Micro-machined thermal sensors are revolutionizing IR imaging with uncooled thermal sensors. This is achieved by the realization of 2D arrays of low-cost micro-machined sensor pixels with very small thermal conductance as well as thermal capacitance that yield high sensitivity and response fast enough for imaging. A comprehensive comparison between the different thermal sensors, i.e. bolometers, pyroelectric and thermoelectric sensors, is beyond the scope of this summary.
However, thermocouples have some inherent advantages, including low power consumption, no 1/f noise and less stringent requirements of controlling the chip operating temperature. The thermoelectric sensors are constructed of one or more thermocouples that respond with spontaneous voltage to temperature differences induced by absorbed IR radiation. In order to allow temperature differences to develop, the "hot" contact of the thermocouples has to be thermally isolated from the "cold" contact heat sink. Such thermal isolation can be obtained in thin-film thermocouples fabricated on silicon substrates by micro machining of the substrate below the "hot" contacts. If CMOS circuits are also realized on the same substrate, whole micro-systems can be fabricated monolithically.
Since thermoelectric sensors can be realized using standard CMOS layers, they have the best compatibility with CMOS technology and therefore have the potential of achieving low-cost uncooled thermal imagers. Traditional designs of CMOS compatible thermoelectric sensors suffered from two main shortcomings. One disadvantage is the relatively small signal. In order to increase the signal several thermocouples are connected in series to form a thermopile. This, however, reduces the achievable thermal resistance and thus performance.
The second disadvantage is the problematic realization of 2D arrays of sensors. As opposed to traditional micro-machining methods that used wet anisotropic etching processes for silicon bulk micro-machining, the silicon bulk is used as a sacrificial layer and structures made of CMOS process thin films constitute the sensors. This technique allows better yield and better thermal isolation in small pixels. Uncooled IR sensors have been realized integrated in standard CMOS chips using this technique.
Conclusions
The building block for the motion detector were developed and manufactured. The lab-scale post-processing provides a yield of 95%. Nevertheless the current design is very sensitive to manufacturing yield. For that reason an off-the shelf pyro-sensor was demonstrated.
Due to the fact that the design specifications were higher than the actual application needs and that the size of the unitary elements was decided based on the snapshot camera, many technical improvements can be done. That is, the yield can be improved by making bigger elements and by moving from a serial to parallel-serial electronics architecture. A balance between the size of the unitary element, bandwidth and responsivity can be achieved.
The front-end electronics was designed, manufactured and tested, the single element TE sensor was successfully tested and expected performance was validated. The vacuum package was designed and validated with a reference sensor. The TE sensor array was designed, manufactured and tested. However a problem related to the VLSI design was found, which prevented the reading-out the sensors signal. The problem was fully analysed and corrective actions were determined.