Jonathan Borremans (program manager at imec) and Piet De Moor (senior business development manager imagers at imec) present a new approach to charge-coupled-device (CCD)-based time-delay-and-integration (TDI) imaging. The new imager implements the CCD TDI pixels together with advanced CMOS drivers and readout in one single chip. This CCD-in-CMOS technology is compatible with backside illumination for enhanced sensitivity, and multispectral filters can be added for increased imaging performance. With this highly sensitive and high-speed imaging solution, high-end applications such as remote sensing, life sciences and machine vision are targeted.
The principle of time-delayed imaging
Time-delay-and-integration (TDI) imaging is a preferred line-scanning technique used in applications such as industrial inspection and earth observation. In general, this imaging technique can be used when an object or scene is moving linearly over an imager (or vice versa), and when the content of the scene is not changing within a short period of time. Conceptually, a 2D image can be captured by using one linear row of photosensitivity pixels. The image is then taken line by line, delayed in time – the delay being synchronized with the linear scene movement. The resolution is determined by the number of pixels in the row. A smarter way of capturing a 2D image is not to use a single row of photosensitive pixels, but a number of rows. Each row detects the same information as the previous rows, but now delayed in time. By adding the correct scene data of all rows, an increased signal and signal-to-noise can be obtained.
Schematic representation of the time-delay-and-integration principle, using 8 rows of photosensitive pixels
CCD – a ‘noiseless’ way of collecting and transferring the signals
The technology of choice for collecting and transferring the analog TDI pixel data is charge-coupled-device (CCD) technology. CCD uses pixels that can detect light by collecting photon induced charges under a CCD gate – i.e., in a potential well. Interestingly, the collected charge packages can be transferred easily from one gate to an adjacent one, without adding noise. In a TDI implementation, this allows to move a charge packet from one pixel to another (i.e. in a column, along the track) in perfect synchronization with the scene movement. An analogy is often made with the term ‘bucket brigade’, used for a line of people passing buckets of water. As a result, the signal from the different rows (also called stages) at different times are finally collected into the same charge packet. This way, the signal is added in hardware. Importantly, as the charge collection and transfer is noise-free, the only added noise is the one at the final readout stage, when charge is converted into voltage.
A comparison can be made with CMOS imager technology – a technology that is preferred over CCD for consumer imager applications due to its high degree of electronic circuit integration. In a CMOS imager pixel, the light induced signal is converted into the voltage domain, and this step induces a noise increase in each pixel. Contrary to CCDs, when combined with TDI, the signal transfer and the signal storage cannot be done in the charge domain. Rows (or stages) representing the same scene but taken at a different time, are now to be added digitally – in software. This ‘digital TDI using CMOS’ solution not only comes with a noise penalty. It also requires a large memory and fast data processing, as 2D frames are to be read-out at line rate and added in the digital domain. For a large number of TDI rows, this process becomes very complex and power hungry. An advantage of this technology is the high degree of integration, as fast readout electronics can be integrated on the imager chip itself.
TDI CCD-in-CMOS – combining the best of both worlds
Traditionally, a CCD system requires a board with multiple chips, including the CCD, drivers and readout chip. Imec has been developing a new solution that combines CCD pixels and CMOS readout in one technology, enabling a single imager chip with both CCD and CMOS functionality.