Binary mechanical devices
Binary mechanical devices
(OP)
One class of mechanical devices not widely recognized is mechanical, but their essential operation is binary. It might be fun to discuss such mechanisms.
===
The teleprinters (Teletype [tm] in the USA, also Kleinschmidt, Creed in the U.K. (others?)) were, of course, partly binary. Keystrokes were converted mechanically to serial binary data, and also punched, in parallel, into paper tape. The serial data was converted back into parallel and decoded; likewise, paper tape was read in parallel, and its contents decoded.
Flexowriters were similar in function to teleprinters although not often (if ever?) used for telecomms. (I know a fair amount about Flexowriters; was once a technician on them.)
The IBM Selectric typewriter, totally mechanical except for its drive motor, internally used a binary code for every typing key, and contained two mechanical digital-to-analog converters, which used a principle identical to mechanical analog linkage-computer adders/subtractors.
Referring to a multi-animal scheme for pulling a wagon (afaik!) this linkage is sometimes called a "whiffletree".
The output of those converters tilted and rotated the typing element ("golf ball").
---
Encoding keystrokes into binary is trivial, but decoding binary data to select a key is not so elementary. A basic mechanism that (afaik) is never listed in a collection of basic mechanisms, a sensing finger, is the usual way to decode. It's not a new concept; the Jacquard loom used sensing [wires, iirc]!
Essentially, in the decoder, each bit of the parallel input operates a code bar that has projections ("teeth" with flat ends) only in some places. Each decoder output (say, a key) has a pair of projections for it.
Depending upon the code, each decoder output's "teeth" are selectively removed (or omitted, in the first place) from every code bar. If a legal code is set into the code bars, one output will have no teeth blocking the movement of its sensing finger, and the full sensing motion becomes the output.
Other sensing fingers will be blocked, so they move only slightly, to contact at least one tooth, and therefore be blocked by it.
[Another basic mechanism omitted from the usual collections is an interposer. The driver part moves back and forth, driven by a crank or cam, most likely. The driven part is likely to be held in home position by a spring. If the interposer is set for no action, it's out of the way, and the driven member stays where it is.
However, the interposer can be moved (when the driver is retracted) between the driver and the driven members with no force, but when the driver cycles, in this case, the driver pushes the interposer against the driven member. Of course, the interposer is made so it can move along with the driver with no damage.]
The Flexowriter's decoder used sensing fingers; a finger that moved all the way was in position to be "snagged" by a cam-operated moving bar that pulled down the corresponding key.
===
Some paper-tape punches made by Soroban Engineering used interposers to either drive the punch pins or leave them retracted. IIrc, their fastest punches ran at 400 chars/sec, with the data input timed by the punch camshaft. Punching data in bursts, the tape (which exited horizontally) didn't have time to sag; it took a gentle curve downward.
===
Many good pipe organs have a nonvolatile RAM that in earlier times stored binary data mechanically. Called a combination action, it permitted the organist to make major changes of stop selection (and other configuration changes) simply by pushing one button.
These combination actions had very few addresses, typically, but each address had a bit for each stop and other function (such as a coupler).
Writing data was simply done by setting the organ exactly as desired, pushing and holding a [save] button, and then the [address] button. Reading was even easier, just a poke of the [address] button. (I'm using electronics terms, not organ terms, in part!)
===
I know truly almost nothing about it, but apparently the first computer built by the German genius Konrad Zuse was both mechanical and binary. I have seen Web photos of it, but such a mechanism is so unusual that it's really hard to understand without explanations; one can only admire.
===
An extremely-interesting educational toy called the Digi-Comp was a three-bit mechanical J_K flip-flop with (iirc) six-input programmable gates for set and reset. Those six were the true and complement outputs of the flip-flops.
I had one, and remember setting it up as a plain natural binary counter, a circular shift register with a half twist, and other things.
There is some thought currently about nano-sized logic mechanisms.
Any more info/ideas?
===
The teleprinters (Teletype [tm] in the USA, also Kleinschmidt, Creed in the U.K. (others?)) were, of course, partly binary. Keystrokes were converted mechanically to serial binary data, and also punched, in parallel, into paper tape. The serial data was converted back into parallel and decoded; likewise, paper tape was read in parallel, and its contents decoded.
Flexowriters were similar in function to teleprinters although not often (if ever?) used for telecomms. (I know a fair amount about Flexowriters; was once a technician on them.)
The IBM Selectric typewriter, totally mechanical except for its drive motor, internally used a binary code for every typing key, and contained two mechanical digital-to-analog converters, which used a principle identical to mechanical analog linkage-computer adders/subtractors.
Referring to a multi-animal scheme for pulling a wagon (afaik!) this linkage is sometimes called a "whiffletree".
The output of those converters tilted and rotated the typing element ("golf ball").
---
Encoding keystrokes into binary is trivial, but decoding binary data to select a key is not so elementary. A basic mechanism that (afaik) is never listed in a collection of basic mechanisms, a sensing finger, is the usual way to decode. It's not a new concept; the Jacquard loom used sensing [wires, iirc]!
Essentially, in the decoder, each bit of the parallel input operates a code bar that has projections ("teeth" with flat ends) only in some places. Each decoder output (say, a key) has a pair of projections for it.
Depending upon the code, each decoder output's "teeth" are selectively removed (or omitted, in the first place) from every code bar. If a legal code is set into the code bars, one output will have no teeth blocking the movement of its sensing finger, and the full sensing motion becomes the output.
Other sensing fingers will be blocked, so they move only slightly, to contact at least one tooth, and therefore be blocked by it.
[Another basic mechanism omitted from the usual collections is an interposer. The driver part moves back and forth, driven by a crank or cam, most likely. The driven part is likely to be held in home position by a spring. If the interposer is set for no action, it's out of the way, and the driven member stays where it is.
However, the interposer can be moved (when the driver is retracted) between the driver and the driven members with no force, but when the driver cycles, in this case, the driver pushes the interposer against the driven member. Of course, the interposer is made so it can move along with the driver with no damage.]
The Flexowriter's decoder used sensing fingers; a finger that moved all the way was in position to be "snagged" by a cam-operated moving bar that pulled down the corresponding key.
===
Some paper-tape punches made by Soroban Engineering used interposers to either drive the punch pins or leave them retracted. IIrc, their fastest punches ran at 400 chars/sec, with the data input timed by the punch camshaft. Punching data in bursts, the tape (which exited horizontally) didn't have time to sag; it took a gentle curve downward.
===
Many good pipe organs have a nonvolatile RAM that in earlier times stored binary data mechanically. Called a combination action, it permitted the organist to make major changes of stop selection (and other configuration changes) simply by pushing one button.
These combination actions had very few addresses, typically, but each address had a bit for each stop and other function (such as a coupler).
Writing data was simply done by setting the organ exactly as desired, pushing and holding a [save] button, and then the [address] button. Reading was even easier, just a poke of the [address] button. (I'm using electronics terms, not organ terms, in part!)
===
I know truly almost nothing about it, but apparently the first computer built by the German genius Konrad Zuse was both mechanical and binary. I have seen Web photos of it, but such a mechanism is so unusual that it's really hard to understand without explanations; one can only admire.
===
An extremely-interesting educational toy called the Digi-Comp was a three-bit mechanical J_K flip-flop with (iirc) six-input programmable gates for set and reset. Those six were the true and complement outputs of the flip-flops.
I had one, and remember setting it up as a plain natural binary counter, a circular shift register with a half twist, and other things.
There is some thought currently about nano-sized logic mechanisms.
Any more info/ideas?
Nicholas Bodley |*| Retired technician
Eastern Mass.
RE: Binary mechanical devices
I am looking forward to my retirement. As of now, I do not have the time to dig deeper into those fascinating devices that actually were very modern (if they even existed) when I left school.
All can say right now is that my first printer was an old IBM Selectric (I think) with a ball that tilted and rotated. There were some solenoids attached to the linkage and I built the computer interface from a single chip intel microprocessor (the 8748) and some transistors to actuate the solenoids. I do not think that there were more than five bits for characters and then one shift bit. Probably extra bits for CR and LF - I do not remember. The printer was connected to one of the first commercially available PCs - an ABC 80 - in 1979. There were no "personal printers" available at that time. At least not with the neat output that the IBM could produce.
It is a great wonder that these machines could be built and maintained at all. There were (I think) at least a thousand little levers and links and shafts and springs - not counting all screws, nuts and washers - and most of them needed to be just right to get the desired function.
Model trains? Forget about that. The challenges that an old IBM or an analog mechanical computer offers are of quite another order of magnitude. And the beautiful thing is that the internet makes sharing of knowledge and answering questions an easy matter.
RE: Binary mechanical devices
Thanks kindly for your comments. They are interesting reading. As to complexity and reliability: The Marchant Silent Speed calculators were said to have about 7,000 parts, and one version almost double that.
When you have competent design and mfg. engineers and machinists, complexity need not be a major problem. As mechanisms go, the Selectric doesn't seem too bad.
Unfamiliar mechanisms look a lot more complicated than they do after you know them well.
I have a Selectric that's in rather bad shape which I plan to dismantle, and probably not restore; maybe that's a sin. Likewise a big, heavy Remington Rand 10 key printing adder with mult. keys and even a divide key, the latter probably for semi-auto operation.
I'm really enjoying the "feedback" here! Again, thanks!
My regards to all,
NB
Nicholas Bodley |*| Retired technician
Eastern Mass.
RE: Binary mechanical devices
Having such unusual playthings almost certainly steered me into an engineering career, rather than something more lucrative like law. At least engineering is an honest profession!
-----------------------------------
Start each new day with a smile.
Get it over with.
RE: Binary mechanical devices
You cannot argue a mechanical device into working order. You have to find the REAL reason(s) for the problem and then apply the CORRECT remedies. Arguments are for the court-room - facts are for engineers. That's why we are so dull - and so poor...