week ending August 25, 2013


I was assembling the red button I had made for Bob Rosenbloom when I spotted a major foul-up that made all of my buttons unusable. The button and light cube are superficially identical in outline, the main differences are material and opaqueness. However, there was a third difference I did not spot until today.

The two example buttons I had in my collection that had blank faces, which were essential to allow me to label buttons for any purpose since IBM etched the labels into all of its buttons.  However, those were both designed to snap into a housing but to not slide in and out - they stood at a fixed height just like the light cubes, whereas buttons must be able to move inward when pushed to depress the switch underneath. Buttons have a cutaway on the side that allows them movement in the holder, while light cubes have a fixed short notch that fixes them in place. Where I ran afoul was that the blank button I used for a mold model was actually shaped like a light cube. It had the fixed short notch.

I had no buttons without labeling - which makes sense as IBM would not set up a button to push in if it had no function, and if it had a function it would be labeled. What I did have was a surfeit of various buttons, both the usual labeled kind and two of the fixed position pseudo-buttons with blank faces. I sawed a blank pseudo-button apart and sawed a real button apart symmetrically. These two were epoxied together to make a frankenstein button, having a blank face but the movement-permitting long cutaways on the side. That new model was used to create molds for casting my buttons.

The buttons I had produced, pretty as they were, would not move up and down in the holder. Cosmetically good, but functionally useless. They all have to be replaced. At least I could use the existing molds for casting light cubes, which must stay in fixed positions in their holders.

My first test cube used pure transparent dye, but was far too transparent to match the frosted appearance of the IBM light cubes. I then mixed a small amount of opaque white pigment in with the transparent green dye and got a similar appearing resin. I have two castings sitting in their molds and will be able to verify its correctness tomorrow when they are fully dry.

Mostly my castings but a couple of IBM parts - green start button and red file check light cube
This sets back the schedule. I will attempt to make two molds today from my frankenstein button model and then a third mold tomorrow morning. That way, I can cast two buttons tonight and add a third mold for a round of three castings tomorrow night. As well, I now have two good molds for light cubes and can produce them two per day. There is only two more keycaps to make, one of which is already in the mold. Tomorrow I will pour the last one and by Wednesday it will be ready to demold. All this requires another trip to Tap Plastics for more mold making silicone, stirrers and mixing cups.

My first few light cubes didn't hit the colors I wanted. The two green ones are a noticeably lighter color and the two white cubes had to be thinned so much to switch from opaque to translucent that they seemed to take on a slight grey tinge. I will need to see them once they are out of the molds tomorrow night.  I have two button molds, the second still setting up, and the face of the first might not be smooth enough because I pulled it off a bit early to try to get two molds produced in one day.

By Tuesday I had one perfect button mold and a second mold starting to gel. Once I remove the casting from the original button mold from yesterday I can see if it produces acceptable buttons or I have to make a third mold overnight.

Molds for light cubes and buttons, plus a blue button and two failed light cubes
The mold was not good enough when I checked the result, so the third mold is now setting overnight. I made two buttons, one full size and one short, both intended to be red, but for some reason the coloring came out orange, very orange. Rejects. Along with the third mold, I cast the last of the tan-gray keycaps, two more white light cubes, a green button, a short red button and a full size red button. The keycap, white light cubes, short red and full green button came out fine.

My frankenstein short button at top - one anomaly on side to cut off but otherwise good.
I had flaws on the face of the full sized red button. Continuing on, I cast three blue buttons, a red light cube and a yellow light cube. One of the three blue buttons I cast had flaws and must be replaced. The yellow light cube came out as an orange cube - I don't have a transparent dye that is a light enough yellow, all my choices have too much red in them. The red cube is good, although not quite frosted enough for my taste.

The light cubes remain the most vexing part of this project, because I am having difficulty getting the hues and the degree of frosting using the dyes and pigments at hand. I might try a bit of sand blasting to create the frosted look, but that will also give me a rough feel to the surface which is not present on the IBM cubes.

One idea I am toying with is to cast the light cubes in pure translucent colors of the desired type (assuming I can solve the problem of a good yellow), sandblast the surface to get frosting, then swipe fresh resin on the outside while sticking them back in the mold for a day. That might give me a smooth outer surface over frosting, but somehow I doubt it would come out good. The resin would sink down the sides, not evenly gelling around the cube. I really need something to add to the resin that gives it a milky translucency prior to adding color.

If I figure out a good way to get the light cube look I want, then I will have to re-do all eight. Another option at hand is to fill in the etched face of the existing IBM made colored cubes, hopefully with some kind of clear resin, then just title them as I wish. I would also need to remove all the paint from the etched letters as a prep step, but that is probably easier to do that getting a resin layer over the face to be even and smooth.

It seems the smooth face might happen if I dribble a small amount of resin into a mold and then jamb the IBM light cube into it. Gravity would hold the resin at the bottom of the mold, where I want it, although the concave face means a convex bottom allowing the resin to sink away from the letters at the high point of the face.

My thursday night casting, removed Friday evening, included what I hoped were the final two red buttons, but when the red buttons came out, their face had yellow flecks that were really visible. Apparently a bit of the yellow pigment that was mixed with the red didn't dissolve, instead dropping to the bottom of the mold, which is the face of the button. My luck with the red buttons has been extraordinarily poor. I poured another two red buttons Friday evening, to be removed Saturday evening. In addition, the Friday night pour included the last blue button and two light cubes.

The cubes lightened in color during the curing, enough to turn a perfectly colored red translucent cube into an orange cube, and to take the light yellow cube and make it clear white. Drat, I have to figure out how to compensate.

However, even the two red and one blue button gave me poor results once again. One of the red buttons had a nice smooth surface but still had some flecks of orange on the face. They are much smaller and I am just going to use it as it is, because I have no reason to expect it will come out any better on the next dozen tries. The blue looked nice but has a small ridge across the face, something that seems to have cropped up on the mold itself. The last red came out smeary, wet and blotchy on the face.

Saturday night I did a casting of one blue and one red button, plus two green light cubes. Sunday while they are curing, it appears that the light cubes might be a bit too blue to be acceptable. However, they continue to lighten so the jury is still out. The buttons are the proper colors, but the quality of the faces won't be known until I 'decant' them all tonight.

Well, Sunday evening and the two buttons came out great! There is just the tiniest of ripples on the face of the blue button, a little indentation in the mold causes this. The red button has a flawless face and nice even coloring. The two green light cubes are a bit too translucent, not as much frosting, and the tint is off a bit but not too bad. These are probably good enough to use. Once I put on the lettering I will make the decision.

I still need to make several buttons and light cubes:
     two full sized blue buttons (to complete the typewriter black plate)
     one yellow light cube
     one red light cube

I poured an attempt at the remaining two light cubes but will wait to try the blue buttons until Monday. Just emptied another can of resin, the final can will go into use tomorrow. It has to cast the two typewriter knobs once I make their mold, and it has to cast the typewriter tab set/clear knob when that mold is created. Should be plenty even with some failures and experimenting along the way.


That afternoon I attended a class at Techshop on designing molds for plastic injection molding using Autodesk Inventor Pro - as I as a bit unsure as I had worked through the tool on my own, this small investment seemed worthwhile.

Knob design that will be produced by injection molding
I learned a tremendous amount - not only about the software process but about plastic manufacture, molds, milling and other tips and best practices. Unfortunately, I learned that plastic injection molding yields parts that begin to experience sinking surfaces when the thickness of any feature is more than about 3/16" and my typewriter knob is about twice that thick. I would either have to hollow out the knob proper to get below that thickness or go to a plan B for making the knob.

A flash came to me as I was thinking about this problem. The work I would do milling a mold for the knob produces a mold. That is, a mold that could be used for casting with resin, not just for injection of hot melted ABS! If I can produce a smooth enough negative shape, it will act as a mold for resin. There are a few minor details for how I build the mold - it would be a two-part mold that fit together at a parting line but perhaps it doesn't have to be milled out of aluminum. Anything that can be milled to a smooth finish will work. 

I could 3D print a model of the knob, with the slightly rough finish it produces, then try to get it smoother by spraying on a primer, perhaps with a bonding enhancer first, then smooth sanding it down until the surface is what I want.

I did exactly that - modeled and then printed the knob - although for mechanical reasons I did it as an upper and lower half so that I can create a two part mold. The only part that doesn't have a good enough finish is the domed 'top', exactly what I suspected. It consists of concentric bands of plastic rather than a smoothly rising surface. I just need to cover it with a smooth dome shaped item long enough to make the mold. If I can fine some kind of cap for the molding, I don't even need to smooth the actual surface of the printed knob, but I am still working out how to move forward.

3D printer output of the top and bottom halves of the knob

I worked on a set of dry rub transfer decals to label all the buttons, light cubes and keycaps I have made. The process is finicky, everything has to be done just right, and it is easy to mess up a decal by a small slip or mistake.

The decals came out well, most worked right the first time. I completed labels for four of the six buttons on the black plate and for all eight of the light cubes. I have a few more decals to do, the three buttons and a knob on the typewriter plate, the two buttons on the blackplate and the four keycaps, but rather than slog along and begin making mistakes, I chose to stop Saturday night and do those remaining ten on Sunday.


I took the two front upper sections to the shop, welded the seams and ground them down to a reasonable surface. First, I have to report that I am absolutely terrible at welding, particularly these thin sheets of metal. The raw result of the welding is quite ugly, but after grinding it gets down to a more reasonable level. As an indicator of the final condition of my parts, I did a quick spray of gray primer and some texturizing on one of the ground down edges.

My horrible raw welding with thin plates

Test of what my welds would look like after grinding them smooth


My toggle switches (POWER and KEYBOARD switches) need to be mounted to the button holder under the black plate. I designed a small bracket to hold the two switches I bought at Radio Shack onto the button holder plate. That was cut on the waterjet at the shop and bent to rough shape. However, the switches can slide around a bit and wedge at a tilt, which requires an additional bit of bracket to be added. I will work on that, because once I get these mounted to the button holder plate, I only need to install the base plate plastic and press-fit on the handles to finish this up.

Toggle switch mechanism with my switch holder plate in place
If I used epoxy on the handle, it would mean this assembly is one-way - once glued together, I couldn't remove the button holder plate or the black plate from the keyboard assembly without breaking that handle off. To get a good press fit, I have to be very accurate drilling the slot into the back of the toggle handle. I will practice with my keycaps which also need that level of precision, since they too are press-fit onto a 1/16" thick tang from the keyboard.

I worked on the 3D printing files and will give a try at making the base plates and new toggle handles with the press fit slot built right in. If the base plate works well, I will make a second one with some red filament that is on order from Amazon.com.


The switch to set and clear tabs has to fit over the bat handle of a switch I bought at radio shack, be cosmetically the same as the 1130's handle, and work properly when the switch is swung up or down to set and clear tabs. I did the design work on Autodesk Inventor to create the 3D model, which I will fabricate on a 3D printer as a model. That model will be used to build a mold, ultimately letting me cast the handle in the correct blue color to match the other buttons on the typewriter.

Design of the handle for tab set/clear on the console printer black plate

Rear of the Tab Set/Clear handle where the bat handle of switch is inserted


After carefully protecting the two painted faces of the logo plate, I polished up all the edges with steel wool, then sprayed a potective coat of clear gloss enamel over everything. The last step was to epoxy the threaded rods into the holes in the plate. With this done, I mounted the plate using lockwashers and nuts inside the pedestal box.


One of the forums I frequent, for vintage computing enthusiasts mainly in the midatlantic region, has a member who snared a 370/145 console panel on ebay and wants to get it operational. His first version will be connected to a PC running a 370 emulator program called Hercules, with the panel displaying the register contents that are tracked by Hercules. The challenge he faces is that the prior owner sawed the lights apart behind the panel and hot-glued some mini christmas lights in place. That created blinking lights, but not a realistic pattern. He tore those xmas lights out and wants to put in incandescent lamps to be driven by the PC. He will build the driver logic to operate the lights but the physical restoration is a problem for him.

I have been exchanging emails with David and will help him deal with the physical part of his project. I have a 360 or 370 light, of the type that is on the console, with a round lens and plastic ring, which I could use to figure out a way to build usable replicas. The first phase of David's project can take advantage of the remnants of the IBM lamps, which are still in place on the panel, but open on the back which would allow use of some kind of lightbulb to shine through.

IBM light similar to the ones in the 370/145 panel, from my stock

Sockets for the lights, from my stock. Similar to the sockets on the 370/145 console panel

I made a suggestion for how to proceed - depending on his wishes, I can help by using a lasercutter to shape some acrylic to hold lightbulbs and isolate the light from each bulb. He is sorting out his ideas for how to proceed - lots of factors he has to consider. I will wait to see if he wants any further help. 

No comments:

Post a Comment