ROBUST DESIGN FOR SOFTWARE DEVELOPMENT - Software Design for Six-Sigma: A Roadmap for Excellence - page 469

Transfer Function

FR

2 1

DP

FIGURE 18.3 Robustness optimization definition.

illumination required by the camera. This value represents the lowest illumination
that the camera can operate under with acceptable image quality, as defined by this
method.
The light sensitivity of a camera is affected by many design parameters and their
variation (noise). These include the aperture, the quality of the lens, the size and
quality of the sensor, the gain, the exposure time, and image processing. When using
several criteria, it is difficult to compensate for the quality of the camera with gain
and image processing. For example, increasing the gain level may provide better
luminance, but it also may increase the noise in the image. Illumination measures
[in lux (lx)] the visible light falling per unit area in a given position. It is important
to note that illumination concerns the spectral sensitivity of the human eye, so that
electromagnetic energy in the infrared and ultraviolet ranges contributes nothing to
illumination. Illumination also can be measured in foot-candles (fc). 2
Consider two settings or means of the minimum illumination parameter
(DP)—setting 1 (DP * ) and setting 2 (DP ** )—having the same variance and prob-
ability density function (statistical distribution) as depicted in Figure 18.3. Consider,
also, the given curve of a hypothetical transfer function relating illumination to image
quality, an FR, 3 which in this case is a nonlinear function in the DP. It is obvious
that setting 1 produces less variation in the FR than setting 2 by capitalizing on
nonlinearity. 4 This also implies a lower information content and, thus, a lower degree

2 1lx

10.76 fc. Note that 1 lx can be interpreted as 1 “meter-candle.”
3 A mathematical form of the design mapping. See Chapter 13.
4 In addition to nonlinearity, leveraging the interactions between the noise factors and the design parameters
is another popular empirical parameter design approach.

=