Continued construction week ending July 7 2013


On Tuesday I was able to weld the mid and front frame segments, plus cut and prep the tubes that will connect these in the front to back direction. That evening I did some grinding to clean up the welds - some of these will need a front-to-rear tube welded to them, but they all will look better if ground to be fairly flat.

Once the frame is completed, I expect to spray paint it black, consistent with the IBM frame color. My frame uses 1" square tubing but IBM built the 1130 with 1/2" square tubes - mine will be heavier and more rugged.

Friday I welded almost all the remaining tubes at the shop and did the grinding of those welds at home. I believe I will have a total of six welds to do over three tubes, after cutting one of my remaining tubes in half. This should a snap to do on the weekend, at which time I can grind the six down, clean the frame, prime it and spray it gloss black.

 On Saturday I finished up the welds, smoothed everything, and mounted the swivel casters to the bottom. I find I need to add one more swivel caster to get full stability, due to the cantilever desk which is right where a user will be applying downward pressure.

I cleaned it of oil, primed it, and finished painting it black. Here it is, on casters, ready to get internal parts and metal covers.
Frame of 1130 replica, freshly painted.


I was able to use a photo of an IBM made logo plate as a source to create a vinyl paint mask of the text "1130" - thus achieving the rectilinear shaped outline text that I had trouble locating as a font. I cleaned up the "IBM" logo, moving the letters much closer together in a new vinyl paint mask that will lay down this plus the text "Computing System". The fabrication now involves two masks applied serially - the "1130" applied second, once I have sprayed down the other text as that yields a better alignment reference to locate the 1130 text.

I decided to reprime and recoat the logo plate on the side where I made my first try at the lettering. Once it dries, I can put down the new version on both front and rear faces of the plate. There may be some raised surface on the rear face, since I didn't sand down or machine away the initial try, but it is out of sight when looking at the replica machine from the front or sides.

Here is the outcome of applying the new vinyl masks to the rear face. Although the alignment is off a bit, the letters skewed and so forth, it is good enough to convince me that with a bit more care aligning the masks, the front face will be quite acceptable.

Second try with masks for the rear face of the logo plate
I bought a stainless steel 1/4" rod to use for the stand holding the logo plate above the display pedestal box. I had to cut off the existing threads, in order to cut the 10-24 sized threads I have tapped into the logo plate. Once I have the threads cut, I will trim this to size and install in the logo plate.

My first try to cut the threads on the lathe this Friday wasn't successful - I need to add a live end to suspend the rod from both ends, as the stainless steel is too tough and flexes while cutting, producing a messy thread.

I put together a quick alignment jig for the plate and got better results for the front face, but not ideal. For one thing, the image I had to use to recreate the '1130' outline font was at an angle, so the letters reduce in size from left to right. It was not possible to align them ideally on the plate, but I did the best I could trying to match things up.

Quick alignment jig to install vinyl paint masks on logo blank
Here is my final front face on the logo bar - I consider this good enough to use
I could quibble about the rounded holes in the IBM logo's B, the overly wide tail in the middle of the M, wish the IBM was justified slightly to the right and that the font letters were a tad stretched vertically. More caviling about the 1130, whose numerals shrink in height from right to left, which could be justified a bit more to the right and once the shrinking was fixed, should have even spacing top and bottom.


I was able to grind down the horrible welding I did on the pedestal stands and pedestal box sides, which convinced me I don't have to junk these and make another set. I was able to flow some solder to make the surface even smoother, where there were holes burned into the stand sheetmetal because I was welding with the .035 wire configured in the Techshop MIG welder.

Smaller wire, .025 or less, is needed for thin sheet metal but I am not sure if they have the rollers for the Lincoln welder to handle the wire or the adapter for a smaller spool. I bought the wire itself in anticipation of using it but tried instead with the default wire and suffered burnthrough.

I have a bit more grinding to get the surfaces as flat as I can, then I will use some JB Weld as a filler. It has been noted that this filler works well with powder coating, for two reasons. First, it contains steel particles thus participates in the high voltage charge needed to attract the power. Second, it can stand up to the 350-450 degree baking temperatures of the powder. I picked some up at a local hardware store and applied it to create a smoother surface for both stands and pedestal box. It had to cure overnight before I could work it.

I sanded down the JB Weld, found three small spots that could use more and applied a second coat Saturday morning. It should be cured on Sunday and can be sanded flat so the parts are ready for power coating.


As soon as I figured out my method for attaching these to the pedestal box, I could clean, prime and paint the plates. I chose to solder two nuts on the back of the plate, into which I can thread a bolt to cinch the plate onto the front of the pedestal box. Using a propane torch in my garage, I quickly soldered some 1/2-24 nuts to the two plates.

They were then primed and given the base coat of white paint. That is all that is needed for the left plate, which has only the Emergency Pull knob emerging from it. The right plate had a circular vinyl paint mask placed and a coat of black sprayed.

Painted plates for front of display pedestal box
The remaining step is the installation of rub-on decals for the lettering for the seven mode switch settings: Run, SI, Disp, Load, SS, SMC, and Int Run. I am still working on the process and not yet getting high enough quality decals.

My knob for the mode switch does not have a pointer. I removed one from a different IBM 360 era knob and glued it to the base of my knob. I have also completed design files for both decals and vinyl paint masks to create the lettering around the mode switch. If I get the decal process fixed, I will go with that but my fallback is to use the vinyl cutter to make masks that I will spray some white paint through.

After cutting the fallback vinyl masks, I found that the lettering is too small to reliably work with vinyl - the little circles of vinyl that are on the inside of the letter "R", for example, are so microscopic it is effectively impossible to remove the outline of the letter while preserving that circle.


I have completed the Autodesk Inventor 3D modeling of the 16 gauge steel plate that will be cut to allow the keyboard and various buttons and lights to be mounted. I will use the CNC waterjet cutter to cut out the openings, after completing my training on the machinery at Techshop. I need to move the 3D model into Flowshop to create the CAM paths that will drive the CNC cutter. The cutter software wants a 2D model, which I built by exporting the face of the modeled part. This machine is very popular, so the biggest challenge is to book time to do my cutting.

At the same time that I am cutting the black plate, I will cut two copies of the button/lamp/switch holders - these will take the IBM button/lamp housings I recycled from some 360/370 era surplus equipment, holding them at the proper spacing. Each side of the black plate has openings for a grid of such buttons, two across by four down, each button fitting through an opening that is 1" wide by 3/4" high.

The button holders are a bit wider where they will snap into the holder plates I am making, for example they are 1 3/32" wide while the button or light block that sticks up out of the black plate is exactly 1". My two holders will be mounted under the black plate with spacing adjusted until the button and lamp blocks stick up above the plate the desired amount.

The updated files were processed through Flowpath, yielding the cutting path that will be used by the waterjet machine. I was able to reserve time for Sunday on the cutter, also reserved the sandblaster and powder coating oven for that evening.

The process worked out extremely well - here are the steps and the final result. I used the CNC waterjet to cut the steel plate, which you can see in process and then the completed blank. I sprayed it with powder in the paint booth and then baked it in the oven.

CNC waterjet cutter
Plate in the middle of being cut by 55,000 PSI stream of garnet sand
Plate after cutting was completed
Powder spray booth where 14KV attracts powder to stick to plate
Oven in which the plate was baked at 375F for 15 minutes
Plate cooling after finish baked in oven
Plate atop keyboard, with loose lamps and buttons sitting there for artistic effect


I ironed out some small wrinkles and behavioral tics in my adapter logic, validated operation of all functions and proper typing of all characters, then created a sample program to produce typed output for videotaping.

The carriage return function had been malfunctioning when the distance to return wasn't short, because it went into high speed mode and my arithmetic and detection logic to spot when we reached zero (left margin) wasn't fast enough to keep up. It also appears that stray light is inducing some errors in the emitter counting; this will go away when the proper covers are installed on the typewriter mechanism.

Debouncing the phototransistor based emitter is my biggest headache. The shaft can vibrate slightly particularly at start and stop times, but also any partial binding during a movement cycle may induce just enough rotary vibration atop the average rotational speed that we can get the emitter producing spurious pulses. On the other hand, really long debounce delays can miss pulses when the escapement is moving at high speed mode.

I wrote code that works like a schmitt trigger combined with a smoothing capacitor - it maintains a counter, adding up in any cycle when the emitter is on and decrementing by one in a cycle when the emitter is of. I set two 'rails' - thresholds - for the counter, namely 0 and 5000. If the counter reaches either threshold, the debounced output switches to that value and remains there until the counter wanders to the other rail. Thus, bouncing or noise is smoothed out, requiring 100 microseconds worth of one signal state over another. If the line is noisy, it takes more than 100 microseconds to snap to the other value, but with every third sample wrong it will still flip to a proper state in about 300 microseconds.

The hidden shift functionality was still a bit erratic. I also found that running the console diagnostics at full speed would produce characters faster than the mechanism seemed able to type them. I will continue to probe at this until I am satisfied with the operation, but by and large the interface is complete.


Many of the tasks ahead depend on my ability to produce dry transfer (rub-on) decals of text. Many buttons and switches will need these, but the most critical part that requires this is the plexiglas panel that sits on the display pedestal box. I will build up the image in several layers on the inside face of the plastic.

First will go the decals with mirror image text in black. Next, a light gray is sprayed through a mask to form rectangles behind the text and to make vertical divider lines that separate each group of four lights for a register - that nibble corresponds to a hexadecimal digit which is easier to remember and manipulate than a string of ones and zeroes. Third, a darker gray is sprayed through a new mask, forming long rectangles for each group of lights representing a machine register. Fourth, a general coat of black is sprayed over everything on the panel.

Last, a laser cutter will etch away the paint where each LED will shine through, the opening etched in the shape of the numeral or letter related to that light. For the main registers, the rightmost light has the digits "15" etched through the paint, so that when it lights a '15' is lit up from behind. The leftmost light has a "0" opening, as that is bit 0. Clock cycles are etched as two characters, lights being labeled with pairs such as X3 and IA. Other status conditions have their own two character codes etched through which the lamp shines.
IBM made panel from an 1130 
The decal making process is complex and each step must work correctly to end up with a good decal. I have mastered all but one step, having issues with the bonding of a transparent mylar film to the fused image, which causes the colored mylar fused on the paper to adhere more to the mylar film so they can separate in a water back leaving the image on the clear mylar rather than the paper. I get incomplete transfer, with some parts of the image floating away or staying on the paper.

The image is printed on special paper using a laser printer. That paper with the image has a sheet of colored mylar placed over it and the sandwich goes through a hot laminator. This fuses the colored mylar onto any part of the paper that is black, e.g. has toner applied. It peels away form the 'white' parts of the image.

Any bits of the colored mylar that have remained on the paper where they should not are removed by using a low stick blue painters tape. I have not had any mylar other than where it belongs, thus haven't needed this step to date.

The fused image on the paper and a thin clear mylar film are wiped down with alcohol and them buffed to create a static electric charge on the pair. They are put together and then passed through the laminator to intensify this static adhesion. Here is where I am having the problems - my mylar film and fused image are not sufficiently bonded after the passes through the laminator.

After trimming any excess mylar film off the bonded sandwich, it is dropped into a bath of water. The sandwich curls up and then gradually loosens, with the clear mylar film floating clear of the paper, taking the fused colored mylar image off the paper. In theory. The mylar film comes away without parts of the image.

The mylar film from the bath is dabbed dry, then placed with the image side up atop a paper towel with a small amount of spray adhesive. More adhesive is sprayed across the image side of the film, which is now a decal.

To apply the decal, the image side with its tacky adhesive is placed on the target surface and the back of the mylar film is rubbed. This sticks the image to the target surface, which remains there when we peel away the clear mylar film.

I did a test to make sure that the decals will stick to the plexiglas panel and that spray paint placed over the decal will not melt the letters or effect them. I used a failed decal, where you can see that parts of the image are missing, plus a white spray paint, on a leftover scrap of the plexiglas. The result is good - the part of the image that was on the decal is clear and well defined on the plastic, with the white paint background as expected. This gives me confidence in the complete process I will use to fabricate the display panel.

Process test of text decals and spray paint sandwich behind plexiglas
I am hoping I can sort out the issues and get my decal process dialed in over the next few days. I am eager to complete the display panel, keyboard and buttons for the machine. If I can break through the decal problem, I should be able to finish those parts of the system in the next couple of weeks.

I am working with the creator of the decal process, we are painstakingly working through possible causes for the problems and hopefully will soon have the production process humming along. Right now it seems I have two issues impacting the process:

 1 - my heat gun is not getting the paper dry enough, thus it is shrinking while going through the laminator and causing wrinkles in my mylar coverings. It is a cheap unit I bought at Harbor Freight, so its failure to do its job is not surprising.

2 - the laminator is not applying enough heat and pressure to fully bond the clear mylar to the paper with the fused image, but when I took what came out of the laminator, wrapped it in a paper towel and gave it a couple of additional passes through the machine, it seemed to bond well enough to stay together in the water bath.

I can solve problem one by buying a better heat gun. Problem two may require me to buy a different laminator, a second-hand model of the type that this process was developed around. The model is no longer made, resulting in a substitute that is not working adequately for me. I am waiting for the creator of the process to come to the same conclusions.

No comments:

Post a Comment