2002-Jan-10, 02:26 AM
I see a lot of coverage on up & coming photo computer models, but thier doesnt seem to be much enthusiasm surrounding the whole affair.
Comparable I guess to working with silicon when every other Jockey was scratching the cats whiskers with a very long scratcher.
Possibly astronomy has the most to look forward to when considering the implementation of ideal photonic computing systems, both analouge and digital.
1.---------------------------------------
Spectral analysis of a deep field, electron Vs photon.
Digital -> when the first set of pixels arrive at the processor, the number of electrical events needed to get there is disgusting. Even in a RISK or analouge setup you still rely on gadzoogles of seperate electronic signals ( and pathways ) to mutate a photon into a digital or analouge electrical values. Now all you have to do is convert a few billion sets of a million members into a rather large binary sequence.
Photonicly -> in a nut shell, use the data as it is, keep the input simple and leverage photo tables to 'decode' spectral info on the spot. What is a photo table? Using one souce of 'light' to calculate phase of some other light 'source' of similar wavelength takes nano seconds . ( 2 * T source ? ).
---------------------------------------
Bianary addition of 2 bits ( a & b )of laser light. Groovy. ( assume a & b have same F & Amplitude ).
step One:
Use an expensive lens to partialy refract each bit into two 'copies' of the partcular bit. ( a becomes a1 & a2 ... )
Take a1 and b1 ensure they are pi rads out of phase then use another expensive lens to add a1 & b1 . When a = 1 & b = 1 then a1 + b1 = 1. . When a = 0 & b = 1 then a1 + b1 = 0. ( the first bit in the sum ).
Now take a2 & b2, put into phase and treat as a carry ( carry is TRUE when a2 + b2 = 2x watts, FALSE when a2 + b2 = x watts;
Now do it with a large binary number and cascade this result into other photonic words.
Dont forget, when multiplying, the period of a cascade must be greater than the time taken for a transistor to detect that the result of factor 1 minus 1 repeated is zero ( you have added factor 2 to itself untill the transistor signals as above ).
Please add your designs here.
Comparable I guess to working with silicon when every other Jockey was scratching the cats whiskers with a very long scratcher.
Possibly astronomy has the most to look forward to when considering the implementation of ideal photonic computing systems, both analouge and digital.
1.---------------------------------------
Spectral analysis of a deep field, electron Vs photon.
Digital -> when the first set of pixels arrive at the processor, the number of electrical events needed to get there is disgusting. Even in a RISK or analouge setup you still rely on gadzoogles of seperate electronic signals ( and pathways ) to mutate a photon into a digital or analouge electrical values. Now all you have to do is convert a few billion sets of a million members into a rather large binary sequence.
Photonicly -> in a nut shell, use the data as it is, keep the input simple and leverage photo tables to 'decode' spectral info on the spot. What is a photo table? Using one souce of 'light' to calculate phase of some other light 'source' of similar wavelength takes nano seconds . ( 2 * T source ? ).
---------------------------------------
Bianary addition of 2 bits ( a & b )of laser light. Groovy. ( assume a & b have same F & Amplitude ).
step One:
Use an expensive lens to partialy refract each bit into two 'copies' of the partcular bit. ( a becomes a1 & a2 ... )
Take a1 and b1 ensure they are pi rads out of phase then use another expensive lens to add a1 & b1 . When a = 1 & b = 1 then a1 + b1 = 1. . When a = 0 & b = 1 then a1 + b1 = 0. ( the first bit in the sum ).
Now take a2 & b2, put into phase and treat as a carry ( carry is TRUE when a2 + b2 = 2x watts, FALSE when a2 + b2 = x watts;
Now do it with a large binary number and cascade this result into other photonic words.
Dont forget, when multiplying, the period of a cascade must be greater than the time taken for a transistor to detect that the result of factor 1 minus 1 repeated is zero ( you have added factor 2 to itself untill the transistor signals as above ).
Please add your designs here.