OPTICAL PROCESSORS

We all understand that sooner or later the silicon technology used today to create processors will reach its limit. It's like with oil – sooner or later it will run out – that's why they are already starting to develop alternative energy technologies! Exactly the same situation has developed in the world of information technology – they are starting to look for a replacement for silicon technology. Electrons running along conductors like tanks are quite impractical – at the very least, a significant part of their energy is lost, released as heat and electromagnetic radiation, although this is not the only drawback. Moreover, completely different options for replacement are proposed for consideration – from a biocomputer to optical processors. Stop! What is this? It is unlikely that each of us will have a computer made of bacteria on our desk, say, in ten years, but the fact that an optical processor will be installed in the computer of the future is quite real. Today we will talk about this miracle of human hands.

Let's consider the advantages of optical technology:

· Two-dimensional arrays can be transmitted in parallel with one light pulse.

· Possibility of using completely different environments for transmission, storage and processing of information.

· Information processing is possible during its transmission through the optical system, which is the computing environment. Imagine you send a picture for processing – it will be processed almost instantly because it is processed as it passes through the optical system.

· Information that is encoded by an optical beam can be transmitted at the speed of light without the release of a large amount of energy in the logic elements! This is really good – after all, the less energy expenditure, the better.

· The optical system does not allow interception of information, since it does not emit anything into the environment.

All these advantages are achieved due to the fact that photons, not electrons, are used as information carriers.

The basic elements of optical image transfer processors have long been known. These are a lens, a mirror, an optical transparency (a transparency is a transparent plate on which an image is applied in some way, representing the spatial distribution of the absorption coefficient, the refractive index (or thickness), or both at the same time) and a space layer. At present, wave elements have been added to them, as well as lasers, semiconductor multi-element photodetectors, nonlinear optical media, various types of deflectors and light valve devices.

The basic logical function, with which you can build any digital computer, no matter how complex, has many optical implementations. Figure 1 shows a simple example of constructing a multi-input OR-NOT/AND-NOT function. using the L lens and the threshold device-inverter N.

Figure 1 – Construction of the OR-NOT/AND-NOT function

Here, either an optical light valve device (a switching bistable medium) or a simple photoelectronic receiver with a nonlinear transfer characteristic (i.e. a nonlinear dependence of the intensity of the output light flux on the input) can be used as a threshold element.

Figure 2 shows an optical processor that implements the transformation of an input row vector into an output column vector.

Figure 2 – Transforming a row vector into a column vector.

Here LED is a line of light-emitting diodes. They are located on the focal line of the cylindrical lens L1. T is an optical transparency with the transmission matrix T(i,j) recorded on it. The matrix rows are parallel to the generatrix of the first lens. L2 – cylindrical lens, the generatrix of which is parallel to the columns of the transparency matrix. It collects the rays that have passed through the elements of one row on one pixel of the multi-element photodetector D. It is easy to see that the input X and the day off vectors are related by a linear transformation

U = TX.

The optical system can also process two-dimensional structures. Figure 3 shows a diagram of an optical processor that implements the operation of convolution of two images, which underlies the operation of many associative memory and pattern recognition devices.

Figure 3 – Image convolution.

Here S is a flat homogeneous light source, L1 and L2 – spherical lenses, D – matrix photodetector, T1 and T2 – transparencies whose transmission corresponds to two processed images.

The distribution of radiation intensity on the matrix photodetector is proportional to the integral:

Transformation integral.

In the previous examples, light played the same role as electrons in the conductors of conventional microcircuits. In this case, geometric rays acted as “wires”. It is clear that with the same success, light can be driven into a waveguide and a computing environment can be organized according to principles close to the ideology of electronic semiconductor microcircuitry. This is what integrated and wave optics are concerned with.

Fundamentally new possibilities are provided by the use of the properties of spatial coherence of radiation. Thus, in coherent optics the following mathematical operations on complex functions of two variables are easily implemented: multiplication and division, addition and subtraction, integration and differentiation, calculation of convolution and correlation, Fourier transform, Hilbert transform, Fresnel transform and a number of others, it can be shown that even with the help of only two basic operations of multiplication and Fourier transform it is possible to perform a whole series of others (addition and subtraction, differentiation, integration with weight, convolution, change of scale of the function argument, restoration of the function from its spectral density, etc.).

The structure of the coherent optical processor, the so-called 4F-scheme, is shown in Figure 4. Here LS – a laser illumination system that produces a wide beam of coherent radiation. T1 And T2 – amplitude-phase transparencies modulating the phase and amplitude of the passing light wave. L1 and L2 – spherical lenses with focal length F. The resulting signal is read by a matrix photodetector D.

Figure 4 – Coherent optical processor.

The distribution of the amplitude of the light field in the plane of the photodetector is proportional to the convolution of the amplitude transmission of the first transparency with the Fourier image of the amplitude transmission of the second transparency. Processors of this type are used as complex spatial filters in the image quality improvement system, as well as in pattern recognition systems.

The Fourier spectrum of a two-dimensional signal is calculated using the lens L and a layer of space of length F as shown in Figure 5. The remaining elements are intended for data input/output and system lighting.

Figure 5 – Calculation of the Fourier spectrum of a two-dimensional signal.

For a typical computer using the fast Cooley-Tukey algorithm, the length of the Fourier transform increases with increasing sampling points n proportional to: n log(n). In an optical computer, this procedure, even in the two-dimensional case, is performed in just one machine cycle, which makes the optical computer indispensable for solving problems that require rapid situational assessment and real-time control.

BIBLIOGRAPHY

1. http://bsfp.media-security.ru/science/index.htm

Irkutsk Branch of the Institute of Laser Physics SB RAS. Malov S. N. The role of diffusion illumination in subtraction of images of phase objects.

2. New materials of optical informatics, photonic crystals, optical memory: 2002 No. 7. Saint Petersburg State University. Zadvorkin A. V. Optical computers, information processing method.

3. webmaster@media-security.ru. Problems and tasks of optical information processing. Gurevich S. B., Sokolov V. K.