section is collecting light and the charge is not moving while the previous information in the storage section is shifting down to the readout section line by line. After the shifting is complete, the two sections are connected to the same clocks and the entire image is quickly moved down from the exposure section to the storage section. The line shift rate during the transfer is typically several hundred times faster than the readout line shifting rate. Since the charge pattern moves down very rapidly, the vertical striping is reduced relative to the FF striping. After the shift, the scanning is disconnected from the exposure section and the readout resumes at a slower shift rate.
Generally, the CCD cells in the storage section are smaller than the pixels in the exposure section because there is no need to efficiently collect light, no need to provide a square matrix and no need to provide anti-blooming protection. A few extra rows are usually added at the top of the storage section and covered with the opaque layer to provide a buffer between the edge of the light-collecting pixels and the storage cells. This prevents stray light from getting into the first few rows of the storage area that contain image data where it can generate spurious charge that can produce white background patches in the image.
Interline Transfer Architecture
In Interline Transfer sensors, the exposure and storage sections are alternated column by column in the same area of the silicon as shown in Figure 3. Each column of photosensitive elements has next to it a vertical transfer register covered with aluminum. The group of transfer registers makes up the storage area. Light is collected in the photoelements, which may be CCD wells or photodiodes, and then the accumulated charge is moved from all photoelements simultaneously into the neighboring transfer registers. After this move, the charge is shifted vertically, one line at a time, into the output register for readout. For faster operation, IT devices can incorporate multiple taps and can support split vertical scanning.
Figure 3 - Interline transfer CCD architecture
Many variations on this basic arrangement exist, each with its own characteristics. Although the photoelements appear to be in columns as in FT sensors, the photoelements in IT sensors are not connected together vertically because there is no need to shift vertically in the photosensitive areas. Fundamentally, the photoelement can be either a CCD well or a photodiode. Typically, the CCD well has lower noise because it does not have to be reset but the photodiode has higher quantum efficiency because it does not require phase lines on its input surface. The large variety of configurations for photoelements now available has blurred the lines significantly so when the details of operation are important, thorough investigation of the actual configuration in candidate sensors may be required.
Transferring the charge from the elements to the transport registers may occur at a specified time or continuously until the time to scan has arrived. There may be some advantage to continuous transfer because the possibility of leaving some charge behind is reduced and because continuous transfer may allow the use of pinned photodiodes, which can have lower noise because they do not need to be reset.
IT sensors also offer the possibility of electronic exposure control. If the photoelements are equipped so that all of them in the sensor can be simultaneously reset upon application of an external signal, then this signal can be used to define the beginning of an exposure. Since the end of the exposure is defined by end of the time that the charge is transferred to the transport register, the duration of the exposure is set by the difference between these two times. If the scanning is periodic, then the end of the transfer time will be periodic and fixed at a particular point in the exposure-readout cycle. This means that although the duration of the exposure can be varied, the position of the exposure in the cycle cannot. As a result, if the exposure must be synchronized with an external event, it is necessary to time the entire cycle to this event by operating in an external trigger mode in which entire images can be produced on demand.
Because the photoelements are all isolated vertically from one another, the possibility of blooming along the photoelement columns is essentially eliminated. Antiblooming drains can be added between each column of photoelements and the transfer register for the column to the left as part of the isolation structure required between them. When the photoelements are isolated on all sides like this, blooming can be kept under very tight control.
Frame Interline Transfer Architecture
The Frame Interline Transfer (FIT) Architecture (Figure 4) was developed to provide both very low vertical striping and electronic exposure control. Striping in FT sensors, which have no exposure control, can be reduced only so far as the image can be moved rapidly from the imaging section to the storage section while IT sensors, which do have exposure control, have more striping as the exposure time is reduced. The FIT sensor is a combination of IT and FT in which the exposure section of an FT sensor is replaced by an IT sensor array. After the exposure, the charge can be shifted to the vertical transfer registers under the aluminum shields as in an IT device, but is then immediately shifted at high speed into an FT storage array below. As a result, the time available for accumulation of unwanted charge from light leaking under the shields is reduced to typically less than 1 millisecond with any exposure setting. The striping is thus reduced by both the effect of the shields and the rapid transfer out of the exposure section.
Figure 4 - Frame interline transfer CCD architecture