LI-RAM Uses Light Instead of Electricity

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,634
What's faster than good old electricity and those electrons we rely so heavily upon? And as we know a little bit of light can go a long ways and of course which speeds can be determined by substrate. However when it comes to household electronics, the use of light to store and transmit data can be very beneficial. This article is very light on details, but this report has a bit more granularity.


The University of Victoria materials scientist developed the item as part of an international effort to reduce the power consumption and heat produced by modern computer processors. “It really is the holy grail of computing,” CTV tech analyst Carmi Levy said. “This could be the breakthrough we’ve been waiting for.”

Frank filed a Patent Cooperation Treaty (PCT) Patent Application along with Green Centre Canada. She is currently "working with international electronics manufacturers", and expects LI-RAM to be commercialised in 10 years.
 
I was reading about this in the 80's with using light in processors, memory, etc. I guess this whole light thing is either for a show of technology or for research funds.
 
Just hurry up and build this into universal memory. I want a 100 TB of universal storage that is also memory to cut down on transit times from one part of the computer to another, we need optical interconnects between the cpu and gpu and need to let them all access the same storage/memory block. Memory will cease to be a constraint, and we will push closer to realistic vr worlds.
 
tybert7 said:
Just hurry up and build this into universal memory. I want a 100 TB of universal storage that is also memory to cut down on transit times from one part of the computer to another, we need optical interconnects between the cpu and gpu and need to let them all access the same storage/memory block. Memory will cease to be a constraint, and we will push closer to realistic vr worlds.
Click to expand...

And then we can ride Lightbikes in our gay little Tron world.
 
Well, you can do majority logic and store a binary result in the phase lock of an optical parametric oscillator.
Electronic parametrons were the solid state rage before transistors became reliable. Japan built quite a few.
Nothing but cost and ridiculous size forbids oscillating logic entirely with light. "Degenerate OPO", go figure.

Parametron PC-2.jpg IR_Optical_Parametric_Oscillator.JPG 150KHzParametron.jpg
Only last photo is mine. Note in all above cases: phase locking pump was 2x the logic frequency.
If locked to even pumps, stores a zero. Locked to odd pumps, stores a one. Entirely arbitrary.
A reference one or zero is phase locked for the system before any further logic is meaningful.

Non-linear saturable inductors, varactor diodes, or optical crystals exploited to pump energy into
an oscillator and maintain the state. So its DRAM of sorts, I guess. I have personal experience
only with the non-linear inductors version. Not tried any of the other ways mentioned above...

You do logic by stopping the pump till the oscillator forgets. Mix the inputs, and re-start the pump.
Whichever phase was in majority gets amplified up (about 40x) and locked for long as the pump
continues to run. Can now take those weak inputs away, or change them, strong local oscillation
won't lose phase lock to itself or the pump.

Majority gate logic is a superset including all normal AND, OR, NAND, NOR stuff. Inverting is a
freebie! Just reverse wires, delay 180 degrees, or otherwise invert the phase. Full addition with
carry can be calculaled in just two gates. Store both carry and sum for long as needed, also free.

So, there are usually three pumps, each assigned to a bank of parametrons. Each bank shut-down
in turn. To allow forgetfullness of the prior state, and allow the flow of logic to be pulled forward on
bank re-start. History proves completely normal computers can be made of this crazy stuff.
 

Attachments

  • NEC.1103.1958102646285.pdf
    810.9 KB · Views: 21
Last edited:
https://phys.org/news/2017-04-optical-micro-oscillator-next-generation-applications.html

This optical oscillator serves different purpose. Important takeaway 1 cubic centimeter!
If OPOs can be built at that same scale, an entire optical parametron computer might be
assembled, about the same volume as historical parametron computers.

Not really practical, or offers any advantage. Just saying that it easily crosses theshold
of possible.

One thing we didn't ganulate is memory: Except OPO to store 1 bit, which is nothing.
Old parametron computers kept memory in non-volatile magnetic cores, which were
more dense and far lower power than if it had all been kept on oscillating logic cores.
For light, I don't propose a crapload of live optical oscillators beyond whats needed
for registers and cache. Nor do I propose MRAM, though its plenty interesting stuff.

Instead, I suggest delay line memory. Send a stream of binary phase shift key'd light
down a mile of fiber, clean up the bits (phase lock and amp them back to full scale as
parametrons easily do), send around the loop again and again. Now we got memory...

Anyone not familiar with the delay line concept, visit this article:
http://hackaday.com/2012/11/09/storing-32-bytes-of-data-in-a-piece-of-glass/

------

And we don't have to go the way of traditonal computers. Oscillating logic can work in
both directions simultaneously, its actually a bit of a trick to make it flow one way only.
So maybe useful for simulating some types of quantum computing situations.

Wether with light, microwaves, or plain old electricity. We really need to re-understand
parametrons and majority logic before the quantum stuff gets here, else we will be lost
when truly mind-bending things happen.

Example1: Start a sensitive parametron, but no inputs other than ambient noise floor.
Makes a pretty good random number generator. XOR two of them for really random.

Parametric amplification is exponential, so anything too small to merit a gain greater
than one, decays away to nothing. Anything over that threshold, amplifies up. To the
pump amplitude, beyond which energy spills back across the pump. So, exponential
gain can't build-up infinately, but diminishes to unity at known repeatable amplitude,
we will call "full scale".

It is possible to make a parametron that isn't sensitive to the noise floor, and will
store a superposition state of not oscillating, till a strong enough input comes along.
Trinary logic anyone???

Example2: Bidirectional multiplier. Drive inputs, and it multiplies. Drive outputs, and the
inputs give you two random factors that multiply to that answer. Can be used to check
itself. If after significant trials, all you ever get is X and 1, or 1 and X, then the number
in question might be prime.
 
Last edited:
You must log in or register to reply here.
Back
Top