The first photonic logic components built using laboratory test models of Photonic Transistors produced the Boolean logic functions: XOR (pronounced 'exclusive-or', NOT, and an OR. According to the rules of Boolean switching logic, various combinations of only two of these basic functions are all that are needed in order to construct all of computing!
That first transistor was also used to produce an elementary photonic signal amplifier for increasing the amount of energy in a modulated input beam. Amplifiers are also needed in order to interconnect working logic functions into functioning computers. Of these, the component with the fewest number of complications in its output configuration was the NOT function.
In electronics, the NOT is called an 'inverter.' In photonics, when its input beam is on, it produces an off output...and an off input produces an on output. Since at least one of the transistor's inputs must be on in order to produce energy in the output, the 2nd input beam accomplishes exactly what all the 'experts' said was impossible. One beam was used to turn another beam on and off without some electronic do-hickey in the middle.
The OR function works like the dome light switches in your car. When one door OR the other door, OR both doors are open the dome light is on. All of the doors have to be closed in order to turn off the overhead light.
The XOR is like the light over your basement stairwell. When the upstairs switch is off (down,) along with the downstairs switch, the light is off. When they are both up (on) then the light is still off. Only when one is on and the other off (in the opposite positions) does the light go on. (Unless of course if your electrician put one of them in up-side-down.) With light, one of the most curious things is that darkness (at the minima) can be produced by beams that are on. That is, light makes darkness. It is this basic quality of light that allows us to make an XOR, and do all sorts of other neat things.
To understand how it works recall that the photonic transistor is based on three basic principles:
The best way to really begin understanding photonic transistors is by reading and studying (with an open mind) the various patents that are available through this web site or directly from the Patent Office. The brief examination here provides only a cursory examination of photonic transistors, and is by no means an exhaustive explanation of the subject that is meant to answer all of many pertinent questions that come up. We will cover some of the frequently asked questions, however.
It is commonly believed that photonic computing cannot be accomplished using optical interference because it is a "linear" process. That is, when two or more tiny laser beams are superimposed onto the same spot, at the same time, the energy redistributions that occur follow the laws of linear (algebraic, ie. + and -) addition of amplitudes. However, as will be shown, the holographic photonic transistor affords us the opportunity to use all of the various combinations of electromagnetic energy that the laws of physics allow. And that, while there are complications that occur, we can keep track of these from process to process and provide the needed corrections during the process of designing the computer-generated holograms. These holograms will then be manufactured into commercially available photonic computers ...in the very near future.
To begin the process we need to start some where. The place to start is at the elementary Boolean logic level, the basic photonic switches that are the heart of digital computing.
Many electronic-imitating Boolean logic devices, which are the basic transistor circuits used to build digital computers, have two inputs that at various times are either on or off. This produces 4 different possible configurations:
When the 1st beam is on by itself, a generally consistent distribution of energy occurs over the cross section of the beam having a phase at each location that depends upon the geometry of the optics. The actual pattern has a distribution of energy wherein its amplitude and phase may be precisely calculated for every location in the Dynamic Image. Thus, every possible waveform for every location can be determined quite accurately.
When the 2nd beam is on by itself, a similar energy distribution occurs, however its phase distribution is different from the 1st beam image even if we align the images so that their amplitude variations match. This is because the 2nd slit is not at the same location as the 1st one, so the wavelength-unit distances to each location in the output image from each location in the two inputs will be geometrically different. It is these physical distances that are used to calculate the phase of a ray that is expected to show up at a particular location. Then we can sum its amplitude in with all of the other amplitudes of all of the other rays that are supposed to arrive at that location at the same time.
When input both beams are on together, an energy redistribution occurs so that the energy becomes concentrated in the areas where energy from the two beams is naturally in-phase due to that geometry. The greatest amount of energy concentration is at those places is called the "maxima", and those places that have a minimum amount of energy are naturally called the "minima". The maxima is said to be produced by "Constructive Interference" (which I often abbreviate as CI). The minima is said to be caused by "Destructive Interference" (which I abbreviate as DI).
In between the maxima and minima there is a range of energy distributions from weak low level signals near the minima to strong signals near the maxima. The optics can be designed so that there is little or no phase shift between the various states at the location where the maxima shows up when both inputs are on. At the location where the minima shows up, the phase shifts by 180 degrees between the single beam two single beam input states and goes dark when both inputs are on. The places in between the maxima location and minima location can have all sorts of phase fluctuations that we will put to work in later devices.
The terminology used to describe the continually changing patterns of the dynamic image can get confusing. If we choose a naming convention that is defined during the state when both inputs are on, we can call also call those locations 'maxima' and 'minima'. If we place mask with a hole in one such location, we can thus describe where it is. The problem is that when we modulate the inputs, other images are formed. The energy distributions in these images may change considerably at the location of hole, and thus the amount of energy that makes it out the hole changes. But, once we have placed a hole in one place, and given it a name that comes from the two-beam state...we don't move the hole around. It stays in the same place. To keep track of it, we keep the same name even though it may not be a maxima or minima in those other states.
So, by placing an output hole in the separating mask that encompasses the maxima and its vicinity the output will be very nearly like the precise center of the maxima. And one placed at the minima and its vicinity will very nearly match the activities at the exact minima...when both inputs are on.
The XOR:
An output hole placed at the location of the minima provides an output when either of the input beams is on by itself. However, because 'light makes darkness,' a minima occurs over the output hole. No energy arrives at that location, so nothing goes through the hole. Thus, the output when both inputs are on is off. In Boolean computer switching logic terms, the device is an XOR. Just like the stairwell illustration.
There is, however, a 180 degree phase shift that occurs between the two single beam states. This introduces a phase modulation component that must be compensated for in succeeding components. It is important to note though, that in spite of the phase shift, the needed logic information has been extracted by a combination of its inputs. How we use that information depends on what is needed in succeeding logic stages.
However, without interference, and without energy separation from the components of the Dynamic Image, as detailed in our first patent, no XOR information is extracted! Such steps are vital for the creation of photonic computing.
The NOT:
As with any XOR, if one of the beams is kept on all the time, the device becomes a NOT. That is, when the 2nd input is on, the output is off and vis versa. With the NOT, there is no adverse phase modulated component, because the reverse phase state is not used. Since that one beam is kept on as a power supply to the device, and the other beam causes its energy to either exit the hole or not exit the hole, that 2nd input beam is actually turning the power beam on and off.The OR:
What happens if the hole is moved over to the maxima, the CI position? Now energy appears in the output whenever any of the beams are on. While there is a variation in output amplitude that must be compensated for, this Boolean device is an OR. Just like the car door illustration.
The all Photonic Amplifier:
It requires only two basic Boolean devices in order to produce ALL OF COMPUTING, and we have three of them in that first patent alone. However, in order to make up for energy loss from one device to another, one needs an amplifier. So, if one beam is kept on all the time as a sort of photonic power supply, and the 2nd beam is switched on, the output through the maxima-positioned hole jumps from the single beam level to 4 times that level. (The reasons for this are discussed at some length in our other articles on the basics of interference.) Thus, the information-carrying portion of the output has 3 times as much energy as the original modulated input. Thus, the invention is also a light speed amplifier. If two such amplifiers are interconnected, just as in electronics, the result is a flip flop, a light speed binary information storage device. Two of which are described in that first patent.
| FAQ | Photonics Menu | Home | Previous | Next |
Interference is not something that is easily accomplished on the macro scale. But then, we are usually not interested in making big transistors. Little ones are what we want. So how small can they be made?
The 3M company has demonstrated its ability to produce 20,000 independent holographic-like lens on a single square centimeter of material. While there's no reason to imagine that is the limit for making small scale devices, certainly it's a fine start. Unlike the economic vitality (or lack of it) in the electronics world that depends upon the ever-increasing cost of silicon real estate, the inexpensive glass, plastic and aluminum that will be used to make photonic computers permits one to use as much material as is needed. Certainly there's no reason why photonic computers cannot eventually be made even smaller than today' lap tops. When they do, they certainly will have a lot more horse power.
Most of the physical limitations are addressed through other parts of our intellectual property. Some these will be discussed below under "Patents and Patents Pending".
| FAQ | Photonics Menu | Home | Previous | Next |
Light really moves. In one second, electromagnetic energy can circle the earth seven and a half times in one second. In one nanosecond, (one billionth of a second,) light goes 30 cm, or about 11 3/4 inches. By measuring the dimensions of the smallest working model of our photonic transistor we can calculate the amount of time it takes light to pass through the device in order to accomplish the above photonic logic and amplifications functions. In a working photonic computer, these will be the switching times used to determine how fast we will be able to make a photonic supercomputer go.
First by way of comparison, so that we can realize the import of what has been accomplished, the electronic transistors that perform logic in a $5,000,000 Cray III supercomputer are able to switch in about 0.25 nanoseconds using expensive gallium arsenide transistors. That translates into a 2 nanosecond clock cycle time which in more familiar terms is 500 mhz. So the new 200 mhz Pentium is getting up there. Of course it's tough to make $2000 Pentium machines out of materials that are needed to build a $5,000,000 Cray.
Our test transistor was made using a piece of glass so that we could easily hold the image component separator still in relation to the beam combining optics. By calculating the distances through the glass, and the attached mask, the transit times were able to be calculated. Each one be built was progressively smaller until we reached the point there the transit time for 632.8 nm, red, laser light is 0.007 nanoseconds!
The Cray reportedly has transistors that switch in 0.25 ns. In that case or roughly 35 times faster than the Cray.
Some will call such a device 'crude.' Well, have you ever seen a picture of the first electronic transistor? It looks 'crude' as ever. However, it had one over-riding quality...IT WORKED!. That first example of semiconductor based amplification was never used to do anything other than to demonstrate how transistors work.
The crude examples that we have produced, likewise have one over-riding feature, THEY WORK! They work exceedingly well. They are real and functioning transistors that do what no prior photonic device is capable of doing without expensive and pulse-time consuming processes. They use light beams to turn other light beams on and off. A thing that all the so called experts said was 'impossible!'
For a long time after the electronic transistor was invented, AT&T couldn't get a dime out of the major electronic companies, all of which had a considerable investment in vacuum tubes. They had loads of people, with all sorts of 'proofs' that transistor computers would never work. Not until they contacted a little- known company in Texas that made electronic instruments...Texas Instruments, who had no stake in vacuum tubes, did the technology take off.
I guess people are always that way. When Thomas Edison invented the phonograph, he took it to France to demonstrate it in front of the French academy of sciences. After the demonstration, two of the most prominent members of the society stood up to pronounce, the phonograph they had just witnessed working, as a hoax. They said that it was impossible to make a machine talk, and that the American doing the demonstrating was a ventriloquist!
But, back to the subject. What really is Photonic Transistor's speed limit?
Pipelined Pulses:
If the transit time through an electronic transistor is one nanosecond, the input must remain either completely on or completely off for that full nanosecond. Otherwise considerable noise will be introduced into the system. The Photonic transistor, however, is able to operate using pipelined pulses.
That is, a continuous stream of very short pulses can be introduced into a single transistor, pulses that are much shorter than the transit time of the device, and they will all be processed independently without any noise buildup.
Just as information pipelining is an important part of the architecture of the Pentium and many supercomputers, so too, pipelining information into the various light beams that make up a photonic computer can greatly increase its throughput. But how fast? What is the theoretical limit?
The theoretical limit for the shortest pipelined pulse would be equal to the period of oscillation for a one-wavelength-long pulse. Now I didn't say that this limit is easy to reach. I merely said that it seems to be a limit. If it can be reached, the switching time for that same red laser light would be 2.1 femtoseconds! A 'femtosecond' is one millionth of a nanosecond.
If a shorter wavelength is used, the pulse time is shorter. If 300 nm ultraviolet light is used, the switching time is 1 femtosecond!
However, such switching time comparisons to today's electronic computers do not take into account light's ability for accomplishing massively parallel computing. That is, by doing millions of things at the same time, far more work can be accomplished. Channelizing the visible part of the spectrum provides over 4 billion separate channels. Photonic transistors are capable of operating using all of them individually and all together. They can be manipulated as easy as forming the right kind of dynamic images and separating the appropriate energy patterns from them.