|
Substantial activity over the past few years has been devoted to improving the performance of uncooled IR detectors. These efforts have been rewarded by impressive progress in the reduction of pixel size, with sustained further improvements in NETD. Recent bolometer arrays are only a factor of two or three away from the temperature-fluctuation noise limit for current values of thermal isolation, and optimism remains high for further advances. However, essentially all bolometer-based thermal imaging sensors are currently limited by spatial noise, not by temporal noise. Whenever temporal noise is visible on the display of an uncooled IR imager, it is a good bet that spatial noise will be visible as well. Although fixed-pattern noise is an important part of spatial noise, it is not the only part. Spatially coherent temporal noise, 1/f noise, and drift all contribute to degrading sensor performance, and, like fixed-pattern noise, the degradation exceeds the expectation set by simple numerical comparison. These effects degrade imager performance more significantly than an equal variation from truly random temporal noise, because the eye integrates both temporally and spatially to suppress the latter. By including in our analysis a simple model of the eye-brain response we can determine the specifications necessary to suppress these effects so that true background-limited performance can be realized. This paper examines the sources of spatial noise, analyzes the effect on performance, and attempts to set expectations for performance limits. To limit the scope, the analysis is restricted to pulse-biased, current-mode bolometers.
|
