Printer responds to EDSAC code
Post this page to popular social media
Bill Purvis, with the help of John Pether at TNMOC who acquired a spare Creed teleprinter, has been working on the printer for the EDSAC Replica Project. The original printer for EDSAC was a modified Creed, model 7. The one being used in the replica Project is a slightly later model (7E). (If definitive information about the original can be discovered, a copy can probably be found, and the modifications simply transferred across.)
Bill Purvis describes what he did:
"To start with I wanted to determine that the printer would work, and to do this I had to provide power to operate the receiving magnet. I built a small interface card which could be used either to drive the printer from the keyboard or from my Raspberry Pi, which has a PiFace add-on unit. Before powering on the printer I stripped it down, did a little cleaning, and then re-assembled it. This was to establish sensible procedures and for familiarisation. When I applied power, nothing happened apart from a slight hum. It turned out that a resistor had been removed from a spark filter across the centrifugal governor which left the motor effectively disconnected! A trip to Maplins found a suitable replacement, and then the motor ran.
"After a certain amount of debugging my interface, I was able to type characters on the keyboard and to my surprise the characters appeared on the paper! I had expected the ink ribbon to have dried out and need replacing or revitalising, but it produced quite decent characters.
"Next I connected the interface to the Raspberry Pi and with a bit of home-grown coding managed to send characters from the RPi keyboard to the printer. At this point the character coding on the printer was International Alphabet No 2 which is not what was used on EDSAC. Fortunately, the control characters (CR,LF,LS,FS,SP) all use the same coding, so it was just a case of re-arranging the type bars in the type head. This proved easier than I had feared, the only tricky bit was replacing the three springs which return the type bars to the normal position after they have been struck. Once this was done, the printer now prints using the same character coding as we need. One slight difference: 4 characters from the EDSAC set are missing, but these ($,",#,;) are rarely used and should not upset any of the existing programs.
"At this point I made the short video which shows the printer printing a line of text typed in on the RPi. The next step, and which I am currently working on is the modifications visible in the photo mentioned. This consists of a superstructure with 5 solenoids and levers. The 'normal' Creed operates using a single bit-serial connection, with the five bits per character being sent as positive and negative pulses to the receiving magnet. An Additional start pulse and a final stop pulse complete the sequence. For EDSAC, it was decided that a parallel interface would be more convenient. The five solenoids are activated by the 5 bits of each character. These link down into the selection mechanism, lifting the selected levers, while a single start pulse is sent to the original magnet.
"Unfortunately, the only photo showing this modification doesn't give us the full picture, so I am re-inventing the linkages down into the machine. The main problem seems to be getting all five levers
close enough to operate the selection levers which are very close together. I have made several rough sketches and made a start by constructing a base plate and supports. The 4 photos (see Gallery top right) show:
Creed1: the printer with three fixing screws removed (marked in red)
Creed2: the printer with the support pillars screwed into the holes
Creed3: the base plate mounted on the support pillars
Creed4: a side view which approximates to the view seen in the photo mentioned above.
"The Base Plate (see photo in gallery top right)) is a piece of 3/32" thick steel plate cut to size. Three support pillars have been made from 1/4" A/F Hex Steel, turned down and threaded 4BA to screw into the three fixing holes, and with 4BA tapped holes to take countersunk screws. Drawings of these parts are in BasePlate.pdf and StandOffs.pdf (see below). That for the Base Plate will develop as I start to add things to it.
"The original photo suggests that the base plate and support frames were made from a single casting, I felt it more expedient to make mine from separate pieces, screwed together. If this is felt to be too divergent from the original, we can get a casting done later, when I have been able to get a working model. I suspect there may well be a number of trials before I get something which will serve our needs and making a number of castings would be wasteful of time and material.
"I have yet to do any detailed work on the electronics, but Chris Burton has identified the relevant panels in the photo: EDSAC plate 13. Four panels, just right of centre, three of which contain a number of EF54 and EA50 valves, but also a couple of KT61 valves. These are believed to be the Magnet drivers. The logic that I put together for the Elsie simulation should map reasonably well onto this configuration and I expect to progress this fairly soon."