Siemens T36Si

[Funksammler]

I would like to introduce the restoration of a Siemens T36Si teleprinter.

By 1930 the development of teletype technology had developed into viable commercial business and the US Morkrum Kleinschmidt company (later merged into AT&T) produced the state of the art teletype machines. Lorenz aquired a licence to use their technology exclusively in Germany, forcing Siemens to develop a different design.

The Morkrum Kleinschmidt machines use only one clutch release magnet for the transmitter and receiver mechanisms meaning they are mechanically coupled. The pulses from the transmitter reach the receiver almost immediately but the pulses from a remote machine only reach the receiver after a short delay which increases with the length of the transmission line. If the transmitters and receivers from both machines are electrically connected to the same line there comes a point that the two machines can no longer synchronise due to the delay. To avoid this problem long teleype connections use four wires, two wires for messages from A to B; another pair for message from B to A (it also means that the machine no longer prints its own messages).

In the Siemens machine the transmitter and receiver are mechanically independent, they both have their own clutch release magnets so in theory the Siemens T36 could be used over a single pair of wires over long distances (as the delay between machines no longer poses limits). So in theory a network based on Siemens machines could be built more cheaply, but since the German Telex network had already developed to be compatible with mechanical machines needing four wire connections, this potential commercial benefit could not be exploited.

Another benefit of the Siemens design is that it is essentially a digital electrical machine; it becomes relatively easy to connect it to peripheral equipment like punch tape readers and writers or indeed cypher units as used in the T52. Each T36 machine has two connection sockets to connect the transmitter to a punch tape reader and the printer to a tape puncher. The use of punch tape allows a message to be pre-recorded on tape and be transmitted at full speed afterwards. Since hand typing could never keep up with the speed of the punch tape transmission times (and with it cost) can be reduced.

Whereas the mechanical machines essentially used a mechanical typewriter as their printer, Siemens decided to go back to the rotating wheel printer first developed in the 1840's. All the characters are placed on a wheel which makes one full rotation every cycle. To print the intended character, a striker pushes the paper against the wheel at the exact time that the character passes by during the rotation. To achieve this, the T36's printer has to convert the 5 bit code into a precise timing moment. Since the wheel rotates in less than 140 ms the striker has to be able to hit with milisecond precision. The resulting printer is a lot more compact and probably cheaper to make than the typewriter type used by the competition but the amount of electrical switching contacts involved make it complicated to adjust and maintain.

Key in the printer mechanism are the five polarized relays that store the five-bit code throughout the printing cycle. The relay coils receiver a short pulse of stored energy from the receiver camshaft contacts and flick the switch contact to the T or Z contact where it stays until the next pulse is received. So in theory the polarized relay can remember the bit information forever. However to achieve the sensitivity and correct balance the relays require regular careful adjustment, one of the weaknesses of the machine.

The attached photographs show the overview of the machine, in later posts I will focus on some of the details

IMG_3950.JPG

IMG_3951.JPG

IMG_3952.JPG

IMG_3953.JPG

IMG_3954.JPG

IMG_3955.JPG

IMG_3956.JPG

[Funksammler]

The first problem to solve on the machine was a seized keyboard. The oil and grease in the mechanism had hardened and someone had probably try to force down the keys, causing the coding cams to slightly bend. So after cleaning and oiling they would still not run freely and I had no option but to dismantle the keyboard completely. So all the springs, keys and cams were remove and cleaned, bend back into shape and reassembled. To be true it took about three goes to get the mechanism to operate properly. After that the lugs that engage with the electrical contacts had to be re-aligned as well within very tight tolerances.

IMG_3847.JPG

IMG_3850.JPG

IMG_3851.JPG

IMG_3852.JPG

IMG_3854.JPG

IMG_3855.JPG

With the keyboard removed, the net filter and resistors fitted to the base become visible. Hardly any of the components or contact points are numbered, making fault finding extremely difficult; very uncharacteristic of German equipment of the period...

IMG_3859.JPG

IMG_3860.JPG

[Funksammler]

Now that the keyboard was working again, the transmitter circuit could be tackled. The keyboard cams engage with six contacts in the code generator. The first contact (which is moved with every keystroke) is the starting contact. The other five cams are swithed to -60 or +60V dependent on the keystroke to generate the 5-bit Baudot code. The starting contact closes the AR relay which in turn releases the clutch of the transmitter camshaft which start to rotate. The transmitter camshaft connects the -60V and the five coding contacts to the output one after the other to generate the transmitted code. It also has a mechanical feedback to the code generator so that the start contact is released (and with it the AR relay) so that each keystroke is only processed once. Once the AR relay is released, the clutch coil is released and the transmitter camshaft is again held stationairy until the next keystroke.

IMG_3960.JPG

IMG_3961.JPG

IMG_3865.JPG

IMG_3962.JPG

IMG_3963.JPG

I had a bit of a problem with reconnecting the keyboard wiring, even though the terminal board is marked, none of the wires are and I spend a day scratching my head when I had accidentally reversed the starter and the 5th bit contacts...

IMG_3861.JPG

The coding and transmitter circuits contains about 20 switching contacts which all have to be cleaned and adjusted. By using the Y (10101) and the R (01010) each bit can be monitored on the oscilloscope and the contacts adjusted until the signal is as clean as possible. It proved inpossible to completely eliminated the contact bounce in the code generator, which affects the first bit only (the green trace shows the second bit, so the contact bounce has disappeared by the time it is sampled by the transmitter cam). You can also see that the signal is a bit noisy, which only gets worse once the receiver and printer circuits are running at the same time...

IMG_3878.JPG

IMG_3879.JPG

[Funksammler]

Now that we have a usable signal coming from the transmitter, we can feed it back into the machine to test the receiver. The receiver has to inverse the actions of the transmitter, it has to convert the serial code into five separate bits. It uses another rotating camshaft to operate a series of contacts. The negative start pulse is used to release the clutch of the receiver camshaft. The next five contacts each take a roughly 2,5 ms sample during each pulse and store the sampled voltage in a capacitor. At the end of the cycle the five voltages stored in the capacitors are fed to five polarized relays. These relays are the 5-bit memory of the machine which is used by the printing circuit as an input. So the T36Si is a real digital machine with a 5-bit memory and databus between the receiver and the printer.

So the restoration again had to deal with multitude of contacts. The five capacitors used for storate need to be checked, fortunately they are of a vacuum sealed type and they still proved to be within tolerance after 80+ years! Not so for some of the lower value capacitors used to store the energy to drive the clutch release coils, these had to be replaced with new ones. The first picture shows the capactor bank in place while the second shows the removed capacitor bank. The new orange capacitors are small enough to be fitted between the on top of the old ones.

IMG_3894.JPG

IMG_3892.JPG

The energy stored in the capacitors is enough to switch the polarized telegraph relays to one side or the other. While this worked relatively well in the fist machine I tackled I encountered a curious issue in the second machine, in about half of the relays the magnetic polarity of the armature had reversed, meaning that the relay switched in the wrong direction. I guess after years of non-use the permanent magnetism in the armature metal has faded. Fortunately I could find a set of five relays that were working in the correct direction. It did though cause a lot of headscratching to find out why half of my memory was not working correctly...

Another issue to was the aforementioned contact bounce in the coding mechanism as it makes the first bit sampling unreliable (this contact bounce only affects the first bit). In the original wartime setup, the received signal is fed through another polarized telegraph relay in the "Fernschaltgerät" (which I have unfortunately not found yet) which would probably have filtered most out this problem. Since I do not use this extra relay, I had to make sure that the coding mechanism was optimally adjusted and a placed a capacitor in the signal line to filter the worst of the bounce.

Needless to say the individual telegraph relays needed cleaning and adjusting as well.

IMG_3864.JPG

IMG_3863.JPG