This week, we’re taking the wayback machine to 1940 for an informative, fast-paced look at the teleprinter. At the telegram office’s counter, [Mary] recites her well-wishes to the clerk. He fills out a form, stuffs it into a small canister, and sends it whooshing through a tube down to the instrument room. Here, an operator types up the telegram on a fascinating electro-mechanical device known as a teleprinter, and [Mary]’s congratulatory offering is transmitted over wires to her friend’s local telegraph office hundreds of miles away.
We see that the teleprinter is a transceiver that mechanically converts the operator’s key presses into a 5-digit binary code. For example, ‘y’ = 10101. This code is then transmitted as electrical pulses to teleprinters at distant offices, where they are translated back into alphanumerical data. This film does a fantastic job of explaining the methods by which all of this occurs and does so with an abstracted, color-coded model of the teleprinter’s innards.
The conversion from operator input to binary output is explained first, followed by the mechanical translation back to text on the receiving end. Here, it is typed out on a skinny paper tape by the type wheel shown above. Telegraphists in the receiving offices of this era cut and pasted the tape on a blank telegram in the form of meaningful prose. Finally, it is delivered to its intended recipient by a cheeky lad on a motorbike.
http://www.youtube.com/watch?v=HcMHam54EOI
Thanks for the tip, [Mr Name Required]!
Retrotechtacular is a weekly column featuring hacks, technology, and kitsch from ages of yore. Help keep it fresh by sending in your ideas for future installments.
wow, now that’s retro!
I wonder how was the time synchronization performed. I mean, we have the MOSI line explained, where is the SCLK? :)
AC line maybe?
My experience is only with Teletype machines, which I have repaired/kept running…they also (well, came up with) Baudot code and do the whole 45.45 baud tty thing. They synchronize/align decode with the first character received and reset after each clutch using mechanical shift registers. The decode timing is based critically upon the gear ratios of a couple of gears and the use of a fairly large shaded pole synchronous motor — line frequency was indeed the method used. For DC based teletypes, they used swing-weight governors to hold motor speed constant (once it reached speed, the governor would turn the power off to the motor, at which point it would slow down infinitesimally and turn back on again). Large flywheels on DC machines made this remarkably effective.
There is an older (early 90s, I think?) BBC show called the “secret life of machines”. Aside from being generally awesome and humble, they did a great episode on the Fax machine, and synchronizing the two machines was a huge design problem in the early days.
Fax machines, like most long-distance serial communications use asynchronous communications, so all you really need are clocks at either that are of the same frequency and a synchronizing pulse (aka, the “start” bit). Once you have that, the transmitter can simply assume that the receiver will stay in sync, especially over the short time required to send a single character.
SLOM was awesome, Tim Hunkin’s website is well worth a look, for a while you could download the entire series for free there too.
Shame they couldn’t make that show now, every answer would be “It’s got a computer in it”.
They DID say it was a simplified model they showed. In actual machines, there was an additional cam that encoded to “on” for all characters. The camshaft that is shown that combines the bits does not turn continuously – it is rotated through a slip clutch, and is prevented from turning until a key is pressed, which allows it to turn one revolution, sending one character. The receiving printer has a similar slip clutch, and receipt of the first “start” bit enables IT to turn one revolution.
That’s what the START bit is for — at the beginning of each asynchronous character, is a start bit.
When there is nothing being transmittted, there is current flowing through the wire…that keeps a lever attracted to a magnet. The lever keeps a clutch opened, so a cam on the main shaft doesn’t rotate.
The start bit drops the current, disengaging the magnet, releasing the lever, which causes the cluch to engage, allowing the cam to rotate with the shaft.
As the cam rotates, it samples the lever (which is being pulled in and out by the current in the magnet, corresponding to the bits in the character) and sets up the receiver bars for each bit in the received character.
The rotation speed of the main shaft is set to match the duration of the serial character. So, one rotation matches the 8 data bits, plus one start bit (at the beginning) and one or two stop bits (at the end, which allow some “slop” time to get things set up before the start bit for the next character comes in)
How do they keep in sync ?
Line frequency and a synchronous AC motor. Same way as a clock.
Nope. They resynchronize using a relay that engages a clutch at each start bit.
Both are true. Character sync is via a START bit, bit sync is by synchronous AC motors (and gearing, which accounted for 50/60 Hz in international use)
I wondered the same thing so I had to look it up on Wikipedia. Looks like it used start and stop bits. I guess what was common for western union was 1 start bit, 5 bit letter or code, and 1.42 stop bit. The stop bit was commonly replaced with 1.5 or 2 instead of the 1.42 in more modern equipment. Hmmm, sounds like common serial configuration. Interesting stuff! Great Retrotechtacular!
It IS the common serial configuration. Serial ports are a direct descendant of teleprinters. Sometime in the 1960s, the 5-bit code was superseded by an 8-bit code that included a parity check bit. The 8-bit code enabled lower-case, upper-case, and other symbols to be included without requiring shift codes to select between groups of characters. But aside from that and bit rates, the protocol hasn’t changed since this film was made.
Before the 8 bits code was the ASCII code which is 7 bits. 8 bits is a further extension to ASCII.
Which started as people stuffing more data into 7bit+1parity systems.
Teleprinter code cards: http://www.rtty.com/CODECARD/codecrd1.htm
How do you get a fractional bit? Does it have a shorter duty cycle than the other bits or something?
Don’t look at it as pulse width modulation, it is not. Each bit is generated by a machanical cam. It is easy to set the length of each bit by varying length of the ditch in each cam.
And they are a subset of Unicode
Al well and good, upto the point where a motorbike kills your ping-time…
That “Last Mile” is still a problem. :D
Still, a motorbike has a high data capacity. The Great Old Ones used to say “Never underestimate the bandwidth of a 747 full of DECtape”. Our best archaeologists are currently trying to work out what this could mean.
Quote: “Hmmm, sounds like common serial configuration”
lol that’s where serial started these things were originally 300 BAUD rate. Note: BAUD is not equal to Bits per Second. especially when you can have 1 and a half stop bits in BAUD and 5 bits per character.
Synchronisation: This is why there were weird things like stop and start bits. Unlike modern serial where MegaBytes are sent as a single serial stream, these units ran at the speed of the typist. They type a letter, it’s sent via serial and then the serial line goes dead waiting for the typist to type the next character. All this stopping and starting was the reason stop and start bits were required for synchronisation.
Also, because these Teletype machines had to work over a single pair of wires there couldn’t be other handshake lines like RTS, CTS, DTR, DSR.
Timing was controlled by relays. For things like stop bits a relay had a copper core so that eddy currents would cause it to release slower than other relays and hence the one and a half stop bits.
More interestingly, this era was the start of digital. Digital was first implemented in this retro electro-mechanical era using relays. Often they used grey code binary http://en.wikipedia.org/wiki/Gray_code so that relay timings weren’t so critical.
For those who recognise that this is a lot like serial, you are correct. This was the very beginning of serial transfer. Later, stop and start bits were made the same length as other bits. Then additional handshaking lines were introduced (DTR, DSR, RTS, CTS) and COMM’s became more or less standard at 8 data bits one stop bit and not start bits (8, n, 1). Baud rates expanded to 600, 1200, 2400, 9600. 9600 was the introduction of Quadrature Amplitude Modulation. Along the way, the AT command set was implemented and is still used today in modern interfaced like GPRS modems on the GSM network.
I think 5-bit teleprinter was more like 75 baud. If you watch the woman typing, she’s going as fast as the teleprinter will allow. Teleprinter keyboards had lockouts so you couldn’t press a key until the previous character was finished sending, so skilled operators had very consistent typing rates.
Model 33 Teletypes (8 bits, but only 6 were used) were 110 baud — 11 bits, Start, 8 data and 2 stop.
The older 5 level machines were 45 to 75 baud I believe, at least three different speeds.
And, older ones were measured in words per minute. (using a standard 5 character word)
I had to do some research to see why people where even still sending telegrams in the 1940’s. I knew that the telephone had been around for a number of decades, and according to various sources, the US install base was over 15Million by WWII. I can only deduce that the cost of long distance phone calls coupled with the long delay for mail is what what gave the telegram market continued life, at least until the 1980’s when long distance calling became relatively cheap, at that point I believe the telegram market began to drop steeply.
Anyone else have some insight?
The phone wasn’t common place. They existed in hotels, motels and some pubs but not homes (except for the very rich) so the post office / telephone company was the main communication hub. As you see in the video, the last leg of the journey was the motorcycle.
Phones were also less time efficient. The call was manually routed via successive exchange (switch board) operators for each leg of the journey. Teletype machines had dedicated lines for high traffic routes.
Shortly after the era of this video Teletype machines were modified so that the code could be transferred to paper tape. Then the the routing became even more efficient because you could have defecated lines along high traffic segments rather then end to end. In transition points the tape was received on one machine and then sent from another machine in the same office to it’s final destination or a closer point.
You refer to punched paper tapes that could be read by the relaying TTY. When I started programming in the 70’s those TTY with punched paper tape were used as mainframes and minicomputers terminals. Programs where kept on those tapes and reloaded using the TTY tape reader or a standalone reader.
Yes, Then computers went to punched cards (fanform) and then to magnetic tape. I see on pictures of old main frames that they used a Teletype machine for input and a printer. I would have made sense to utilise existing telecommunications technology that was already in production.
The first computer I programmed was manual card feed. It had 8 nixie tubes and I remember that you could use the nixie tubes as extra bytes when needed. It only had four of five buttons on it. I wish I could remember what it was.
Was it a Wang calculator?
Wow, I just saw my ‘spellcheck’ error. My sole for an edit button!
Oh! I was wondering what you meant by that!
I figured thats what happens when too many birds are on the wires.
Excellent typo. Some lines are still defecated.
OH! My sole is burnt!
You place a lot of emphasis on the bottom of your foot.
Amazon. Only instead of words on paper, it will be merchandise selected by a robot and dispatched by drone from a warehouse.
Haha, BT Heritage, I bet they still have some of that stuff in their network – it was still around in 2000’s when I was there, no idea who was using it (if anyone).
The Telex network stuck round for over a decade longer than you’d think in the UK due to there being some legal precedent that a Telex message had similar legal standing to a paper letter, whereas a phone call or e-mail (or fax, presumably) did not.
Here ya go — Bulletin 310B in all its glory:
http://www.soemtron.org/teletypemanuals.html
Everything you ever wanted to know about the good old Model 33.
The office in the film is my town (Bournemouth), and I know the road in the address. I know that’s not very meaningful, but there you go. Anyway, great to see such old technology, and the beginnings of the world we live in, and also to see the differences in such information films made – the pacing and detail isn’t something you’d see today.
I’m assuming that there was no switching for the send/receive stations. That they were dedicated to each other, and the one one typist only was sending to one station or she had several machines depending on where the message went?
Trivial detail: Teletypes are current loop devices, and half-duplex by default. This is because they are directly descended from , and replaced, wireline Morse systems. With half duplex, the printer prints what you send and what the remote station(s) send,
just like on a Morse loop, except without the (relatively) slow, error prone and highly paid Morse operators. (Now you see why Teletypes replaced Morse code for telegrams…it was accuracy and elimination of skilled Morse operators)
There can be several transmitters and receivers on a single loop (and not necesarily the same number of each), and current normally flows (20 or 60 mA) in the loop, keeping the selector magnets on all the printers pulled in. When loop current is interrupted by the start bit of a transmitted character (from any transmitter on the loop), all the selector magnets in the loop drop out and all printers print the transmitted character.
Anyone old enough to remember “Final Score” on the BBC’s “Grandstand” on Saturday afternoon? Live video of the football scores being printed on the “Teleprinter”. That really was a hack. Was it the same machine?