New typewriter mechanism and continued physical construction


I happened upon an advertisement in Craigslist for a local seller who listed a Selectric typewriter plus extra "wheels". The picture, however, showed the rotary knob on the right side of the keyboard that IBM used with the Memory Typewriter series.

After checking various sources, I could see that it was most likely a Memory 50 typewriter. Those used the selectric II mechanism with only a few changes. Most of the changes are to make it more heavy-duty and able to handle long sustain printing runs. This model added solenoids and feedback switches like the console printer typewriter used with the 1130 and 360 computers.

Craigslist ad for "selectric", but could see it was a memory model
It stored documents on a long wide loop of magnetic tape - think of an eight track tape but wide enough to be a 100 track unit. The operator turns the rotary dial to select the location where a document will be stored or retrieved. The 1130 does not use anything like this, but it might be a useful mechanism for some other purpose. As well, it is a possible point for interfacing to the typewriter, if I can decipher the protocol and coding used on the tape.I could inject text that seemed to be a stored document coming off the tape, but is instead something I send from the 1130 adapter.

However, this will be too cumbersome. The typewriter retrieves the contents of a tape track when the RECALL button is pressed, moving the data into an internal 4000 character electronic memory where it will be typed when the AUTO button is pressed. If I were to use the tape mechanism to inject text, I would need to bracket the data with pushes of RECALL and AUTO. This seems very cumbersome to transfer a single shift or printed character at a time.

This typewriter is normal width, so that it can be built into the 1130 replica at the same dimensions as the IBM built machines. My current mechanism is too wide and throws off the proportions of the machine. It uses a mechanism much closer to the real 1130's printer, although this is based on the Selectric II but fully interchangeable 88 character type balls. If I had an actual typeball for an 1130 printer, it would work on this unit. . I still need to build some adapting logic, converting and dealing with a different location on the typeball for some characters, which might include the need for hidden shifts. Hopefully I can do away with the need for those extra cycles. My current typewriter uses the incompatible Selectric III (96 character) typeballs.

Switch-plate for front of typewriter needs to be redone, currently too wide
I ran it manually, verifying that all the letters typed correctly, it shifted up and down, the carriage return and index (line feed) worked, tabbing appears solid, and it was able to retrieve saved documents from the tape and type them. It is dirty and has solidified grease, so the first step for me is to degrease this and then properly lubricate it. The adjustments all seem right, although I can run more comprehensive diagnostics when it is safely cleaned up.

Selectric Memory 50 typewriter ready for cleaning

Cover of Memory 50 typewriter
I did a first round of cleaning using 99% isopropyl alcohol and some rags, to remove as much gunk as I could from various linkages. I will need to force fresh oil and grease at the proper points in the machinery to complete the dislodging of solidified old lubricant.

The built in diagnostic routine comes from track 51, an extra track on the tape which is normally blocked from access by a setscrew on the rotary knob. It has a prerecorded set of output that verifies functionality as well as accuracy of positioning. I ran it and found some errors in behavior, all of which seem triggered right after the end of a carrier return operation. Hopefully this is simply some residual grease that is impacting a portion of the mechanism, where further cleaning will restore the machine to perfect operation.
Failing "track 51" test, showing failure to latch to left side on carrier return
After more careful examination, I found a small spring detached at one end, that spring is connected to a shoe that mates with the escapement lead screw. I saw a mounting post that looked like it was where the spring attached at its loose end - put it on and reran the tests. Perfect! That seems to have been the problem with the machine, which appears to be operating flawlessly. I still have a bit of cleaning and replacement lubrication to do before considering the restoration complete.

Spring stretched above leadscrew, reattached to post in center of picture

Successful output of "track 51" test, all functions perfect
To clean and lubricate the typewriter, I have to shift it to its 'service' position, tilted up in the air at the front to give access to the underside, for some spots and return it to horizontal for others. Somehow in the back and forth, I must have let the mechanism come down skewed to the side; I heard a crunch from underneath as I lowered the unit.  There were shards of a brown plastic cover which had been broken off its location covering a pair of fuses on the power supply which is to the right side of the typewriter mechanism.

Typewriter tilted up in its 'service' position
Shattered fuse cover
I cleaned out the shards, looked the machine over and everything seemed ship shape. However, when I turned it on, the carrier would not space, return, type nor would most functions work. Yikes! Here I had the machine working perfectly and only a bit of oiling away from done, and I managed to hose it up.


I spent a few bad hours trying to figure out what was wrong. The information available on the typewriter is paltry at best - a couple of charts showing voltages to check, a partial wiring diagram, and three 'quick check' charts. No adjustment manual, no diagrams of the machinery, making it hard to figure out what is going on. I am hopeful that a selectric enthusiast on the golfball typewriter group will scan some of the hefty manual that he has in his possession, because digital copies are not currently available. He will start with a few pages that I asked for as a first priority. I would go do the scanning except he is located in central Europe.

Fortunately, the static voltage check turned up an anomaly - the feedback signal was at 0 but should be floating up at 8.5V when the typewriter is idle. From the wiring diagram I have, that signal runs through three switches, any one of which can pull the signal down to ground. They are used to block activity while the mechanism is moving - a switch at the index mechanism, one at the shift mechanism, and one monitoring print cycles. I had to locate the three, which was a bit of a challenge by itself since I have no diagrams showing what the mechanisms look like or where these switches are placed.

The switch that was grounded when it should not be was the switch at the shift mechanism. That makes sense, because it is about where the brown cover would have impacted when it was broken. Although the switch looked fine and the wires were solidly in place, when I checked it with a meter it wasn't right.

The switch uses a very thin reed switch - a narrow glass tube with the switch wires inside, a magnet on one of the wires so it is pulled away from its mate when a permanent magnet is nearby. The shift mechanism has magnets that should have been breaking the contact, yet it wasn't working. I pulled off the switch board and noticed that the glass tube was fractured, out of place and holding the switch closed in spite of the magnet.

Powering up the typewriter with the wires disconnected from that switch, I verified that it now worked well. I have a spare reed switch from the Electronic Typewriter model 50 I had been using - it allowed selection of alternate spacing (pitch) using several of the reed switches and a magnet that swings on a lever as the user selects various settings. I removed a good reed switch from the ET50 board and soldered it onto the Memory typewriter board.

Board from ET50 that 'donated' its reed switches
Reed switch out of position on my first try to replace the broken reed switch

Once in place, I tested it and discovered it wasn't working right - assuming it was mispositioned, I unsoldered and moved it, but broken it and its mate when I tried that one. I need to buy a replacement reed switch of the appropriate type. Alternatively, I could buy the board with the switch mounted, if anyone has this spare part sitting around. However, the only source I know that still has stock of old parts for these machines is out.

One of several reed switches I broke during the day

I picked up a burglar alarm reed switch, removed the small glass switch from inside the big plastic housing, and started to solder it onto the board. Unfortunately, during the soldering I managed to break it. Last try with another switch from an alarm sensor and this time I got the position close enough to work with the board held in place as I tightened the screws.

As well, I had a good day lubricating the mechanism. Now that the feedback board with its replacement reed switch is back in service, I can begin to investigate the best way to use this in the 1130 replica.

One of the eagle eyed members on the Golfball Typewriter forum noticed that my tab test lines are not perfectly aligned, indicating a slight problem with the escapement mechanism that should be adjusted. I will get to that soon.

Board I repaired, slightly below and left of center of picture
Since I intend to disassemble it in a way that allows me to put it back together later as an intact working Memory Typewriter, part of my planning will cover how to separate the pieces nondestructively. The other aspect I will plan is how to drive the machinery from the fpga - where to connect, how to interface electrically, and what adapter logic might be needed to convert the 1130's typewriter adapter logic to this machine.

This typewriter has feedback signals very similar to the typewriter used in the 1130, which means that I will simply feed that over to the 1130 logic. The 1130 adapter looks for an interlock signal, generated while carrier return-line feed, tab, or line feed operations are in process. There is no set of switches to detect this on the Memory 50, thus I can't directly generate this. I will need to carefully study its behavior to figure out a way to produce a useful interlock equivalent signal.

To generate the interlock signal for carrier return-linefeed is easy, because every return ends with the carrier activated the overbank switch as it bounces off the left margin to settle at its home position. Interlock will start when the carrier return magnet is activate and will drop when I see the overbank switch cycle on and then off. It will be when the overbank ends that we can consider the carrier to be back at its proper position and ready for new typing commands.

The other interlock is during tab movements, which I will consider started when a tab command is processed and then the feedback signal changes. That means we triggered the movement and the escapement (movement) mechanism is released. We end the tab command when the escapement magnet is de-energize. I will add a bit of buffering delay to these conditions to account for mechanical settling - about 15 ms should be more than enough.

The typewriter was designed with the keyboard logically separate from the printer mechanism - pushing a character produces the tilt and rotate codes that will be issued to the printer, but the 1130 is sending tilt and rotate codes to trigger typing a character. The only complexity is that the typeball and assignments are different - thus I must do a simple translation, using the value from the 1130 to look up the proper codes for this physical mechanism.

The control operations, (space, backspace, tab, carrier return, index, tab set, tab clear, and shift) are commanded by special combinations of tilt and rotate codes so I can translate the 1130's individual control signals to the prpper codes for the memory typewriter.

In essence, I believe I can make the typewriter believe I have the "typewriter" button pressed and am typing away on its keyboard, but the signals it will receive come from 1130 not the disconnected wires of the keyboard. Because the processor board in the typewriter remains in control, it will set and clear tabs when the right input signals are delivered, move the carriage the right amount on a tab, handle spacing and returns, etc.

Keyboard PCB - wires connected on right are where I will hook my interface lines
I listed the characters that are printed on an 1130, recorded the tilt, rotate and case of each, then listed the corresponding values for the Memory 50 mechanism. The outcome showed that there are a handful of hidden shift requirements as well as a few characters that can't be reproduced. The missing characters are vertical bar (or), not sign, < and > which are now mapped to produce 1/2 1/4, cent and $ respectively as these are less likely to be used in an equation thus allow disambiguation of typed output. The hidden shift requirements are for ' ; = # ampersand $ and @ characters.

The codes used for various characters are quite different, the location on the typeball must vary considerably. The IO Selectrics have two types of typeballs, BCD and correspondence, with different mappings and locations around the typeball, the 1130 is a close variant of the BCD typeball but has a few differences.

Producing logic signals for the simulated keyboard input will be easy - taking the 8.5v signal in a resistor divider network and using a power transistor to short around the lower resistor to ground, just as the physical contacts short the lower resistor of that divider on the actual keyboard controller of the typewriter.

The other interface requirement is to detect both logic signals and the activation of various solenoids/magnets. The logic level flips between 8.5V and 0, while the solenoids operate at either 24 or 12V. The fpga input would be damaged by all three of those non-zero voltage levels, requiring some form of protection on the interface.

I will design a board that drives the simulated keyboard signals and senses all the required voltages, channeling those logical states into my main interface board to the fpga. My intent is to allow this to connect without severing any cables or permanently changing the typewriter mechanism. I will fabricate the same type of push-on contacts on my PCB for the wiring that normally goes to the keyboard controller, but will need to evaluate what kind of nondestructive 'sense' wire is needed for each of my targeted signals. Some may attach via push-on contacts but not all of them do. Many magnet leads are wired onto terminator strips which have round pins that push into make contact.

Physically, the large PCB with all the typewriter electronics, which IBM calls a planar, sits across rear of the typewriter but will be replaced to sit in the frame below the typewriter. I don't want to interfere with the cables that run up to plug into the top, so it will need to sit underneath close enough for the cables to reach.

The power supply for the typewriter sits to the right of the mechanism, with a fan blowing air into the inlet of the planar housing. This will also be relocated to below the typewriter housing, sitting next to the planar to still provide cooling.

Power supply just visible, flat on base of TW and to the left of the tape drive - fuses visible where I broke off the cover
The typewriter also has a built-in tape drive with 51 channels that can each store about 4000 characters. I will not use this with the 1130 replica, thus it will be carefully removed and stored away for future restoration of the typewriter.

The keyboard itself is not needed and has to be removed to fit the printer into the 1130 enclosures. I will find a way to separate it cleanly and nondestructively, saving it for a future restoration of the typewriter. The solenoids that activate the tilt, rotate and velocity controls of the typewriter sit underneath the keyboard and extend outward beyond the front of the console printer faceplate location. I believe they are low enough that I can extend them under the covers of the machine, but I will need some careful measurements and experiments before I can confirm this.

Side of mechanism - will detach keyboard (right in pic) and planar (foam at extreme left)
With the planar assembly and keyboard removed, this should easily fit within the 11 1/2" approximate front to back dimension of the typewriter enclosure on the 1130. It appears that I will have about 2" of free space between the front of the typewriter carrier and the black faceplate. That gives me plenty of room for the switches and buttons, but also requires an extension of the 'marker' flag that shows through the clear plastic ruled stripe on the front - to indicate at what character position the carrier is currently located.

The entire mechanism needs about 7 1/2" from floor to top of the enclosure, while I have a budget from the real 1130 of about 8 1/2".  The most serious constraint will be fitting the solenoid bracket underneath the tabletop.

The solenoid assembly partially extending under the keyboard (to the top in this pic)

I wedged the light matrix assemblies into place inside the box using stiff rubber sheet segments, folded over to add springiness. The positioning of the assemblies, the front plastic panel and the two metal side plates were tweaked until it all sat together well.

I wired up the signals and power lines to the LED driver board - they attach to the board in a single 8 position connector but I want the two power lines to run down one pedestal while the signals lines for this board and my switch are protected in their own pedestal channel.

The LED driver board was placed inside the pedestal box behind the light/mask assembly. Carefully folding the long runs of wiring to make it all fit inside the back cover of the box. This was painfully tedious work, as the stiff wires twist and bend the connectors, popping them up or off the LED board with the least adjustment of the wires. After about an hour of very frustrating work, everything was in place.

Pedestal box (upside down) as I complete connection of driver board in rear
Wires run down the left pedestal stand to the light cubes on the main black plate, as these lamps are also driven by the LED driver board. The left pedestal also contained the main power lines for the LED board, to separate all power or driven lines from all signalling lines which are on the other side. The serial protocol lines to the driver board as well as the mode switch cable were routed down the right pedestal stand.

Manipulating wiring to fit inside pedestal box
I didn't want to closeit up until I had done a full test of the lamps, wiring, connections and everything else inside, because the rear plate is held on by quite a few sheet metal screws that make the opening and closing of the plate a slow process. The weekend is over, however, and I will be away most of the week on business, so this will be set aside and wait until I return for those final steps.


Getting the acrylic box to be relatively water-tight was quite a challenge and after too much time spent fiddling around adding layers of hot glue, I jumped to a plan B. I found a small plastic box that I could use to make the first half mold, with a perfectly flat bottom and sides high enough to ensure the model was fully submerged. I attached the model of the inner side of the knob to the bottom, with its epoxied screws and nail that will serve as air holes and the gate for pouring in resin to make knobs in the completed mold. The mold solutions were mixed and poured into the box up to its rim, then left to solidify.

The next day the half-mold was pulled out of the box, the model of the knob inside separated from the base plate while remaining embedded in the mold. Then the'features' (plastic rectangular solids) were pulled off the bottom plate, and that half-mold was inverted. The top of the inverted mold, with the exposed face of the knob inside's model, became the base or floor for pouring the second stage that makes the other half mold.

I glued the two halves of the model together as this was needed to complete the mold fabrication. With the model still fitted into the half-mold and the outer side with dome sticking up above it, I poured the second half of the mold. It formed protrusions matching the features in the lower half mold, so the two halves can be registered in alignment when I am casting.

Having created the two part mold for the typewriter knobs, with the completed mold ready to use by Tuesday afternoon. The first pour of blue resin seemed good, but I will need to wait the 24 hours before I can really determine how well the knobs will turn out.

First try to cast a typewriter knob using new two part mold
When it was removed, I found a few gaps where air bubbles were on the inside face. That might have been acceptable however there was one that affected the rim - thus highly visible as well as affecting the functionality. I will drill a few additional air relief holes then try again.

First try, inside face, air pockets left gaps including at rim

First try outer side, stepped ridges rather than smooth surface

With the new holes and a much more careful pour, pausing frequently to swish the resin around to chase out air bubbles, I am hopeful to have a cast that is usable as a knob. I took the first, failed, version and used it to try out a process to smooth the dome. I alternated the resin spray to melt down sharp edges with sanding, fist with 180 grit dry paper and then with 240 grit. It appears I can get the dome to be much smoother, although the symmetry and surface appearance is unknown until I am done.

Initial smoothing test, seems with some more work can get the dome smoothed
I produced two passable knobs from two days of casting - several small air bubbles on the inside face, with a very minor incursion on the edge much smaller than the void in the original trial version. I know have a sequence of finishing steps to apply: smooth the dome, flatten the inner face, cut a rectangular hole for the platen axle, drill a hole for the setscrew and tap the setscrew hole.

Two castings for the typewriter knobs, before cleanup and finishing


The last of the light cubes are out of the mold and deemed acceptable. Once labeled, I will have a full control panel exactly as found on a 2501 A1/A2 card reader - the type used with the 1130. I already had the READY and POWER ON lights, plus the three switch buttons with the correct labels. I just needed to fabricate the ATTENTION, READ CHECK and FEED CHECK lights. With the labels applied, the panel is now complete.

Control panel from 2501 card reader, ready to install in replica reader

I did some looking, research and thinking about ways that golfball elements might be made, to recreate the balls used on the 1130 and 360 console printers, including the special APL ball. This would be injection molded nylon, with a cap added that had been liberated from a broken IBM or GP made typeball.

Injection molds get pulled apart, the top coming up off the part while the bottom may be pulled down away from the bottom. To allow separation, the part has to slope inward from the midline, to the top and bottom, causing a gap as the mold is pulled away from midline (called the parting line). The slope is called 'draft'.

On a type element, the natural spot for a parting line is in the middle of the characters for one of the four 'bands' of type that correspond to the four tilt positions of the typewriter mechanism. The parting line produces a noticeable separation line, which would ruin the quality of the characters.

Thus, the mold design will require some additional inserts that generate the typeface, with the outside of those inserts having the draft to separate from the mold halves. These type of inserts are generally created as sliding elements in the mold but could be free attached and pushed off the resulting typeball.

That manual prying process would be unacceptable for any production volume casting, but is fine for the very limited volumes I might make of these elements. Further, the inserts can be swapped in the mold to create different character sets and arrangements around the ball. Since the typeball is hollow inside and open on the bottom, I would have clean entry to pry the inserts away from the bottom.

IBM type elements are coated with a metal covering, but it is said that this was a cosmetic decision given customers association of plastic with cheap and poor quality goods back in 1961 when the Selectric was first introduced. I can find a way to make these silver colored if necessary for the esthetics, but will begin with plain white nylon type elements.

The typeball has precise placement of each character for the tilt and rotation of the typeball produced by the typewriter, plus the characters' face is curved to match the curvature of the cylindrical platen that the paper is rolled around. The character will thus strike with equal depth of every part of the character, whether it is in the vertical center where the platen is closest or at the top or bottom of the character where the platen is curved further back.

The most challenging aspect of milling the mold for this is to produce very clean surfaces for the font - instead of the milling marks that would be left by milling a metal surface. As well, type often has sharp angles and edges on the inside which are very difficult to make with round milling tools.

Until I have a process in mind that has a reasonable chance of success for a reasonable level of quality, I won't be moving forward to buying supplies for this. Right now it is just ideas I am playing with.

No comments:

Post a Comment