Special Report: Optical Patterns

Incoherent /coherent multiplexing with a SELDA An incoherent /coherent multiplexing approach to reduce the noise due to multiplexing by use of an array of microlasers has been developed by Jenkins and Tanguay [78]. The concept of the method can be understood as follows: When multiple gratings are recorded simultaneously by a coherent light source, all the cross terms are added coherently. In other words, the amplitudes of all the gratings are summed first and squared later. As a result, the number of noise gratings (or the number of cross-talk terms) becomes N2 , where N is the number of gratings. On the other hand, as in this proposed method, if each grating is recorded sequentially one by one, the cross-terms are incoherently added (each grating amplitude is squared first and summed over all the gratings later), and thus the number of noise gratings becomes N; therefore cross-talk can be greatly reduced. Such an incoherent recording can be achieved simultaneously with an array of independent (not phase locked with each other) coherent light sources, such as an SELDA. Jenkins and Tanguay applied this noise reduction method to photonic interconnection and holographic optical elements for 2-D wavelength-division multiplexing applications.

This technology of using microlasers for holographic storage will be extremely useful when suitable holographic recording materials become available in the future. Recent developments in biological holographic recording materials, such as bacterial rhodopsin, show a great new promise for future computers [79].

12.5 Microlasers for holographic associative memory

12.5.1 Introduction Recent advances in neural networks opened many new possibilities for optical information processing for broad application areas [80-82]. Optics, especially coherent optics, has found an excellent match in implementing the neural networks that require parallel and analog computing. Fig. 12.14 illustrates an associative memory for word-break recognition to generate a readable text from a continuous string of words. In the system, an input word stream is presented at the input plane of the system. Autocorrelation peaks that appear at places where there is a match between the input and the memory words pre-recorded in a hologram are detected. All the spurious sidelobes are removed by a threshold operation in the correlation plane. Next, the separation between the peaks is magnified along the word direction. This stretched correlation output is reflected back to illuminate the hologram and reconstruct the whole memory at the output plane. The output through a window that is situated at the origin of the output plane is the desired readable text, with all the errors corrected and spaces inserted between words.

Fig. 12.14. Holographic associative memory for word-break recognition.

12.5.2 Holographic neurons Such a neural network system normally consists of three parts: (1) a recognition part to compare an input with all the memories, (2) a reconstruction part to retrieve the corresponding memory, and (3) nonlinear thresholding elements to make decisions. Recognition (pattern recognition in subsection 12.3.2) and reconstruction (holographic memory in 12.3.1.) have already been described. Below, we focus on the third part, nonlinear thresholding elements, or so called optical neurons. There has been much research done on optical neurons. Recently an integrated vertical-cavity surface emitting microlaser array with HPT's to yield optically controlled lasers has been demonstrated by Chan et al.[30]. The structure of the device is shown in Fig. 12.5, as explained above. Fig. 12.5(b) shows the output vesus input optical power relationship for various bias voltages. As shown in the figure, the device displays a suitable nonlinear threshold function as an optical neuron with the threshold input optical power of 0.1 mW (at the wavelength of 855 nm). It also provides an optical gain of approximately 5, making it an excellent candidate for an optical neuron array. Besides its nonlinear threshold function, the output light from the optical neuron is coherent. It is a laser; therefore, it can reconstruct any type of hologram. In Fig. 12.15(a), the neuron with an input power above threshold is fired to reconstruct the corresponding memory. Fig. 12.15(b) shows an experimental result. If the output light signal is below threshold, nothing appears in the output plane (left). On the other hand, as soon as the input light is increased above the threshold level, the neuron is fired and the corresponding holographic memory is retrieved (right).

Previous    Next    Table of Content for report: Optical Patterns    Home

Optical Patterns

 

 

 

Photuris.com - Optical Data Networking