Manufacturers continue to increase frame rates while improving sensitivity. A scientific camera does much more than simply capture a picture of an object under study. If an acquired image is to reliably reflect measurements of chemical concentration, cell growth, etc., the digital values associated with each pixel must directly relate to the amount of light detected. It is imperative, therefore, that cameras designed for scientific imaging maintain this quantitative relationship between input and output. Although any digital camera that preserves this relationship might be considered scientific, these cameras distinguish themselves based on their ability to do so at very low light levels, over a wide dynamic range and/or at high frame rates. High-performance scientific cameras are engineered to deliver photon-noise-limited imaging down to near-single-photon input levels. Ultimately, the perfect scientific camera would be capable of detecting every incident photon, with precision limited only by photon shot noise across the entire dynamic range of the image. Scientific cameras allow researchers to image individual green fluorescent protein molecules in a polyacrylamide gel. The data is from the lab of W.E. Moerner, department of chemistry, Stanford University. High-performance scientific cameras improve sensitivity by running the CCD slowly to minimize on-chip amplifier noise. Over the past several years, improvements in amplifier design and signal processing have boosted attainable frame rates. State-of-the-art on-chip amplifiers provide camera noise floors of three to six electrons at digitization rates ranging from 1 to 20 MHz. Quantitative performance, meanwhile, is preserved over a full 16 bits of dynamic range with very high linearity. A new approach to reducing camera noise even more is being pioneered by Marconi Applied Technologies in Elmsford, N.Y., and Texas Instruments Inc. in Dallas. These CCDs provide a gain structure prior to the on-chip amplifier. When the amplifier noise is referenced to the light input, it is effectively reduced by the gain factor. An on-chip gain of 100 times reduces the effect of amplifier noise by 100 times, so that a noise floor of 10 electrons would drop to an equivalent 0.1 electrons. However, this strategy introduces excess noise during the charge-multiplication process. Although the on-chip gain should allow single-photon events to be detected (providing incredible sensitivity), the excess noise means that the number of photons measured is more uncertain. Fortunately, the gain mechanism can be turned off when additional precision is required. Improved cameras based on the next generation of this technology are slated for release in early 2002. New detection technologies are also being used to overcome the traditional compromise between keeping sensor noise very low and running a CCD output amplifier as fast as possible to obtain high pixel rates. Reading the CCD through multiple output ports is one solution to this performance trade-off. The parallelism increases frame rates, but the implementation of associated off-chip electronics becomes increasingly less efficient as the number of outputs grows beyond even a few. CMOS sensors excel in this area. Amplifiers and analog-to-digital converters can be placed on every column, increasing the parallelism during readout by 1000 times or more for a large sensor. For low-light applications, however, CMOS does not yet have the low-noise amplifier performance and high quantum efficiency required for low-light scientific imaging. System integration is also driving innovation in scientific imaging. Many design trends are found in the consumer camera market. For instance, numerous companies are adopting FireWire to simplify the computer/camera interface. Also, 12- to 16-bit analog front ends are facilitating the placement of much of the difficult low-level signal processing into a single integrated circuit, allowing smaller, less expensive packages. Some popular CCDs now come in hermetically sealed packages with integrated thermoelectric coolers. As manufacturers continue to push the limits of performance, they must be careful not to sacrifice the quantitative advantage that underpins the cameras they seek to perfect. Meet the author Robert LaBelle is head of life science business at Roper Scientific Inc. in Tucson, Ariz.