Proceedings Article | 5 June 2020
KEYWORDS: Staring arrays, Quantum well infrared photodetectors, Sensors, Nonuniformity corrections, Infrared detectors, Manufacturing, Mid-IR, Imaging arrays, Imaging systems, Infrared sensors
The III-V material system has proven to be a solid base for building infrared detector focal plane arrays (FPAs), enabling well-established designs such as bulk InSb-photodiodes, GaAs-based quantum well infrared photodetectors (QWIP) and GaSb-based type-II superlattices (T2SL). The remarkable advancement in the sensitivity and stability of such FPAs seen in the past decades calls for revision of the merits and the acceptance criteria used for the performance evaluation of infrared imaging arrays. Specifically, the early skepticism toward QWIP has largely been focused on their low quantum efficiency, and consequently low specific detectivity of the FPA pixels, as compared to the counterparts with bulk design. It was later demonstrated that unrivalled spatial uniformity of QWIP arrays (typically with a residual non-uniformity <0.05 %) not only compensated for this drawback, but rendered them superior in terms of the sensitivity, with typical noise equivalent temperature difference (NETD) <30 mK (when staring at 30 °C source through F/2.24 aperture, FPA on 15 μm pitch operated at 70 K and 60 Hz frame rate, peak absorption at 8.6 μm). This introduced QWIPs to the demanding applications, such as tactical or gas sensing, in spite of the design’s inherently low detectivity. In this paper, we propose the tightened criteria for marking the defective pixels and demonstrate how IRnova’s III-V based long-wave QWIP and mid-wave T2SL FPAs with 15 μm pitch live up to them. High spatial uniformity is reflected in symmetric and almost tail-free NETD histograms for both detector types. Only 0.13 % of the pixels exhibit NETD above ×1.5 of the median value for QWIP and 0.14 % - for T2SL detectors. At the same time, there are virtually no pixels falling outside less stringent limit of ×3 the median value. For QWIPs, this observation is consistent with flat noise spectra, which are completely free from the low-frequency “1/f”-component, as measured on the test cells. Finally, our attesting procedure allows tracing even solitary pixels exhibiting deviating types of noisy behavior, including random telegraph signal (RTS)-like, whenever they are present in the array. Low number of RTS-like pixels is a key advantage of III-V based FPAs that enables them to be used for high-operating-temperature (HOT) applications. Another characteristic feature of III-V based detectors is the outstanding stability of their performance. For IRnova’s mid-wave T2SL FPA on 30 μm pitch, using a 3-year-old non-uniformity correction map even without refreshing the offset is penalized by a mere 100 % increase in the residual fixed pattern noise relative to its temporal NETD. In the paper, we complement this with a study on the stability of the non-uniformity correction of IRnova’s QWIP and T2SL detectors on 15 μm pitch.
