chapter 14, Design Examples

Excerpt

Several different design problems and approaches are presented in this chapter to bring together the concepts that have been presented thus far. The problems are necessarily brief and incomplete, but they show the approach and some of the tradeoffs.

14.1 The Design Process

Infrared system design is not a process of synthesis. Rather it is a process of invention, analysis, and iteration. The process I have found to be most productive is to analyze first the geometry—calculate the field of regard, the field of view, and the pixel size. Next calculate the time line—the frame time, dwell time, and the bandwidth. Then calculate the sensitivity on the idealized basis described above. The last step is the optical scheming. If all of this goes well (and it probably won't the first time) then repeat the sensitivity calculation with more accuracy and realism and do some real optical design.

We will apply these principles to the problems delineated below.

14.2 A Night-Driving Imager

The problem, stated simply, is to design an infrared device for driving a car in the dark and during times of poor visibility. This is a real problem. At least one company has hinted that it has such a project in place, and similar devices have been used by the Portland, Oregon, police.

What should the field of regard be? One answer is the full angular extent of the windshield as seen by the driver. This is approximately 90 degrees horizontally and 30 degrees vertically. Another answer, as we shall see, is that the field should be as much as you can get, perhaps with slewing. The field of view should be about 30 degrees, because that is about what we can get with a reasonable optical system. What should the angular resolution, the angular pixel size be? It should be the angle a man subtends at a reasonable stopping distance. We assume (and there are arguments based on the Johnson criteria for this) that we should get five pixels across a man at a distance of 100 m. If the man is about average in size, each pixel will have to be about 5 cm. This requires a pixel subtense of 0.5 mrad. Then, since 30 degrees is approximately 0.5 radians, there are 1000×1000 pixels in the field of view. We assume that the system is a real-time imager, so that the frame time is 1/30 of a second. If a single detector were to be used, the dwell time would be 0.03 μs. We know that this is too short and will give a bandwidth in the many megahertz region. So we immediately plan for an array of some sort. The input SNR in the 8- to 12-μm region is about 10⁷ (see Chap. 6). When this is divided by the square root of the bandwidth, we have idealized NETD values in the MWIR and LWIR regions, respectively, of 3.5 and 0.006 K for a staring system with a bandwidth of 15 Hz. One reason the MWIR looks so bad is that I assumed a quantum efficiency of 0.01 as in PtSi.