My youngest boy, who is crazy about dinosaurs, once asked me if all dinosaurs perished right after the great meteor impact. “No,” I said, “some survived long after.”
In 1993, Eric Fossum delivered the provocative but seminal presentation “Active Pixel Sensors: Are CCDs Dinosaurs?” It was controversial at the time, but nearly twenty years later, CMOS imagers now outsell CCDs.
There is no doubt that once the highest volume imager application – mobile phone cameras switched to CMOS, an enormous amount of investment was made to develop, fine tune, and perfect CMOS imagers and the fabrication processes that manufacture them. As a result of this investment, we witnessed great improvements in image quality, even as pixel sizes shrank.
In machine vision, area and line scan imagers rode on the coat tails of this mobile phone imager investment to displace CCDs. For most area and line scan imagers, CCDs are indeed history.
There is, however another important type of imager, the time delay and integration (TDI) imager. TDIs, which are commonly used in machine vision and remote sensing, operate much like a line scan imager except that a TDI has many, often hundreds, of lines. As the image of the object moves past each line, each line captures a snapshot of the object. TDIs are most useful when signals are very weak, since the multiple snapshots of the object are added together to create a stronger signal.
Currently, CCD and CMOS TDIs sum the multiple snapshots differently. CCDs combine signal charges, while CMOS TDIs combine voltage signals. The summing operation is noiseless in CCDs, but not in CMOS. When a CMOS TDI has more than a certain number of rows, the noise from the summing operation adds up to point that it becomes impossible for even the most advanced CMOS TDI to have less noise than a modern CCD TDI.
One path forward for CMOS TDIs is to emulate CCD TDIs by having CCD-like pixels that can then sum charges. We’ll call this the charge domain CMOS TDI. Charge domain CMOS TDIs are technically feasible, but will require significant investment to develop, fine tune, and perfect. Unlike CMOS area and line scan imagers, the economics do not favour charge domain CMOS TDIs. Mobile phones neither need TDIs nor charge summing. Hence, there is no coat tail for CMOS TDIs to ride on.
Whether or not a technology survives is often determined as much by economics as by technical considerations. The economics do not favour CMOS TDIs, so like the dinosaurs that survived the great meteor impact (which many now call “birds”), I am betting that CCD TDIs will stick around for quite a while. Who knows—they may even develop a few new tricks.