JPH0285831A - Parallel light arithmetic unit - Google Patents

Abstract

PURPOSE:To execute high speed processing by laminating a data liquid crystal panel and a program liquid crystal panel whose optical property is switched by an applied electric signal in the advance direction of a beam light and executing simultaneously the processings of plural pieces of data. CONSTITUTION:From an electronic control part 5a, an electric signal corresponding to program data and input data is outputted, and supplied to each transparent electrode of program liquid crystal panels 17a - 17c and data liquid crystal panels 16a, 16b of a liquid crystal arithmetic part 1a. The optical property of these liquid crystal panels is switched by the inputted electric signal, and as a result, plural light beams from a light source part 2a are allowed to transmit or cut off. Subsequently, the light beam which transmits through is made incident on a photoelectric converting part 3, converted to an electric signal by a CCD sensor array 31, and this electric signal is inputted to the electronic control part 5a and a large capacity memory 8 as a result of operation. In such a way, since the processings of plural pieces of data are executed simultaneously against array data of a large capacity, high speed processing can be executed.

Description

[Detailed Description of the Invention] "Industrial Application Field" This invention relates to a parallel optical arithmetic device that is capable of processing array data such as images in parallel using light for each arithmetic element. It is.

"Conventional technology" Conventionally, in array data processing in electronic computers, array data is... Data is processed in chronological order.

That is, first, one set of data in the array data stored in the mass memory is accessed by the processor and read out via the data bus. After this data is processed within the processor according to the program, it is written to the large-capacity memory again via the data bus. In this way, processing is performed on one set of data, and when this processing is performed on all sets of array data, the array processing is completed. In such a method, the number of times the computer processes the data increases in proportion to the number of array data, and the total processing time increases accordingly. In multidimensional simulation and image processing, CPUs (central processing units), which vary from 1 to 100, are used.
Alternatively, registers are used to perform calculations by repeating the above process as many times as necessary.

"Problem to be Solved by the Invention" By the way, the above-mentioned conventional arithmetic device has the following drawbacks.

(1) The process of importing the data to be processed from the storage device, performing calculations, and writing the calculation results to the storage device is performed in chronological order, so as the number of data in the array increases, the time required for processing increases depending on the number of data. becomes longer in proportion to Therefore, when calculating a large amount of data, the time required is enormous.

(2) As a countermeasure for the above problem (1), semiconductor LSI (
Processing speed has been increased by using multiple stages of processors (large-scale integrated circuits) or multiple stages of registers for arithmetic operations within the processor, but due to implementation issues such as wiring, conventional In electronic circuit technology, the limit is low.

(3) Since the semiconductor LSI processor and storage device are connected by a data bus that performs bidirectional communication, the time utilization rate of the semiconductor LSI processor and storage device is low, resulting in poor processing efficiency.

The present invention was made against this background, and therefore enables high-speed processing and easy-to-configure parallel processing by simultaneously processing multiple pieces of data for a large amount of array data. The purpose of the present invention is to provide an optical processing device.

"Means for Solving the Problems" The present invention provides a parallel optical arithmetic device that processes a plurality of digital data in parallel using optical means. a light source unit that emits a plurality of light beams corresponding to the number of light beams; a liquid crystal operation unit disposed in a traveling path of the plurality of light beams, the liquid crystal material being sealed between a polarizing plate and two glass substrates; In addition, transparent electrodes are formed on the glass substrate corresponding to the passage positions of the plurality of light beams, and have the property of (1) transmitting only a specific polarized component of the incident beam, or (2) polarization of the incident beam. A liquid crystal operation section comprising a data liquid crystal panel and a program liquid crystal panel, whose optical properties such as the property of transmitting the light by rotating the surface thereof, are switched by an electric signal applied to the transparent electrodes, stacked in the direction in which the beam light travels. , detection means for detecting each beam light emitted from the light source section when it passes through the liquid crystal calculation section and is output; The present invention is characterized by comprising an electronic control unit that converts program data specifying an operator to be applied to the input data into an electrical signal and supplies the electrical signal to the transparent electrode of the program liquid crystal panel.

"Operation" The electronic control section outputs electrical signals corresponding to the program data and input data, and these are supplied to the transparent electrodes of the program liquid crystal panel and the data liquid crystal panel of the liquid crystal operation section. The optical properties of these liquid crystal panels are then switched by the electrical signals. As a result, each of the plurality of light beams emitted from the light source section is transmitted or blocked depending on the combination of input data and program data supplied correspondingly.

Then, the detection means detects the presence or absence of the transmitted beam light from the liquid crystal calculation unit, and obtains the calculation result. Here, the transparent electrodes of each liquid crystal panel are provided corresponding to a plurality of beams of light supplied from the light source section, and an independent electrical signal can be supplied to each transparent electrode. Therefore, by supplying electrical signals according to the data data 5 and the program to the transparent electrodes corresponding to each beam light, and determining whether each beam light is transmitted/blocked, multiple calculations can be performed simultaneously in parallel. can.

"Example" Hereinafter, the present invention will be described in detail with reference to the drawings.

Embodiment 1 FIG. 1 is a block diagram of a parallel optical arithmetic device according to a first embodiment of the present invention. In FIG. 1, la is a liquid crystal calculation section, which includes data liquid crystal panels 16a and 16b and program liquid crystal panels 17a to 17c. This liquid crystal calculation section 1a
This will be explained in more detail later. 2a is a light source section, which includes a laser diode 21, a lens 22a, and 2a.
2a, and consists of. Here, the light beam emitted from the laser diode 21 is turned into a parallel beam by the lenses 22a and 22a, -, and is made to enter the liquid crystal calculation section 1a. In addition to the laser diode, a high output light emitting diode may be used as the light source. 3
is a photoelectric conversion section, which is composed of a lens 22b and a C0D (charge coupled device) sensor array 31. 4a is a national polity, on which a light source section 2a, a liquid crystal calculation section la, and a photoelectric conversion section 3 are mounted. 5a is an electronic control section. Further, 8 is a large capacity memory.

Data input/output between the large-capacity memory 8 and the electronic control section 5a is performed via a memory bus 9.

Further, the data signal DI output from the electronic control unit 5a,
D2 and program signals P1 to P3 are sent to data liquid crystal panels 16aS to 16b and program liquid crystal panel 17a via data bus 7a and program bus 6, respectively.
=17c. Further, the output signal from the COD sensor 31 is transmitted to the electronic control unit 5a via the data bus 71.
It's about to be taken in by

FIG. 2 is a detailed diagram of the liquid crystal calculation section 1a. In the figure, lla and Ilb are polarizing plates.

Here, the polarization planes of the polarizing plates 11a and llb are as shown in the figure.
As shown by the arrow, Ila has p-polarized light, and z
In b, the light is S polarized. 12a to +2d are ferroelectric liquid crystal layers, 13 is a guest host liquid crystal layer, 14a to 1
4f is a glass substrate, and 15a-15j are transparent electrodes.

Here, the glass substrates adjacent to the ferroelectric liquid crystal layers 12a to 12d are separated from the edge of the glass substrate from the upper right to the lower left.
Rubbing processing is performed in the direction of 2.5 degrees (two lines in the figure). Further, both of the glass substrates 14c and 14d enclosing the guest-host liquid crystal layer 13 are subjected to a rubbing treatment in the vertical direction.

Among the transparent electrodes, 15a, 15c, 15e, 15g,
151 is formed in a horizontal stripe shape, and 1
5b, I5d, I5r, 15h, and 15j are formed in a vertical stripe shape, and a matrix electrode is constituted by a horizontal electrode and a vertical electrode.

In addition, each liquid crystal panel 16a% t6b, t7a to 17c
, each intersection of a horizontal electrode and a vertical electrode constitutes one pixel. By applying a voltage with a predetermined polarity to the horizontal electrode and the vertical electrode, a signal "l" or "0" is applied to the pixel at the intersection of the picture electrodes.
” is written. Also, among the pixels of each liquid crystal panel, a series of pixels lined up on the travel path of one beam of light incident from the light source section 2a constitute one arithmetic unit.

That is, this liquid crystal calculation section 1a has a number of calculation units corresponding to the number of pixels per liquid crystal panel.

Now, the operating principle of the liquid crystal calculation section 1a will be explained. First, the operating principle of this liquid crystal calculation section 1a is based on a completely new calculation concept obtained by changing the conventional calculation concept, so this will be explained. According to the conventional concept of operation, for example, in the case of logical product operation,
Operators have an operation/sequence relationship on data, such as ``If the logical product operator AND is applied to data A and B, a logical product operation result C is obtained.'' The operation of a logic operation circuit using a conventional electronic circuit is consistent with this operation concept, and each transistor constituting the circuit is switched ON10FF depending on the state of human data A and H, and an output signal corresponding to the input state is generated. is obtained. However, in the above-mentioned logical product operation, it should be possible to treat the data ASB and the operator AND in the same way in that they both influence the operation result C. Therefore, from this standpoint, a new concept of computation is derived in which "the computation result is determined by the combination of data and operators."

The liquid crystal calculation section 1a according to the present invention operates on a calculation principle based on this new calculation concept. The outline of its operating principle is that the optical properties of the pixels of the liquid crystal panel are switched depending on the input data, and the optical properties of the pixels of another liquid crystal panel are switched depending on the type of operator applied, and these pixels are connected to the beam of light. The calculation result is obtained by detecting whether or not the
By doing so, an arithmetic unit that can obtain an arithmetic result corresponding to a combination of data and an operator is realized.

Next, the operation of this parallel optical arithmetic device will be explained.

A plurality of sets of digital data each consisting of 2 bits stored in the large capacity memory 8 are read out by the electronic control unit 5a, and the first pit of each set is used as the data signal DI, and the second bit is used as the data signal D2. As, memory bus 9
and the data liquid crystal panels 16a, 1 via the data bus 7a.
6b. A program that determines an operation for each set of data is transmitted via the program bus 6 as a program signal PI-P3 to the program liquid crystal panels 17a to 17a.
17c.

Each beam of light output from the light source section 2a is sent to the liquid crystal calculation section 1.
At point a, it is transmitted or blocked depending on the combination of data signals D1, p2 and program signals P1 to P3. When the parallel beam of light transmitted through the liquid crystal calculation section 1a and outputted enters the photoelectric conversion section 3, it enters the lens 22.
b, and an image is formed on the COD sensor array 31.

Then, it is converted into an electrical signal by the CCI sensor array 31, and this electrical signal is stored in the large capacity memory 8 via the data bus 7b, electronic control unit 5a, and memory bus 9 as a calculation result.

Next, the operation of the liquid crystal calculation section 1a will be described in further detail using FIG. The parallel beam light from the light source is passed through the polarizing plate 1
Only the p-polarized light component is extracted by 1a. Then, the parallel beam light consisting only of the p-polarized light component is incident on each pixel of the program liquid crystal panel 17a disposed within the liquid crystal calculation section 1. Here, when the parallel beam light incident on the program liquid crystal panel 17a has a signal "1" written in that pixel, its polarization plane is rotated 90 degrees and passes through the program liquid crystal panel 17a, and the signal "0" is transmitted. is written, the light passes through with the plane of polarization maintained.

The data liquid crystal panels 16a, 16b and the program liquid crystal panel 17c have the same structure as the program liquid crystal panel 17a, and operate in the same way.

Next, the operation of the program liquid crystal panel 17b will be explained.

In the guest-host liquid crystal layer 13 in the program liquid crystal panel +7b, a nematic liquid crystal is used as the host liquid crystal, and a black dye having absorption in the entire visible light range is used as the guest material. Therefore, when the signal "l" is written in each pixel of the program liquid crystal panel 17b, either the p-wave or the S-wave beam light is applied to the program liquid crystal panel 17b.
7b and if the signal “0” is written,
Only p-waves are transmitted.

In the liquid crystal calculation unit 1a, data signals D! , D2 and program signals PI-P3 are respectively supplied, and depending on the combination of these signals, each beam of light from the light source section 1a is transmitted or blocked. Then, whether each beam light is transmitted or blocked is detected by the photoelectric conversion unit 3 as a valley calculation result.

FIG. 3 shows the calculation process in the liquid crystal calculation section 1a, and shows examples of calculations at eight pixel positions.

In FIG. 3, a horizontal bar (-) indicates that a p-polarized light wave is output from the pixel, and a vertical bar (1) indicates that an s-polarized light wave is output. Further, cases in which the beam light is blocked and not output are indicated by diagonal lines. Hereinafter, a representative one will be selected from among these calculation examples and explained.

The arithmetic unit consisting of each pixel at pixel position l is supplied with (1, I, l) as a program signal (Pi, P2, P3), and exclusive OR is specified as the operator. As data signals, DI=0 and D2-1 are supplied. The beam light incident on the pixel position l is polarized by the polarizing plate 11a.
The light is p-polarized and enters the program liquid crystal panel 17a. Here, since the signal "I" is written in the pixels in the program liquid crystal panel 17a, the plane of polarization of the p-polarized light wave is rotated by 90 degrees to become an s-polarized light wave, which is supplied to the data liquid crystal panel 16a. Next, since the pixel signal of the data liquid crystal panel 16a is "0", the incident S-polarized light wave is directly supplied to the program liquid crystal panel 17b. The pixel signal of the program liquid crystal panel 17b is “
1'', the S-polarized light wave is supplied as is to the data liquid crystal panel 16b.

Since the pixel signal of the data liquid crystal panel 16b is "l",
The polarization plane of the S-polarized light is rotated, and the polarized light is turned into a p-polarized light and supplied to the program liquid crystal panel 17c. Since the signal "1" is applied to the pixels of the program liquid crystal panel 17c, the plane of polarization is further rotated and an S-polarized wave is output. Then, this S-polarized light wave is transmitted through the polarizing plate 11b and detected by the photoelectric conversion section 3. 2, 1, data D
The exclusive OR operation of I=O and D2-1 is performed, and the operation result "l" is obtained.

Next, the arithmetic unit consisting of each pixel at pixel position 4 is supplied with (l, 0.0) as the program signal (Pl, P2, P3), and AND is specified as the operator. . The data pl-〇 and D2=1 are supplied. The beam light incident on the pixel position 4 is p-polarized by the polarizing plate 11a, and enters the program liquid crystal panel 17a. Here, since the signal "l" is written in the pixel of the program liquid crystal panel 17a, the p-polarized light wave becomes an s-polarized light wave and is supplied to the data liquid crystal panel 16a.

Next, since the pixel signal of the data liquid crystal panel 16a is "0", the incident S-polarized light wave is supplied as is to the program liquid crystal panel 171). Since the pixel signal of the program liquid crystal panel 17b is "0", the S-polarized light wave is blocked and no beam light is supplied to the subsequent liquid crystal panel. Therefore, the photoelectric conversion unit 3 does not detect the beam light. In this way, the AND operation of the data DI=0 and D2=1 is performed, and the operation result "0" is obtained.

In this way, in this liquid crystal calculation unit 1a, calculations are performed for each pixel position according to the set program and data, and depending on the presence or absence of transmitted light at each pixel position,
Each calculation result is obtained. Further, as shown in the above example, by switching the combination of the program signals P1 to P3 supplied to the program liquid crystal panels 17a and 17c, it is possible to switch the operators that act on the data DI and D2. FIG. 4 shows this liquid crystal calculation section L a
It shows the types of calculations realized in and the calculation process thereof, and for each type of calculation, the corresponding combination of program signals P1 to P3 is shown, and the data D
The operation of each liquid crystal panel for each combination of I and D2 is shown. In the same figure, p is the polarization plane of light that has passed through the polarizing plate, data LCD panel, and program LCD panel.
/s, and O indicates no transmission.

Note that the ferroelectric liquid crystal used in this embodiment described above has memory properties in its own material. Therefore, once a data signal or a program signal is written to the data liquid crystal panel 16a116b or the program liquid crystal panel 17a, 17c, that state is maintained unless another new signal is written. In addition to ferroelectric liquid crystals, twisted nematic liquid crystals can also be used for data liquid crystal panels or program liquid crystal panels, but in this case, the material does not have memory properties, so
Even if the data or program does not change, it is necessary to supply the data signal or program signal from time to time.

In addition, in this embodiment, the program liquid crystal panel +7
Striped transparent electrode 15a, I 5b at a=17c
, I 5e, I 5r, l 5i, and 15j, so that each pixel can be programmed independently. However, if uniform electrodes are used over the entire surface instead of matrix electrodes, the same calculation can be performed for all pixels. A liquid crystal calculation section can be configured.

In addition, in this embodiment, the program liquid crystal panel 17c and the polarizing plate 11b are removed from the liquid crystal calculation unit 1a, and
A polarizing beam splitter is arranged inside the photoelectric conversion section 3, and p
By separately photoelectrically converting the output light of polarized light and S-polarized light using two CCD sensor arrays, a function similar to that of this embodiment can be realized.

Embodiment 2 FIG. 5 is a diagram showing a second embodiment of the present invention. In this figure, the same reference numerals are given to the parts corresponding to those in FIG. In this parallel optical calculation device, a diffusion plate 18 is arranged on the output side of the liquid crystal calculation unit 1a in place of the photoelectric conversion unit 3 in FIG. Also, as data signals DI and D2, CCD camera IO
The image signals obtained by a and tob are sent to the liquid crystal calculation section! It is configured to be sent to a.

Next, the operation of this device will be explained. The two images A and B on which calculations are performed go up to the CCD cameras IOa and tob, undergo photoelectric conversion, and are further converted into digital signals within the electronic control 1115b, resulting in data signals DI (images A) and D, respectively.
2 (image B) is written to the liquid crystal calculation section 1a.

The calculations for both images are programmed by the electronic control section 5a.

The liquid crystal calculation unit 1a is irradiated with parallel beam light from the light source unit 2a, and calculations for images A and B are performed for each pixel by the liquid crystal calculation unit 1a.

Then, the beam light is transmitted/blocked in accordance with the results of each calculation, and the transmitted light is sent to the diffuser! Reach 8.

As a result, the calculation results of images A and B are displayed on the diffuser plate 18. 1- That is, according to this configuration, calculation processing between two images can be performed in a short time, and the calculation results can be displayed.

"Effects of the Invention" As explained above, according to the present invention, array data such as images can be processed in parallel for each arithmetic element, and each arithmetic element can be programmed. calculations can be executed at high speed. INDUSTRIAL APPLICATION This invention can be incorporated into an image processing system, a scientific computer, etc., and can perform high-speed data processing.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a parallel optical arithmetic unit according to a first embodiment of the present invention, FIG. 2 is a detailed diagram of a liquid crystal arithmetic unit 1a in the same embodiment, and FIG. FIG. 4 is a diagram showing an example of the operation of the liquid crystal calculation unit 1a in the same embodiment.
Figure 5 shows possible types of operations and their operation processes.
The figure is a configuration diagram of a parallel optical arithmetic device according to a second embodiment of the present invention. 1a...Liquid crystal calculation section, 2a...Light source section, 3...Photoelectric conversion section, 5a, 5b...
Electronic control unit, 10a, job... CCD camera, 11a, Ilb... polarizing plate, 12a~! 2d・
...Ferroelectric liquid crystal layer, I3...Guest host liquid crystal layer, 14a=14r...Glass substrate,
15a-15j...Transparent electrode, 16a, 16b
...Data liquid crystal panel, 17a=17c...
...Program liquid crystal panel, 18...Diffusion plate, 21...Laser diode, 22a+q2
2a... Lens, 31... CCD sensor array, pt-pa... Program signal,
Dl, D2...Data signals. Applicant: Nippon Telegraph and Telephone Corporation No. 3r1! J Figure 4

Claims (1)

[Claims] A parallel optical processing device that processes a plurality of digital data in parallel by optical means, comprising a light source and a lens system, and a plurality of beams corresponding to the number of operations to be executed in parallel in one processing. a light source unit that emits light; and a liquid crystal operation unit disposed in the traveling path of the plurality of light beams, the liquid crystal material being sealed between a polarizing plate and two glass substrates; A transparent electrode is formed corresponding to the passing position of the plurality of light beams, and has the property of [1] transmitting only a specific polarized component of the incident beam, or [2] rotating the plane of polarization of the incident beam. a liquid crystal operation section formed by stacking a data liquid crystal panel and a program liquid crystal panel in the traveling direction of the beam light, whose optical properties such as the property of transmitting the light beam are switched by an electric signal applied to the transparent electrode; and the light source. a detection means for detecting each beam emitted from the unit when it passes through the liquid crystal calculation unit and is output; and a detection unit that supplies electrical signals corresponding to the plurality of input data to the transparent electrodes of the data liquid crystal panel. and an electronic control unit that converts program data specifying operators to be applied to these input data into electrical signals and supplies the electrical signals to the transparent electrodes of the program liquid crystal panel. .