Nano-Particle ALU Retains Vintage concepts

[RJSV]

Hi, here are some details, on a hypothetical Molecular Scale Computer, focus on Boolean process.
I'm thinking 512 K separate ALU 'channels', running at, I don't know, how about 900 GHz (?).

   Diagram starts off fundamentals, for using particle collision dynamics in a practical ALU subsystem.
It does not show the 3D aspects, but you can see: there are two input, moving particles arranged and timed for accurate collision zone. Now, since depth axis is not shown, please assume each in the particle pair will experience a glancing blow, each off of the other.
   So you could work out the varieties of result:
In a case where only one (particle-signal) has entered the 'collision zone' that single particle can be captured to exit, and literally represents an XOR outcome. That is, an EXCLUSIVE OR signifying that only one of two 'input' signals is present, a binary 'ONE' bit.

Reader's can note, on the various outputs, the XOR function result can be taken from two duplicate places, that is, can be viewed as each incomplete. To solve this minor dilemma one merely has to combine both partial XOR outputs, using a simple funnel structure, in the crystal 'substrate', or whatever that implementation needs, to 'OR' combine for a conventional XOR result.
Considering the case for the (two) logical AND outputs, it appears that either signal can be chosen, while other signal AND can be discarded, assuming not needed further along (as a signal duplicate).
It appears this scheme experiences difficulty producing logical OR signal; that could be generated using combination of AND (either output), with each of XOR1 with XOR2.
Thus portions of full Boolean operatives can be built, the limits being when each type of output is demanded, all at once. So some work-arounds needed for that, general, case.

Now, of course, processing for doing logical invert (NOT) that is problematic... But so is organizing 512,000 ALU units, together, into a full molecular computer.
Yes, that's called 'forward thinking', I heard Einstein got called that, or worse. (Lol)

[RJSV]

So, this thread is very speculative, and perhaps un-professional, not being a specialist. But HEY: maybe I'll take things (nano machine ideas) to unusual but useful places!
  The crystal substrate I am proposing would be porous in the sense that micro 'tunnel' structures would act as conduits for deliberate planned connections, component to component. Maybe even utilize new 'drills' for manufacture of those endlessly numerous paths needed, remember there is a half million 'cores', each I'd like to be 64 bits.
  The particle collisions demand a crazy degree of coordination, and that one collision (illustrated) can be timed by (somehow) controlling a variable delay in the transport conduits.
  Another problem: devices that small are not going to collide and have predictable paths. I wonder if particle shape can be a part of the design, such as hexagonal faced 'sphere-like' nano-particles used for the computer signals.

   Diagram shows conduits combined to generate a logical 'OR' using each of the two 'XOR' signals with also the 'AND' indicating signal. The simple use of nano sized 'funnels' seems silly, maybe wildly impractical.
   Next, I ' d also like to consider a DECIMAL multi-state data transfer (BUS) for the  works.
Pls see diagram.on logical OR.

[RJSV]

EPROM -Like SOLUTIONS  for  ALU
While the (crazy) kinematic contraptions can be entertaining, it's problematic, making two such micro sized particles collide, reliably.
 
  A look-up table can provide an (ALU) - Boolean style response, just maybe not so 'live' seeming.

So, to illustrate (the problem aspect), of using particle processes (collisions), see diagram:
  Of two 'signals', first arriving will set the wheel, and the other signal will get a straight path, while currently, the early arriving gets a slightly 'curvy' path, thus a slight delay gets conditionally added.
And so: What is VINTAGE about such designing, of nano-machines ?
The 'Truth Table's use, in a PROM, helps to avoid this 'flakey' seeming logic !

[RJSV]

Another diagram is included, of a hypothetical 'time synchronizer', for two particles ,( to collide).
In front, the wheel housing for particle 'm.' will have particle travel down through, then using the right side travel path, taking slightly longer time.
   In the second housing, shaft - tandem, for particle 'n.' will have been pre-positioned, for the shorter, straight line path, shown as left side (of the back housing shown).
   
   Now, of course, you would need many stages to do correction, for time coordinating. Possibly, other unpredicted variations could end up actually ADDING to the timing un-certainty.
   At any rate, much of the basic concept can be salvaged, albeit through drastic change: A micro-machine 'wheel' flip flop takes on one (of the particle operands), while a second operand can then do some interaction, just maybe not a wild 'collision'.

   The basic 'wheel' component is scaled fairly large, in nano-particle terms, at something like 2000 atoms across. Each processor having approx 30 'features' across, (also by 30 features, up and down).
So, a complete module would have 4096 simple processors, as 16 by 16 by 16, in a cube.
My est came to roughly 2 mm on a side, but I'm still working on that
  Each processor would be one and two digit numbers, with simple instructions set, and perhaps several hundred 'digits' of storage.
   More like super- state machines, with ALU built-in, really.
And one promising BUS resembles vintage Teletype using 5 state bits, plus a Bank bit.

[RJSV]

The 'Time Synchronizer' (see 2nd diagram back) shows the same type 'toggle wheel' that can be used in other contexts.  So, for a logical AND Boolean function, you write the first operand bit to position the wheel. That is what I mean, by 'half' of colliding the two operand particles.
Once the ALU function wheel is positioned, a read is done, using a 'reader pulse' (nano-particle). So for an AND result, that second particle would emerge, on the output side of the Toggle wheel. Actually, for more conventional cases, having either operand as 'zero' creates a passive situation. Remember: electronic bus data uses 0 volts, which in physics view is a passive thing. Electronic digital circuits often get around that, right away, using 'pull-downs' and sometimes by supplying a default zero, by active means. (Then, a subsequent 'passive' input simply leaves the gate in a sure zero state.

   Now, for more background, please refer to the THREAD on 'Print this 3D mechanical memory.

Readers potentially think, that the nano-scale ALU and system kind of COPIES that other thread on mechanical computing, but it's the reverse:
That is, the former thread is actually a derivative of nano-scale designs, (and signalling standards). It was envisioned that a table-top mechanical computer, using 1/4 inch steel balls, could be a R&D platform.

So, anyway, for some more distinctions, some design is for single bit (Boolean), while another context is the look-up table implementation. There, having one 4 bit operand working against a second 4 bit operand, results in a familiar 256 location (8bits of addressing) PROM.
   For more on that, see Ben's thread on PROMs for doing logic functions.

Next I want to fill out more details, on miniaturization, and various estimates on sizes. I found that packaging issues, involving height, weight, and appearance kind of get a back seat, to all the glorious genius circuitry and math involved, So: Hats off to the Mechanical Specialists, they are the ones making the 'packaging' happen!

[RJSV]

So far, you've read about a bunch of speculation,by an engineer / coder, sitting in a back yard somewhere (wuff wuff).

   Some interesting points can be stated, having ALU logic that doesn't use a conventional BUILD-UP, that is that first you create a logic gate, of some variety, and then next you combine gates to build, say, a binary ADDER, and then, exhaustively, combine hardware and (assembly) CODE to have a floating point SINE Function.

But, then, in creation of a DECIMAL BUS system using a PROM style ALU, you've got direct access to a complicated math function, such as SINE (X), just as easy as A + B is resolved. Or nearly so, as the SINE function table only uses 0 thru 90 degrees, as input.
AND, of course you might want a multi digit answer, in floating point format.

   It's maybe hard to convey, but much of this is in VINTAGE territory, makes me wonder when those floating point ideas started. (Probably in early mechanical calculators).
  I know that Steve Wozniak didn't t invent floating point stuff, but he did write 6502 code doing that.

As to why speculate on a 3 mm SuperComputer using magnetically accelerated nano-particles (and possibly also requiring cryogenic cooling, well, simply put:
  ITS FUN, doing that... That's what Engineers do,
right ?