CPU CLOCK Times Mechanical Computer Systems

[RJSV]

  It's fun (see Dave's blog #1332) to watch mechanical workings in motion.
   This topic covers Mechanical 'Ball Release' clocking, useful for keeping CPU and all the rest, on a timed sequence, very similar to the discrete clocking used in today's machines.
   I started out needing a 'wide' or parallel (word) output, or actually more accurate: 'Outlet', as the particular mechanism for building a mechanical CPU needed conventional acting clock impulses.
   Pls see enclosed diagram:
In the diagram, I show a rotating cylinder resembling a water wheel, only it is using the weight of individual steel balls.
Above the cylinder is a sloped box: it has several lanes and with each lane there is a hole for each ball to drop. When that happens, the ball falls and will strike a release lever, on a pawl latch style release (green lever shown).
   Now, I'm trying to illustrate a CONTROL LOOP here, as each (ball) release step will result in actuation of that pawl shown, with resulting repeat rotation of the gate cylinder, thus causing future release on next cycle.

   This creates a 'closed loop' control, of the start-stop cylinder (but with lots of variable friction elements to it.) By placing an additional 'regulator' into this mechanical timing loop, better control and accuracy can be had. A closed loop device, using fall times, can be placed within the main cylinder-lever parts, for ultimate regulation.
   In the diagram, the 'release pawl' is not shown in detail, but is constructed so that the cylinder moves in a start-stop fashion: moving approx. 22 degrees for each ball release clock time. Most prototypes run at around 700 msec. per (ball) release.

   There were, also several other slightly less stellar performing types, some using springs, but the Mechanical Computer invention required a parallel 'word' output, of 4 bits of clock actuators, for a 4 and 8 bit typical machine.
   The 'regulator' insert, ( not shown) consists of two sets of levers, upper and lower, and 'primed' by having a ball loaded in initially, for regulating solely based in fall time, between the special levers... Each ball hitting the lower lever will cause a release, thus falling out and down, to actuate yet another release.
The cylinder release lever is then acted on, causing the cylinder to rotate and further release ( one ball).

[RJSV]

This so-called 'BALL CLOCK' uses the (gravitational) fall timed lever actuations. In the diagram, item C. indicates a receiver / paddle to be impacted. Then, the purpose is to release, temporarily, the upper wheel / cylinder, for a quick rotation of approx. 45 degrees
  See letter B. showing a tab or 'pawl' release, caused by the impacting ball. This clock must first be primed by hand, and started up by dropping a first ball by hand.
   Now, the components shown on left side will be able to operate, as a closed loop, each ball drop creating the force to actuate the next (drop action) by allowing the cylinder to move, a bit.
   The mechanism in the right side provides further refinement, by timing the ball release wheel. That is, the simple two lever portion serves a similar control, keeping each future ball release going, but now solely the source of errors by way of unsteady friction sources. Certainly, this eliminates the wheel associated (variable) frictions.
   I always found the problems had solutions that are interesting. This project has mechanical aspects, being'solved' by a Software Engineer, wonder where things could go, with proper Mechanical expertise (!).
  A fun and interesting project, definitely.

[RJSV]

People sometimes ask: But what good is a Mechanical Teaching Computer ?
   And, "What does the 'CLOCK' do, anyway ?

   See diagram, outline of a 4 element timing sequencer.
   Each pivoting wheel has an upwards directed VANE with a purpose of directing each newly arriving ball.
The diagram shows positioning, after a first ball had overturned the pivoting wheel (i.e. 'toggle'). Next arriving ball will simply pass thru the top wheel vane area, in order to be ' chain cascaded' down to second wheel.
   So each new ball, in turn, will encounter, and overturn a next wheel. That way, as shown, each of the four pivoting wheels can be the 'active' sequencer step output. So, the next arriving ball, at top, will fall thru and out, the second output, of this re-entrant sequencer. See orange line, indicating the path of this second ball output. According to the positions shown, the sequence is at 'Step 2'.
  In detail, each pivoting wheel has an actuating paddle on only one side, for sequencer. In more general terms, the wheel design does include a note symmetric look, having a paddle on each side, for totally conventional, two state binary counting. In that case, the right and left outputs (outlets) take on the typically S-R flip flop signal names, Q and Q-not.

CAN ANYONE TELL ME HOW TO POST MORE THAN
   ONE PICTURE ??

Thanks, RJHAYWARD

[RJSV]

Ball Dispensing Wheel Needs Simple Escapement:

It's a bit ackward, doing mechanical designs, but a good CLOCKMAKER could address the problems, creating a design for start-stop intermittent rotation, with ball dispensing every 45 degrees of rotation.
  In the attached diagram, there is an existing ball, falling, that is soon to impact a trigger lever (see letter A.). Then, as seen, that blue 'fork' pivots, causing the right side pin to be withdrawn. Now, then, the big wheel (shown in green) can rotate some.
  The big wheel is under some torque stress, having several steel balls, in dimples, around the periferey on the right half of wheel (actually a cylinder in full view).
As can be seen, the (blue) fork will insert on its left (see letter B.)
   The purpose is to bring rotation to a halt, that being after 45 rotation, to be blocked by the left side of pivoting fork.

   Now much of this is ackward, I cannot even come up with proper terminology, for moving machine parts. But logically, the diagram serves to show the purpose, as a start-stop dispensing action.
   Even worse, (remember this is a 'software' project (!??), so I must admit: The design can't work this way, as there needs to be reciprocal actions, like in a pendulum clock.  But readers can see the principal's, being low friction release, and re-capture of the dispenser wheel.
   Besides the fundamental flaw, (needing actuation for both movements of fork, not just the one direction shown), other aspects needing design attention involve the small ARC of travel as each pin is moved, out of captive slot of the wheel. (Needs to be a straight path to avoid excessive friction).
Plus, there can be small rollers employed for further avoiding friction.

 I consider some of these functions to be 'implementation details', but necessary.

[MK14]

CAN ANYONE TELL ME HOW TO POST MORE THAN
   ONE PICTURE ??

Please take the following steps:

Go to the reply to post page or similar.

Press '+ Attachments and other options'

Then click on '(more attachments)', until you have enough for each of your pictures.

Then click 'Choose file', and put the pictures in (follow its instructions).

Three tests/Examples:

Also, if you want to include a BIG PICTURE of the image in the post. Make the post, then click the desired image to expand it, and then right click on the desired image, and select "Copy Image Address".

Then paste that link into the "Insert Image" icon, available in the set of icons, when you edit a post.
Repeat for any other images.
Best to ONLY do this when the image(s) are rather important, because they will eat into peoples bandwidth, as they will fully open each time.
I did it here, to show how it is done,and what happens.

[RJSV]

Here is a diagram showing one version of the clock, releasing one ball per second. A second wheel is running in parallel, (see B.).
On left side, a disk has 7 'pins', each ball action will activate this 'pin escapement'. The lever action 'see-saw' will escape past one upright pin, while inserting other end, into the arc of pins around other side of wheel.

Inside each of the ball catch hoods, is a toggle wheel, that acts main controller, for the next coming bit of discrete motion. The toggle moves approx. 40 degrees total rotation.
The ball delivery wheel moves approx. 24 degrees or one forteenth of 360 degrees.
Diagram shows pin wheel has 7 pins around.

[ebastler]

This guy has marble clocking and control figured out nicely. 
He has many more videos, also more technical ones, on the design and build of rev. 2 of his marble machine.

[RJSV]

 I will post another diagram soon, but closer examination of timing shows a "new ball' just enters the overturn toggle wheel, at near the end, of a nominal-estimated 1 second cycle timing.
Since that ball would be intended for causing a next cycle repeat, and with a typical overturn delay of 800 msec, obviously 'this' ball, in the intended line up of each ball, at the "zero phase time', or 0 msec within the 1000 msec clock time. For start-up of the mechanical clock, cleverly manipulated activation of the gate escapement can get this extra ball placed properly in the line-up of balls for dispensing (at or near each zero phase time.)

  In rough terms, the phase zero time is the exact time of application of force to move (withdraw) one blocking tab (with friction against pin). Meanwhile, the other end, of the toggling escapement gate, is inserted, without friction contact. Each rotation of the ball retaining wheel is about 4 minutes on conventional clock face, being one forteenth rotation or about 24 degrees.
So anyway, the gate tab pull away takes typically 100 msec, after which some acceleration happens, building up speed slowly at first, due to static friction.
All in all though, the main ball dispenser wheel is expected to rotate until striking against the newly positioned toggle gate tab. This occurs at phase time of 800(msec), or that phase could also be viewed as 'minus 200 msec', relative to new clock time interval.

Notice, this scheme will still work, if lacking the formal pipelined effect, of a clock every 1.0 seconds. When run without providing the extra ball, clock cycles closer to 1.8 seconds, per dispensed ball.

[RJSV]

Will post diagram soon, thanks

[RJSV]

Diagram attached, thanks

[ebastler]

Nice project! Do you plan to try these ideas out in practice at some point? I think that might add another level of challenges. Machines, especially those with freely moving parts not attached to anything (like the balls), tend to have their own will... Do you have a little mechanical or wood-working workshop? Or maybe a makerspace nearby, where you could use a 3D printer or CNC mill?

Several years ago I dabbled with the idea of building a Turing machine using purely mechanical elements. Never got around to actually trying to make something though. And, of course, it has been done in the meantime -- much better than I probably could have built it:
https://hackaday.com/2018/03/08/mechanical-wooden-turing-machine/

[RJSV]

Here is my short reply: (thank you for questions).

Got sick, career was peaking (Lyme Disease).
So eventually lost a string of job positions. Now that's all just personal.
  Eventually, I 'picked' the Mechanical Computer Teaching system, as a novel territory for business / patent actions... No competition, since Babbage died.
(Respectfully LOL).

   I got the major patent filed, a jumbo. Kept getting problems with city, homeless and METH creeps liked to 'interfere' by stealing computer (patent work on it!).
Plus, for whatever reason, broke my office secure doors etc.
  But, again, trying to keep personal stuff out of this thread.
The clock stuff, that is a whole aside to the complete system, several US Patents potentially.
Right now, having CAMERA AND FILE attachment problems, think that's partially solved.

ebastier: Thanks for questions, gonna take more time later, to get that (CLEARLY!) put down here for you.
(Many parts of patent application are in 'abandoned' status.)

[RJSV]

The diagram includes a note, arrow indicates location of one of the 'pipelined' balls, hidden inside the overturn toggle wheel.

[RJSV]

CLOCK USES PIPELINED TRIGGER SIGNALS

  Observe in enclosed diagram, many parts are dim, for clearly showing the line-up of balls (I like to call them 'signals').
There is one location, in a line of balls each at 'zero' time phase, within the one second timing cycle, where usual actions need a supplement, at start-up.
That is, the clock mechanism will function, closed loop as each new released ball makes its way to overturn toggle and cause one more 'tick'. However, that action will be late, by almost another clock.

   It was difficult to understand at first: in order for fast clocking there needs to be placed one 'surrogate' clocking ball, basically taking on the place and function, as if the serial order (of mechanism actions) did not have the delay exceeding one clock.  This is a difference of the terms 'throughput delay' vs rate of delivery.
   Interestingly, once that initial seeded extra ball is placed, in the line-up, the Clock Module assembly itself will provide / perpetuate for continued clocking and delivery. Care should be taken, for insert that single extra ball, exactly in middle, timing wise, of the ball before and the ball after. (That way each delivered clock ball is in phase, each separated by 1 second.)

Also helps to note, a dimple at position '96 degrees', relative to gravity wheel, Generally, a ball will drop / roll out of the dispensing wlayheel somewhere close to 90 degrees, although various unwanted effects can cause irregular timing delay on that one process within the escapement.

[RJSV]

Sometimes, you invent, often there are novel aspects that often are 'suggested' by the natural progression of a design. (What's required there is Observation OBSERVATION).
So, the ball clock phasing and pipeline came to resemble and suggest similarities with audio / musical Sampling and Looping. The free-fall area functions like a 'phase memory' (what is THAT ?).
PLS see diagram: when that second ball is ' seeded' by inserting at phase= 700 msec that 'time phase' is preserved during ongoing clocking. Probably drifts some also.

[RJSV]

Further info, re: Question by ebastier
Yes, Hayward has a Makerspace in a new public library, that's nearby, as well as TAP PLASTICS.
  As to your question about machine glitches, one problem involved a small rotating drum, for dispensing a ball with intermittent start-stop-start-stop action. Every few minutes, one ball (1/4 inch or 6mm) would be sticking up too much and collide (with some structural brace).  Interestingly, many similar 'parasitic' motions / actions have a vigor or 'energy' seemingly an order of magnitude exceeding any legitimate kinetic actions. Meaning that a collision fault could send the main clock wheel careening BACKWARDS...(no good).
  Many failure or partial failures resulted from the simple things, like a funnel for mixing signals that could cause horrendously long 'drain spout spirals', taking predictable timing as a sought after character.

   As to any business plans, doubtfull unless I can attract some partners (that'd be the schools stepping up). Some local political leaders were queried, for ' ' EDUCATION ' '... Only response, was some (political) mailers, asking for money...
   Speaking of money, I DID GET EXTORTION perpetrated there against me and downtown office:
Some guy wanted 5 dollars, to keep quiet about what I was doing ((LOL !)). I'm proud of that one: STEVE JOBS likely never experienced that, crack-head threat... 5 dollars ? (Better ask for more money).
   As you can observe, there was too much street drama, but every start-up has their story.

   I had plans, on Sept. 10, 2001, for mailing that jumbo patent application (BASTIAN 2002) but that got delayed, which was, maybe, a good delay.

[RJSV]

CLOSE-UP DETAIL ON START-STOP ESCAPEMENT

Diagram shows the 7 pin gating wheel, each pin rests against the Toggle Tab (on the green toggle lever.)
Now, appearance suggests a 1/7 fractional motion, of a whole circle: However, this disk will move 1/14 of a whole circle, approx.
   The toggle lever is pulled away, on the top end, and that is considered timing phase 0, taking approx. 100 msec to move clear. That allows the main dispensing wheel to start moving, eventually accelerating, until motion is stopped, at time phase 800 msec (relative to established start, with pin at top center.
   Meanwhile, as the main wheel is accelerating and moving (towards the bottom-center) the toggle lever is being inserted bottom-wise, and having no contact friction. This is observed as that's a blank area on the pin wheel disk.
The toggle lever stops there, (See letter B showing bottom of wheel action. Letter B. to the left of letter is lines, showing amounts of angular movement as dispensing wheel accelerates.
   Half-way in terms of TIME is approx 1/4 of movement; Three fourths of Time phase duration has another fourth (that's about 6 degrees each, for 12 degrees, of the total 24 degrees.)
The rest of the distance is covered the fastest, with ultimate collision at TIME= 800 msec. Then to wait out the remainder of clock cycle time, to Time=0.
Then, the action timing repeats, but this time pertaining to the bottom area. Thus a full repeat is taking two separate cycle times, alternating top and bottom gating areas.
   Since the CLOCK wheel is moving by 1/14th rev but does that twice, once for top gating, and once for gating on the bottom (pins), that's where the 1/7 rotation per 2 clocks comes from.

By the way: A useful reference book on clock escapements is:
   "Practical Clock Escapements",
  By Laurie Penman circa. 1998