More physical construction milestones - week ending July 28, 2013


I used the water jet to cut all the notches, tabs and other features on the sheet metal part that will hold up the typewriter. Super fast and accurate, although it is still a very complex part with challenging bends to perform.

As it sits now, the left edge is bent out of position a bit, but I believe I can pull it into line with an interior strut that I will fabricate and install. Similarly, I can brace the bottom plate as it must be able to support the typewriter (fairly heavy) and its enclosure. I still have some verification to do with the typewriter mechanism and the expected enclosure, to be sure I have the height and setback right.

I have screwed the sheet metal parts onto the frame and have been twisting, bending, wrestling and fine tuning the positioning. Once they are in the right orientation, I have to figure out the details of how I will fill any gaps before I powder coat everything.

Any changes or additions I make must still allow the pieces to be dismounted, go in for coating, then re-installed. That rules out welding all the seams in advance, for example. Likely I will have visible seams, but held tight together, where IBM had soldered or welded and filled all the meeting points to create the metalwork.

I worked out a way to weld most of the seams on the upper sheet metal and still get the pieces back onto the frame after they are welded and powder coated. I drew up plans for a few small metal plates to add and identified all the seams that get welded. I practiced my welding technique with sheet metal - something very challenging where it is very easy to melt holes through the metal due to the inability of the thin sheet metal to conduct away heat fast enough. I think the seams will work - they won't be the most beautiful welds, but with a bit of grinding down they will probably be good enough.


I needed to make a mold of a button, so that I could cast buttons in the colors I need and having the added advantage of untitled blank faces more suitable for rubbing decal text onto. I used the TAP Plastic Platinum Silicone mold system to make a "glove" mold of an IBM made button. The buttons can be made from this by pouring in Urethane casting resin, colored with their opaque dyes mixed to the suitable color. For similar reasons, I made a mold of the translucent colored light "cubes" that sit over status lamps. I would use a slightly different casting resin, a clear version, and translucent dyes rather than the opaque type I use for the buttons, but the molding process is identical.

The button I was copying was stood up on its rear side, sitting on a base of scrap acrylic. I stuck it one with a hot glue gun and placed the light cube nearby on the same acrylic plate. The Platinum Silicone mold material is a 1:1 mix of the two solution, A and B. A small amount is mixed and brushed carefully over the IBM made button and light cube to form a thin "skin" coat over the items. Once this sets, it locks in the fine detail and surface textures of the IBM part I am copying.

Next, I mix a larger quantity of the molding material and add a thickening agent sold by TAP. The thicker goop is brushed all around the 'skinned' button, forming a thick mold with walls about an inch deep all around. When this finishes curing, it is twisted and bent to release the button and light cube from inside. The Silicone mold material is extremely flexible and strong, permitting the flexing needed to pop the copied parts out.

Casting a button involves another two part product, TAP's Urethane Casting resin. I mix parts A and B in the right ratio, add in the opaque dye to get the color I want, then just pour this mixture into the button opening of my dual-use mold and let it harden. The finished button is popped out by twisting and flexing the mold, leaving the mold ready to cast more buttons.

Light cubes use a different set of TAP products, their Clearlite casting resin and a hardener, mixed in the appropriate ratio. Translucent dyes are mixed in to get the desired color, then this mixture is poured into the light cube opening of my dual-use mold. I can simultaneously make one button and one light cube.

As a test, I cast both a button and a light cube, but using the opaque resin so the light cube isn't really functional. I was able to check my process and the fidelity of the finished product, which is excellent based on this first test. I didn't put in enough pigment to get the rich red color I wanted but the part works perfectly in the IBM switch assembly.

Once I pick up more pigment and the clear resin for the light cubes, I can knock out the needed supply and apply the labels to them. However, I have found some untitled buttons and light cubes to use as models, so I will first make two new molds. That will eliminate one objectionable defect in the first mold I produced - visible etched channels in the shape of the lettering IBM applied to that button/light.

I made another test run to produce an INTERRUPT button for Bob Rosenbloom. It was with the old mold, so I will redo this once I have the proper molds in hand. I tried to stir the pigment into the clearer of the two solutions that are combined to make the casting resin, hoping to get it more uniform, but I found that the pigment still does not mix well, leaving variations in color in the cast product. Once I sort out the color problem, I will be ready to make the buttons for my system.


I have been mounting all the switches and lights in the button holder plates. I recently sourced and mounted the toggle switches that will be used with the rectangular switch ends that I 3D-printed. These are for the 'power' and 'keyboard' switches. I also picked up three momentary push buttons to use with the three blue buttons on the face of the typewriter enclosure, and a momentary on - off - on switch for the tab set/clear switch.

I bought toggle switches to use with the pyramidal solid shaped handles I built on a 3D printer to copy the IBM implementation. The switches have a short lever that can fit into a notch in the underside of the handle.

The rectangular plate through which the switch arm projects was also printed on a 3D printer, but it needed a dome shaped inset formed on the inside to allow the short arm of the toggle switch to stick out enough to work properly.

The clearance through the black plate for light cubes and buttons was very close, in some cases wedging the button so it couldn't be pressed and spring back. I have filed off a tiny amount on the sides of the opening and fit these all properly. Once the button holders are installed on the underside of the black plate and all the light cubes and buttons fit properly, I pop-riveted the hold down brackets to keep these in place under the black plate.

I can begin fitting the entire black plate, keyboard, button and light assembly onto the support struts on the frame and fine tuning the tabletop recess to fit around the black plate. The hinge locations for the tabletop will depend in part on the black plate positioning, so the sequence of fitting these parts is critical.


I produced another set of decals for the panel, put them into the jig and pressed them onto the inside face of the plexiglas blank. I will need to be very careful with them, particularly as the adhesive has not fully cured yet, lest I rub them off or damage them during the subsequent fabrication steps. I have determined by experiment that spray paint won't hurt the lettering.

20" x 5" blank plexiglas to receive text and paint 

My original idea with the jig seemed sound - put the decal, glue side up, onto the jig, then drop the clear plexiglas into place, then rub. Thoughtlessly, I used painters tape to hold the decals into place on the jig surface, which meant they were stuck there, not to the acrylic panel. I am not sure that it would have worked well without the tape, but adding the tape doomed the attempt.
Decal step one, laser printed mirror image of text

Jig used to control alignment of layers of the panel image
I cleaned up the panel, made a new set of decals and tried another way. The decals were trimmed closer, to make it easier to place them and control them. I put  the plexiglas down first, then lining up the decal and rubbed them on. The challenge with this final method will be to get it lined up exactly right when the first letter touches the acrylic, because any attempt to change position from that point will ruin the decal transfer. For some reason, undoubtedly a defect in the glue application phase of making the decals, the rubbing did NOT transfer the letters.

I tried to rescue the situation by continuing with the clear mylar left in place, adding paint and other layers on top of it. I doubted it would be acceptable, because the edges of the mylar would be visible, but it was worth a try. I couldn't put the paint mask down the normal way because the glue on the vinyl would pull up the mylar with the letters.

Two paint masks and decals used to build up layers of the final image
I tried to install the paint mask glue side up, taping it down in position. It was very hard to adjust everything because I had lost all visual cues to alignment. If the lettering mylar slid on the acrylic, I would have the paint mask centered on the misplaced lettering not on the plexiglas itself.

I finally got it in place, but grabbed the wrong gray paint (the darker color). This was a show stopper. I cleaned off the plexiglas, and will use the blank on a new attempt later this week.

I made some changes to the jig approach. The decals are trimmed closer, to make it easier to place them and control them. I put a thick plain paper backing on the decal, on the 'rubbing' side not the glue side, then set the sandwich in the jig. That extra height should ensure that the decal is pressed up against the acrylic panel, so that the panel, decal and paper all come away together. The plain paper backing is tall enough to extend out the top and bottom of the jig, helping me to lift everything together. Once the paper, decal and panel are lifted, I flip them and rub the decal in the traditional way.

My next attempt began with fabrication of a set of decals. I was extra careful about spray adhesive application, since my last try suffered from lack of transfer of the letter when I rubbed them on the acrylic. I put the two decals atop the thick paper backing, lined up everything, and began lowering the plexiglas onto the jig.

Stacked mask and decal - just placed in position to show how it builds up
Once again, the gremlins struck. After the plexiglas had touched the left decal, which means we are at the point of no return, the right decal began curling and lifting toward the acrylic due to static electricity. I paused with the acrylic pivoted up, tried to settle the decal down, but realized it was not going to behave. In order to save the left decal, I instantly decided to turn this into a two phase process - left decal in the current phase, then come back later to add the right decal. Yanking the right decal out of the jig, I put it aside and continued to install the left side text.

The transfer worked perfectly this time, although there was a small bit of damage on the decal. An 'm' on Accumulator had part of the letter missing and part of the 't' in Arithmetic also went away. Given the hundreds and hundreds of processing runs it took to get to this place, I am giving it a tentative 'pass' as good enough even though it is a bit flawed.

I am worried about damaging those letters in the course of transferring the right decal. I might shoot on a protective clear coat and let it dry. I don't want to make anything worse when I am installing the right hand decal.

I did use the clear enamel paint to protect the lettering, let it dry for half an hour and then transferred the right decal onto the plexiglas panel. That worked very well. Now that I have sprayed clear enamel on both sides, the lettering is protected and I can move on to install the additional layers of paint.

Before I could try the light gray paint step again, I needed to visit Techshop to cut new vinyl paint masks, since I used one in my last failed attempt. Fortunately the files are all ready to go, I just needed an hour at the vinyl cutter area. With the replacement mask in hand, I could proceed to the next step.

First vinyl mask in place atop rubbed on lettering
Openings in first mask to allow light gray paint to be applied as box around text
I trimmed and set the first mask into the jig, then pressed the now-lettered plexiglas plate onto it. The light gray color was sprayed carefully onto the assembly and allowed to dry before removing the first paint mask.

Light gray (actually white paint) sprayed through mask
Once the mask was gone, I inspected the result to be sure that the panel had no unacceptable defects. I did find that a bit of white overspray made it through one area where the mask had bowed up off the panel surface. It was relatively easy to scrape that coloring away with a straightedge and hobby knife.

Second mask in place, seen from the front
The next step is to spray on a medium dark gray paint through a second mask, delineating groups of lights that belong to individual registers. As before the second mask was placed in my fabrication jig and then the plexiglas panel was pressed into place. The newly masked assembly was given a coat of the medium gray
paint and allowed to fully dry. Once ready, the mask was removed and I inspected the panel carefully. Once again, a few small spots where the mask lifted a tad and I had paint overspray, but some work with an Xacto knife cleared that away.

Medium-dark gray sprayed through second vinyl mask
What the second color looks like on the back, where it was sprayed

The front view of the panel after three layers put on back
The next step in this process is to spray a coat of flat black over the entire panel. This darkens the gray colored sections a bit, makes everything opaque, and fills in all unpainted sections with black. Once dry, the almost completed panel was given another inspection. I found a few parts of the black paint to be too thin, but another coat fixed that problem easily.
First coat of black paint sprayed across entire rear side of panel
The last step in fabricating this panel is to etch away the paint in front of every lamp that will light up. The etched space is in the shape of a character appropriate to that lamp position. For most, it is a digit shape, but some of the lamps represent machine states, e.g. ZR, or clock and instruction states, e.g. IX; these exceptional lamps have the letters etched from the paint.

I prepared a file with the characters placed in their relative position from the top left corner of the plate, mirrored the entire image since we are etching from the inner face, then used it to drive a CNC laser cutter which will burn away the paint in those shapes. I did have to wait a day or two for free time on the schedule as these machines are very popular and highly used.

Laser cutter pattern for where to burn away paint
I made a test plate of similar acrylic and the same spray paint. I used that to adjust the power and speed of the laser to get full removal of the paint layer without melting through too much plexiglas. I set up a couple of files that were subsets of all the openings, to fit the smaller test plate, then ran them until I was satisfied

In testing, I discovered that my pattern for the numerals and letters for the righthand registers was shifted too far to the left and didn't align with the panel pattern (or the LEDs that will shine through). I was able to fix the design and carefully measure to verify fit. I also found a power level high enough to reliably remove the layers of paint but cause no damage at all to the acrylic underneath.

Then when I set up the painted panel to do the production run, somehow I didn't have the home position adjusted properly. The cutter began etching the numerals at the very top of the panel, not down in the gray colored rectangles where it belonged. I was able to stop the machine before it finished that row. Since the area is supposed to be pure black, I will just need to mask off the rest of the panel and spray some more black paint to "erase' the etched numerals.

After power cycling the machine and the software, to ensure it was not going to continue the 'wrong' layout or remain misaligned, I ran the etching of all six rows. The alignment left to right is very good. It appears my panel was slightly skewed , thus the rows slant a bit and are not perfectly centered in their gray rectangles. However, it is within acceptable bounds.

The now completed panel was carefully inspected so that any flaws in the letter etching can be addressed with some hand trimming or hand application of paint. It seemed like it would be a simple matter to spray some black to cover the false numerals at the top, a bit of masking to protect the rest. Boy, was I wrong.

The gremlins that plague me every step were very active. Strike one - the first run of paint exhibited some bizarre curling and wrinkling, which might have been contaminants that somehow got on the surface. This is unfortunately visible through the front. I scraped it down flat and went to spray another coat, hoping to hide the wrinkling.

Strike two - the black paint looked fine from the back but the pattern of the wrinkling is still somewhat visible from the panel. Not sure I have any safe way to fix this.

Strike three - when removing the blue painters masking tape, a special very low stick version of the regular painters masking tape, I had a section of the flat black lift completely away. Now there is a transparent irregular hole that goes right to the top edge of the character such as T7.

My only hope is to find a flat object with a good hole that I can carefully lay atop the panel rear face, butted up to the tops of the characters I need to protect, then spray through that. It may need to be weighted down, but it will NOT have any adhesive touching the panel. I also need to be sure the material I use won't stick to and pull up any paint. I will need to be very very careful at this stage as the panel quality is spiraling downhill with every touch up attempt.

After removing more of the paint layer causing the wrinkling and using a metal cutout as a paint mask, I took care of most of the remaining defect. There is still a small open area just above the T clock characters, which comes too close to the letters to risk with spraying. I will probably put an opaque sheet material behind the panel, with the same coloring, and hope it is not noticeable. However, I did create a paint mask out of acrylic for a slit that I might use to try to further narrow the open spot.

My panel, before top row error was erased by black paint

IBM made 1130 panel for comparison to my completed panel
When comparing my light panel to the IBM fabricated panel on the 1130, keep in mind that there are some status lights in the lower middle that I chose not to implement. Eight lamps in the next to last row of the plain black middle area were made available for the field engineer to wire up to display signals, if they needed to see the status while diagnosing a problem with the machine. I have better diagnostic tools and wouldn't use these. The bottom row of lights are installed if the optional communications adapter is installed; I am not implementing that adapter.

Light coming through etched characters where lights will shine through


I rubbed the white decal on the mode switch plate. The lettering is a bit small and I didn't get the alignment exactly as I wanted it. The black circle is 2 3/4" not the intended 2 3/8" of the 1130, and I swapped the positions of the DISP and LOAD labels. Unacceptable, thus will be redone.

Decal applied to mode switch plate

File used to create decal
The knob needs a pointer attached. My first attempt to do this with traditional plastic model making cement failed. I tried with brads, drilled holes and cement, not sure it will work any better but will let it set for a while. After adequate time for setting, I looked it over and was not pleased.

The mounts inside the display pedestal box needed to have their holes widened to allow some degree of adjustment of the plate position. Once done, and with some other tweaking, the plate was put into position on the pedestal box, awaiting its knob with pointer.

During the manipulations mounting the plate as well as working on the balky pointer, I managed to damage some of the lettering on the plate. At this point, I plan to get the knob and pointer completely done, wire the switch and finalize the mounting arrangements before I disassemble, repaint the black circle, make a replacement decal and transfer it on. This way I can get the whole execution more like the 1130, anyway.

At the shop, I cut a vinyl pain mask of the correct diameter. It is applied over the plate, first painted a uniform white. With the mask in place, I can spray black paint to form the circle. When everything is dry, the mask will be removed and the lettering decal is applied. Finally, a coat of clear enamel will be applied to protect everything before I re-install the switch and re-mount the panel.

Vinyl paint mask to spray 2 3/8" black circle on mode switch plate

Mode switch plate with white base coat before circle and text applied.
Unfortunately, vinyl is easily pulled and stretched. It was not possible to lay the circle down without it becoming an unacceptable deformed oval, then pulling up the partially dried white paint when I tried to straighten it. After repainting the plate white yet again, I began to look for an alternative way to make a mask to spray a circle with nice clean edges.

I plan to cut a mask out of black acrylic using the laser cutter, which I will spray the black circle through. That will give me the smooth edges and dimensionally stable circle I need. Once I am back with the mask, I can add the black circle, then rub on the decal, and finish the plate with clear enamel for protection.

The paint mask is done and will be used with a new can of black spray paint early next week. In the interim, I already wired the switch so it is ready to go once the knob issue is solved.


I put in some more lathe time, finishing up the milling of the Emergency Pull knob. The knob was attached to a large aluminum cylinder, with a slanted surface down to a 1/4" diameter shaft. The large cylinder was chucked into the lathe and a drilled point on the shaft was held by a live center tailstock to minimize flex of the shaft as it was milled.

First, I smoothed the slanted rear surface under the knob disk, as this is where the operators fingers would rest if pulling the switch during an actual emergency power-down. Next, I cleaned up the shaft that extends into the pedestal display box. That done, I took the work over to a horizontal bandsaw to cut the face disk from the larger aluminum cylinder.

The cut down knob was now mounted with the shaft in the lathe chuck, reversed from the position used to perform the first few operations. I evened off the surface of the disk, then cut a slight recess into the face which will be painted red and contain the label.

Milled EMERGENCY PULL knob, before painting and attachment.

Another view of the milled knob

A bit of cleanup removed some burrs, then I masked off the rest of the knob in order to paint the recess a bright red color. This red face will have white text on it - EMERGENCY PULL - which comes from a dry transfer decal I made last week.

Recessed face of knob has red color upon which text will be placed

Protecting the rest of the knob from paint overspray.
I have a couple of ragged spots where masking tape over-protected the knob, blocking spray from evenly coating the recess. I will manually touch this up tomorrow once the paint is dry, then apply the decal and protective coat.

Ragged edge at top right needs to be manually filled. Color off due to poor lighting.

Remaining steps after applying the decal are a clear enamel coat and polishing of the mill marks off the circumference and shaft.

Side view of lettered EPO switch knob

The knob ready for polishing, a protective coating and mounting.

One open item in my design is whether this will be a fixed mount for the knob or whether I will hook it to a switch and activate some behavior when it is pulled. I will defer this decision for a while.


My latest run producing decals yielded two misfires, all the rest came out perfectly. On one group of titles for buttons, some bit of dust or other object lodged between the white mylar sheet and my printed image, blocking the transfer of the color to part of one letter. Another group of four titles went through the fusing process but yielded a backwards result - white was transferred everywhere EXCEPT where I had black toner in the image. I have no idea why this happened, but the group as well as the failed title from the other group must be redone in a later decal session.

I have transferred lettering via decals to several items already, then covered them with a protective clear enamel paint spray. The completed parts are:

    Power on/off toggle switch handle
    Keyboard kb/ces toggle switch handle
    Int Req keycap
    Rest KB keycap
   Emergency pull knob

Toggle Switch handles after text added with decal
White text decal applied to switch knob

Applying the decals to keycaps is especially challenging because the surface is hollowed inward to match the human fingertip, thus the decal has to be bowed down and rubbed in that "inverse dome" area. It is hard to get them placed correctly and to apply the rubbing pressure due to the strange surface.

I noticed that the decals with smaller font lettering exhibited jagged edges (at fine detail, not grossly visible) from the rasterization and relative pixel size during the creation, manipulation and printing phases. If any of the results are particularly objectionable to me, I will seek an alternate way to print those text strings that reduces the quantized step effect. 

No comments:

Post a Comment