Technical Abstract: NIR and visible light reflection from food products has long been the basis of scientific research for the detection of defects and contaminants as well as food quality attributes. Most of the research in this area reports derived calibration equations that indicate the potential of using the technology for non-destructive detection. However, rarely is the research taken to the point of actual implementation of a real-time sorting device. Commercially available dual bandwidth NIR/vis sorting devices use onboard computers to implement decision algorithms that can be trained to perform required sorting tasks. As such, they tend to be complicated and expensive. Here we describe methods for building sorting devices for specific applications that require no computer or sophisticated electronics for the application of a decision algorithm. As an example of the application of the principles involved in the design and construction of simple and economical real-time sorting devices we report here the development of a prototype instrument for the separation of in-shell nuts from kernels in a pistachio processing stream. The principles used could be used to develop sorting devices for many other food processing streams as well.