Optical Sources, Detectors, and Systems

8.3 Detector Arrays 185

el. Allowing for atmospheric attenuation, non-normal solar incidence,

and a reasonable range of reflectances in an image reduces the average

expected flux to approximately 10^ photons/pixel for a "sunny" day

image. Light from a full moon, from the example in section 1.4, would

result in an average of about

10^

photons/pixel,

below the noise level of

the vidicon, and only a factor of two higher than the image orthicon

discussed in the previous section.

The sensitivity of the vidicon may be improved by using a return

beam read-out similar to that of the image orthicon. An even more sen-

sitive mode of operation for both the orthicon and the vidicon is the iso-

con return beam technique. During the initial time when electrons are

being replaced on the target surface by the electron beam, a fraction are

reemitted as secondaries and

scattered

over a reasonably large angle.

Once the correct negative charge is reestablished, the remaining beam

electrons are

reflected

in a narrow beam angle. By collecting and multi-

plying only the scattered electrons, the excess main beam shot noise is

therefore eliminated. We have not treated the important topics of retina

uniformity, beam focus and deflection techniques, and vidicon dynamic

range. These can be found in

(Csarba,

1985).

8.3 Detector Arrays

Discrete detectors may be used to form an image by scanning of a

single element or a linear array as shown in Figure

8.5.(a)

and (b). The

detectors themselves are not scanned, rather, the image field is scanned

over the element or array using a rotating cylindrical mirror assembly or

assemblies in the optical path. The techniques are discussed in detail in

(Accetta and

Shumaker,

1993).

A full two-dimensional array may be used

n • nnnnnnn

D ^ DDDDDDD

• • nnnnnnn

D-

D •

DDnnnDL

D ^ nannnnn

n ^ nnnnnnn

(a)

(b) (0

Figure 8.5 (a) Single element, (b) linear array scan and

(c)

area array.

186 Systems II.: Imaging

as shown in Figure 8.5(c). The single detector systenn uses standard

techniques for read-out, with the amplifier bandwidth determined by the

scan rate. Linear and area arrays may also be read out in parallel with

individual amplifiers and data recording for each pixel. Alternatively,

the parallel outputs from the amplifiers can be integrated, stored, and se-

quentially sampled to produce a serial data stream. In either case, de-

tector uniformity may seriously limit image quality and data reliability.

This problem can be overcome by calibration and the resultant stored

gain and bias adjustments for individual elements.

For large arrays, the complexity of individual detector amplifiers

may be overcome by the use of capacitive storage and sequential readout

of the detectors. The sampling shown schematically in Figure 8.6 may

be accomplished by a shift register-driven set of FET switches, but the

added capacitance of the output line increases the value of NEE. For

background-limited operation at long wavelengths, this is not a serious

problem; however, at near-infrared and visible wavelengths, the ampli-

fier noise becomes limiting. From Section

4.5,

the NEE is given by

8m

(8.7)

which, for a read-out line capacitance of 10 p/, a ^^ of 5 mS, y=l, and

a bandwidth of I MHz, becomes 250, if we use correlated double

sampling, that is, measure the voltage immediately before and after each

individual switch is closed, as discussed in Section

4.5.

Otherwise, kTC

noise results in a noise equivalent electron count given by

To integrating amplifier

)))))}

Figure 8.6 Schematic representation of array sequential read-out