Watching the film “Tuesday in November”, a classic (propaganda) film about the way American democracy works (in theory), I noticed something near the end.
When the cameras are rolling in Times Square, 1945, there are moving text displays! How was this done? I know you can do it easily digitally by representing the state of the display as a vector for each line. Each time you refresh the display, you send power to the lightbulbs that have “1s” in that binary vector. You then shift the whole vector right or left or adjust the pointer that points to it and you can get motion. The number of shifts per second determines speed.
How did they do it in 1945, though? Was there a gigantic vacuum tube shift register driving the display? Or, was it something more crude like representing the text as a series of holes on paper and then shifting the entire paper strip over, with mechanic switch pins that control power to the individual light bulbs?
I recall the footage of these displays evidencing artifacts in the text. Cant research from here, but does anyone know if the artifacts can be observed to appear on a bit of text, then follow that text along? If so, I think we can assume that shift registers were in the mix. They may have been electromechanical (based on relays) at that point in history.
Think a player piano. That has a moving piece of paper which causes a piano key to strike. Well what if instead of a piano key its hits a relay which controls a light switch.
They used a large roll of paper with holes punched in it in tracks - usually 8 tracks. It was pulled across an array of micro-switches. Where there was a hole the switch actuated and illuminated its assigned light via a relay. So if you had 1000 lights you had 1000 microswitches and 1000 relays.
I’m not quite sure how they handled the transitions. Possibly slow-release relays?
If I was designing such a system now, with the technology then available I would use a hybrid of what has been suggested.
The system needs a relay per light. That is pretty much a given. Once you have that you already have most of what you need to make a set of shift registers. With a clock line and two relays per light, you can get each light relay in line to latch the value of the light before it and transition. Thus clocking the value forward one light. The trick is to load the lights. That can be done with a tape, but instead of needing a parallel read of the entire tape, you only need to read one column at a time, and clock that into the light array. You could either fill the light array exactly, and have it then cycle the message until you loaded another one, or could have a tape with an arbitrarily long message being read cyclically. The relays could be mounted behind the lights themselves, since there is only local connectivity needed. This avoids thousands of control wires.
But this has not much to do with how they might really have been made.
I don’t see why it coudn’t have been done with vacuum tubes. I would have used bistable multivibrators, which consist of two tubes and a couple of capacitors. Driven most likely by paper tape, similar to a player piano.
When the Philadelphia Evening Bulletin moved to their new building at 31st and Market in the early 1950s, they made a similar display.
After writing this, I googled Times Square News display and found a site that included the quote: “…introduction of an electronic news ticker at street-level in 1928.” This suggests that it was electronic rather than operated by relays. But I could find nothing about its mode of operation. But 1928 was awfully early for electronics.
Quote
*
“With a clock line and two relays per light, you can get each light relay in line to latch the value of the light before it and transition”*
In terms of complexity you trade an extra relay for a microswitch (probably more than one relay depending on relay). I’m also guessing relays are more expensive than microswitches.
In the relay model, In terms of reliability the state of a particular light is dependent on many states of upstream relays. Back then this may not have been a safe choice.
A switch per light is much more reliable but would require a bulky reader assembly. #NEWSFLASH
Reading about Newscasters from my now ancient technology encyclopaedia “How Things Work” I see that the standard tape is 13 3/4 inches wide and and has four color tracks. while the news track is 3 1/2 inches wide in white only. The standard display is 15 pixels high.
The reader is a bath of mercury and a head assembly. There is no relay intermediate. Current goes through the mercury to each 40 watt light. However some implementations use photoelectric cells instead of mercury.
The largest newscaster in London has 30,000 bulbs and uses 9kW and 15 lamps per day need to be changed.
Getting back to topic, that’s an awful lot of relays to implement as a shift-register. No-wonder they didn’t do it that way.
Commercial vacuum tubes would have been pretty new in 1928?
The first serious computer IIRC filled a gymnasium and they had grad students with shopping carts replacing vacuum tubes full time. Every calculation was run several times to be sure the answer was not distorted by a tube failure. So I doubt a “electronic” as opposed to “electromechanical” was a good solution in 1928, especially for something meant to show off technology to the public. I’m sure just keeping the light bulbs functional was a full-time job.
Basically, each bulb had to be directly controlled. Whatever technique you use is going to control a wire to a bulb. Modern matrix displays - controlling horizontal and vertical lines to feed the intersection points - would consume far too much computing power for vacuum tube days. A set of modules for each grouping of lights simply meant hundreds of separate locations to be maintained, instead of centralizing everything in one room and only wires running out to the lightbulbs.
Also remember, electromechanical was the technology of the day. IBM punch card machines were the typical business machine, same idea - the holes in the cards allowed switches to make contact, thus triggering counters, adders, etc.
Maybe like a music box, analogous to a rotating drum with projections that strikes vibrating strings.
The simplest form would be those signs with an neon arrow pointing to the entrance, which has three of four phases of animation making the arrow appear to move in a primitive fashion. A rotating device simply makes a mechanical contact with a circuit that turns light bulbs on and off.
That *Wired *link left out a litle useful info… 39,000 brushes to make or break electrical contact as the frame with its letter elements looped by on the conveyor.
Sounds like no tubes, no relays, just a crapload of contacts for a one-to-one addressing of the 14,800 bulbs.
ETA: The *Wired *item and the one I linked from *EDN *are both citing the original item from the NY Times
So, instead of holes in paper or some other binary system, these “letter elements” were literally just the letter they wanted made out of conductive metal. They then scraped these conductive metal letters between 2 brushes that don’t touch each other because they are slightly misaligned, and whenever current is able to travel from one brush to another, that completes the circuit path and the light can light up. No relays at all, the 40 watts or so goes right through the control system, and there’s a big heavy wire to every. single. bulb.
Yikes. And I can’t imagine all the electric arcs created when you complete a circuit and then break it constantly was good for the life of the brushes.
More likely it was a much lower current and voltage running relays at each light bulb. Running 110V through hundreds or thousands of contacts in a gizmo that you have to handle does not sound clever.
[QUOTE=beowulff]
40W at 120V is only 1/3 amp, so you would only need 22 gauge wire (or less) to each lamp.
[/QUOTE]
Looks like they ran 10-watt bulbs. Even at such low power, the brushes did need monthly attention.
It’s not about the Zipper in particular, but an articleabout Bickley Motograph displays indicates they ran 10 watt Mazda* sign bulbs that would last roughly three years.
