The idea of a computer that uses light for computing is facinating to me.

And I would like to ask the forum as to ideas as to how light can be utilised to generate the differentials needed for computing functions.

From what I understand, which is very little, a micro chip relies on voltage differences to achieve it's switching capacity....How could light be used to achieve this function?

Are we talking about visible light only or other?

Do we use microscopic optical fibres?

Do we use only two colours ( frequencies ) or do we go the whole hog and use all frequencies and dump binary languages all together?

What photo senses do we use to detect frequency changes?

An article about 10 years ago mentioned an interesting concept which might be usable: A light delay line for use as memory storage.

An early computer (Univac) used acoustic mercury delay lines as memory. The idea is that a sound wave is sent down a tube containing mercury. The frequency of the sound is used to encode data. Electronic mechanical circuitry at one end of the tube reads the frequencies, corrects for any distortions, and sends the signals back to the other end of the tube. The data encoded in the sound waves is continuously traveling in the mercury and the electronic circuitry. Encoded data can be retrieved as required and replaced with new data when pertinent. It sounds gimmicky, but it is usable as a memory device. It is analogous to data on a rotating disk or drum.

There were slow, now archaic computers which used rotating disk and drum memories which had a read and a write head positioned 1/20 to 1/4 of a revolution apart. They were analogous to the mercury delay lines used by Univac. The data was continuously read and rewritten. The transfer rate for such a disk or drum was the same as for a standard disk or drum, but the average access time was much shorter. You would wait a maximum of 1/20 to 1/4 of a revolution for the data you wished to read or replace (average latency 1/40 to 1/8 of a revolution). Such devices had a read and a write head for each track. They were used as memory before coincident core memories and other technologies were developed. Even in that era, a standard disk or drum was too slow for use as CPU memory.

Light taking a circular path in an optical fiber can be used as memory in a manner analogous to the mercury & magnetic disk/drum delay lines described above.

The above is reliable information, the following might not be so reliable. A search might turn up some pertinent articles which are more reliable and informative.

There are materials (I think selenium is one such) that allow electron flow when a light shines on them and are resistant to electron flow in the dark. These materials are used in doors which open automatically when somebody approaches. Light activated electronic switches can be made using such materials. I also think that there are materials whose light transmission properties change when a voltage is applied.

The technology used by fiber optic phone lines can be used in computer circuitry. The phone lines require circuitry which reads, writes, and otherwise processes the data being carried in the optic fibers. At intervals along the way, the signals must be read, corrected for distortion or loss of amplitude, and retransmitted. Surely such fiber optic technology circuitry could be used as the basis for computer circuitry. At the short distances involved in a computer, the data rates would be comparable to those achieved with current electronic circuitry.

Thanks Dinosaur for that insight, a queation for you...Do you see any benefit in making binary languages obsolete?

0's and 1's seem so limiting.

Do you think that if we can utilise more than one variable ( 0-1) would be more how can I say....hmmm...articulate?

Quantum Quack: A long time ago in this galaxy, it was believed that some electronic devices with three stable states had been invented, and that transistor switches could be manufactured to switch among the three states. It was decided to design and build computers using these devices. Such a computer would use trinary or Radix-3 arithmetic instead of binary or Radix-2 arithmetic. People were very prudish then, and the entire concept was discarded because the trinary digits would be called tits, corrupting the morals of our children and infuriating the feminists.

Kidding aside, binary arithmetic using zeros and ones as the only digits is inherently more efficient than using another radix due to the binary nature of basic electronic components. A CPU using pure decimal circuitry and components, would use ten voltage levels to represent ten digits when transmitting data. Magnetic media would use ten magnetic flux strengths to represent ten different digits. There would be serious problems if the differences in voltage levels & flux strengths were very small. Slight amounts of noise might turn a 7 into an 8. The If the differences were large, the largest voltages would require using large amounts of power compared to what is used in a standard CPU. There would be all sorts of problems and circuit complexity avoided when using solid state basic components with two stable states. Such components highly favor the use of binary arithmetic.

In the fifties, there were some computers designed using decimal arithmetic. Some used standard electronic components. 4 bits were used to encode each decimal digit from 0 to 9. Obviously, 4 bits could be used to encode numeric values from 0 to 15, making hexadecimal encoding more efficient. Binary coded decimal was wasteful of CPU circuitry components and wasteful of space on magnetic media like tapes, disks, drums. Furthermore, the circuitry to do decimal arithmetic is slightly more complex than the circuitry required for binary arithmetic. Note that one decimal digit can only encode 10 characters. Two decimal digits, using 8 bits, will encode 100 characters. 8 bits will encode 256 characters or 128 characters and a parity bit. Computers using binary coded decimal were given up as a bad idea. The Burroughs Corporation was innovative in many ways. If solid state electronics had been delayed about a decade, they might have built a decimal computer based on a device called the BEAM Switching Tube. It was to be called the Burroughs Electronic Accounting Machine using the BEAM Tube. This device looked very much like the vacuum tubes used in all electronic devices prior to the development of solid state electronics technology. It employed a rotating source of electrons (cathode?) that could point at one of ten targets (anodes?). Excuse me if my memory of vacuum tube jargon is faulty. Each of the targets represented a decimal digit. This vacuum tube was essentially an electronic component with ten stable states, which could be used as the basis for a decimal computer. The idea was killed when it became apparent that vacuum tube technology was on the verge of being obsolete, and a computer based on BEAM tubes would not survive long enough to recover the capital investments.BTW: The BEAM acronym was a play on words, since it was suggestive of the rotating beam of electrons used internally as well as being initials of a term.

If tri-stable basic electronic elements had actually been invented, computers using them would have been built and they would have used Radix 3 arithmetic. For many purposes, Radix 3 arithmetic and logic is more efficient than radix 2. For example, consider a balance scale which allows the placing of weights on either side, which is a trinary scale, while an ordinary balance scale is a binary scale. Four trinary weights (1, 3, 9, 27) allow weighing up to 40 units, while five binary weights (1, 2, 4, 8, 16) allow weighing only up to 31 units. If tri-stable components could have been built and if they were as fast (or nearly as fast) as binary solid state components, a CPU based on them would be more efficient than a binary CPU.

There actually was a brief period of time when it was believed that such components had been invented. I investigated the type of logical design that would be suitable to such systems. The project was scrapped when the researchers realized that they were mistaken about the properties of the solid state devices they had developed.

A radix 3 system is very interesting. It would be most efficient to use what is called balanced radix arithmetic. The three digits values are plus one, minus one, and zero. These digit values are represented by + - 0 in the following. Counting is as follows. 0 ---- 0000 1 ---- 000+ 2 ---- 00+- 3 ---- 00+0 4 ---- 00++ 5 ---- 0+- - 6 ---- 0+-0 7 ---- 0+-+ 8 ---- 0+0- 9 ---- 0+00 10 --- 0+0+ 11 --- 0++- 12 --- 0++0 13 --- 0+++ 14 --- +- - - . . . . 40 --- ++++The rules for addition are as follows. + and + = - with a plus carry - and - = + with a minus carry + and minus = 0 0 and x = x Try testing the above rules by adding + to each of the above to get the next value, or adding a minus to count backwardsTit positions are weighted 1, 3, 9, 27. . . for integers. They are weighted 1/3, 1/9, 1/27. . . for fractional values.

Negative numbers can be obtained by reversing the plus & minus signs, while leaving the zero digits as is. In radix 3, complementing is simpler to implement in circuitry than it is for radix 2, because radix 2 complement requires reversing ones and zeros, followed by adding one, which can cause carry ripple.

In radix 3, there is no sign position: If the highest order tit is +, the number is positive; If the highest order tit is -, the number is negative. No special rules are required for adding positive to positive, negative to negative, or positive to negative. You just follow the rules for adding tits from right to left, implementing a carry of + or - when required. A carry out of the high order tit is overflow.

Note that in radix 2, overflow occurs if the carry out of the high order (or sign) bit is different from the carry into the high order bit. The binary rule is slightly more complicated, requiring a bit more circuitry and time.

Note that carry ripple in a binary CPU tends to be longer than carry ripple in a trinary CPU due to the binary system having all positive digits, while some of the +/- tits cancel, stopping carry ripple. Carry ripple is a serial process, which slows down the average time required for addition, which is otherwise implemented as a parallel operation in a modern CPU. A carry ripple over 12 bits can take 3-9 times as much time as an entire addition with no carry ripple. A trinary system can use less digits and tends to have short carry ripples, making it inherently faster to implement addition in electronic circuitry.

Another slight advantage to balanced trinary is that rounding and truncating are equivalent.

There are some disadvantages to balanced Radix 3, mainly relating to division.

Facinating...will take a while to sink in, but in the mean time can you see, say, three colours being used to represent trinary......say (-) = red. (0)=blue and (+)=green and would there be any benefit in doing so? Using three light frequencies instead of electrical voltages.

Facinating...will take a while to sink in, but in the mean time can you see, say, three colours being used to represent trinary......say (-) = red. (0)=blue and (+)=green and would there be any benefit in doing so? Using three light frequencies instead of electrical voltages.

There is some physical limitations that limits the electric computers development. For example because of processes like interference the size of the mainboard should not be bigger then half the wavelenght, so if we have an mainboard with frequency 10 GHz, its wavelengh will be c/10^9 = 30 cm, and its size can not be bigger then 15 cm. Can you imagine all your PC components - CPU, PCI, AGP, memory slots and etc, nested in 15 square cm?

Not to mension the other problems that arise from increased frequency - heating, transistors break-throughs and etc. So even in the near future (say 20 years from now) optical exchange of information would be perspective for computer industry. They can replace data bus with optical fibers and at least the problem with mainboard size will be solved.

But for a totally optical computers can not be talked at all. Light is very hard to be manipulated, it is not infuenced from voltage changes like currency, how light would be manipulated? Also photons can be absorbed or emitted incidently and such a computer should operate at temperatures close to the absolute zero. I think that a totally optical computer may be efficient only if quantum characteristics of the light are used.