Comptometer Model B - restored

Finishing touch now added to the restored Comptometer Model B from about 1908.

The fixed-up machine still had two replacement keys from the 1920s (amongst many other issues). These keys were noticeably different in design appearance from the original composite keys. They are also a slightly different shape and size. The '3' keys in columns 3 and 4 in image below are these later, celluloid replacements:

After some hesitation, these two old replacements were taken off and replaced by new reproductions of the original key design. The celluloid keys were themselves already a century old and could/should be considered part of the machine's history, an acquired 'layer' of history in archeological terms. However, having already re-painted the case I decided to prioritise the overall appearance of the machine; to optimise for looking as when originally sold and first used in 1908.

(Taking off a key means taking out the complete keystem and then using the slotted wooden bar to 'hammer' the key off. Attempting to pull off a key when the stem is in the machine will almost certainly damage/destroy the mechanism.)

The earlier composite-keys are slightly different in shape from the later celluloid keys, a bit taller and a different lower section - from a different molding process and factory. The lettering is also slightly different, thinner and taller. The lettering on Comptometer keys however also varies a bit over time, per different mold or die that was made. From measurements of the genuine specimen on the machine, a 3D model was made of the composite-key design.

The new replacements then 3D printed from PLA plastic and finished with paints to mimic the appearance of the composite keys - blending in well enough to not be noticable.

The mechanism of this model B still functions flawlessly. In above picture the machine is showing the result of dividing 355 by 113; answer 3.14159 with remainder of 033.

The Comptometer Model B now fully restored - calculating like it's 1908 :-)

An Olympia in a Broadway shop window

During a saunter round Leiden, spotted a grey Olympia SM typewriter in the shop window of a second-hand store.

Rather an up-market thrift store - such a thing exists, surprisingly. It seemed to sit somewhere on the scale between junk-store/thrift and an antique-shop. Located in the city on the Breestraat (i.e. Broadway) with a mid-century modern flair in its display.

The photograph of Mr Rühmann probably also dates from the 1950s. It was after closing time, so no safari of the insides of an up-market thrift store - just the window display :)

Blickensderfer carriage escapement adjustment

The reason for this hole in the base casting of the Blcikensderfer is now clear.

It gives visibility of the meshing of the indexing-claw with the carriage rack. This moving claw in the rest-position is held against a stop-piece that can (must) be adjusted. The stop-piece is fixed with two screws, accessible only from above, so not when the carriage is mounted on the machine.

When this stop-piece is not quite right, there will be a bit of play in the carriage position. When at rest, it will then be possible to manually shift it about a bit.

Because the stop-piece cannot be adjusted with the carriage on the machine, it is a trial-and-error iterative process today. The peek-hole shows how the stop-piece still needs to be moved. The factory almost certainly will have had an empty carriage-strip with access-holes to tighten the screws with a rack in-position.
.

The second adjustment is the static claw. It is held on an eccentric screw that permits left-right adjusting. It should be placed so, that when the moving claw is pulled out of engagement, the carriage cannot be pushed to the right - the static-claw's function is to prevent any pulling-back by the moving claw when indexing the carriage.

With the two claws now adjusted, the Blickensderfer typewriter somehow feels a bit more ' solid'. When the escapement is out of adjustment it will still index and type just fine, but will have a bit more rattle and may have irregular kerning.

Increasingly, I'm amazed how well the little Blick can be made to work; and with increased curiosity for the long-lost factory documentation on manufacture and adjustment :)

Blickensderfer platen knob replica

The re-creating of a platen for the Blickensderfer typewriter slowed down, but a small improvement in the surface finish of the sleeve was tried. 

After filing away more thoroughly the z-seam bulge and a general roughening of the entire surface (fine sandpaper), the surface was painted with a water-based acrylic. The paint was simply brushed-on and left to dry for about 10 or 15 minutes. After the paint was dry (but not hard), rubbed with a cloth/rag to make a smooth surface.

The above image shows the newly manufactured reproduction (top) next to the original ~1905 platen. The repro surface still shows the seam of the printing process, but the surface is smooth and it's less noticeable visually than it was. The thin layer of acrylic should not have any real effect on the resilience of the platen and it does help to even-out the surface. Result of this method is that it looks good and types fine, but it will have to be seen how durable a surface this is. Trial and error :)

Also, an improved replica of the platen knob was made. Because the knob needs to be tightened quite hard on the rod, it really did need a threaded metal insert. In this case, just as in the original, a hex-nut.

For fitting the replica knob on an original platen-rod, a UNF 8-36 nut should be fitted. If making one for a new M4-based platen, there's a model for an M4 nut too. Both versions can be downloaded here.

The 3D model is ideally printed with a small layer-height. For nicest results, of course a small nozzle will create a crisper profile and enable smallest layering.

Enable supports (tree) and set the support threshold to e.g. 20 degrees to avoid annoying supports in the central hole.

To finish the knob, insert a nut and fixate with a small drop of e.g. cyanoacrylate glue on a corner. With a few applications of paint and then light sanding, the surface can be made to imitate the surface of an original. (Pedestal-stand helps with painting.)

A glut of knobs! Of course only one needed per Blick - here on the Blickensderfer 5.

Blickensderfer platen testing, hardnesses

The platen of a Blickensderfer is a bit different from most typewriter platens, in that it has threaded rod-ends instead of a through platen rod. Also unusual is that its rubber surfacing is edged with metal side plates and it has a recessed diameter section for the line-ruler to run in.

The platen of a Blickensderfer is invariably rock hard. Even though the metal end plates and the recess make it a bit more tricky, Blick platens can be professionally re-surfaced for sure. Before perhaps doing that, first wanted to do some testing with new reproduction platens.

Wanted to find out if platen hardness has a noticeable impact on writing and if so; what hardness works well for the Blickensderfer. (The Blickensderfer with 'pressing down' of a Vulcanite type onto the paper is a different process than a typebar slamming a metal type onto the paper -and may have different requirements on platen.)

Using steel threaded rod as the basis, a new replacement platen construction was designed to allow different surfaces/sleeves to be tested. The two componenst that make up the 'core' are 3D printed in PLA - that is a very strong material at room temperature. Brass bushings cut from 4.5 mm OD by 4 mm ID tube form the bearings. For the threaded rod, M4 was used. It should be American UNF 8-36 thread; this is still common and widely available in the US, but simply cannot be had elsewhere. M4 is a close match and will take an 8-36 nut.  (It is UNF 8-36 and not UNC 8-32 as had thought! Thanks GC for alerting me!)

The two component parts of the core, the main (green) and cap (pink), have threaded sections near the sides. The thread could be printed, but cutting makes for a cleaner and stronger thread in the plastic.

The threaded rod length is 268 mm with a hole of about 2 mm diameter drilled through at 27 mm from one end. This rod is screwed into the main core part to stick out from the flange-side about 8mm. A brass bush is then inserted, leaving a length of 3 mm sticking out. A small drop of cyanoacrylate glue on where the bush goes into the core will fix it in place.

The smaller core part, the 'cap', is also tapped and the brass bush inserted, leaving about 5.5 mm sticking out. With the sleeve and a nail, the new platen can be assembled.

The typing surface is a sleeve of rubbery material, i.e. a 3D printed tube from TPU of varying grades. The nail through the rod is used to positively lock the rod against rotating in the core. This way the plastic thread only needs to lock against sliding left-right and will not be loaded from rotating the platen. (Normal paper feeding could be fine, but e.g. when straightening paper by rotating the platen whilst holding the sheet the load on the rod is much larger than what a plastic thread could take.)

The brass bushes form the bearings of the platen in the carriage brackets (holes of about 4.8 mm). The core is held only at the threaded sections at both ends. Over most of the platen width, the threaded rod is free inside a larger hole, only loaded near the ends. In section mid-width, the new reproduction platen looks like this:

When the sleeve is over the core, the nail is locked in place. The right-hand 'cap' fixes everything in place with friction on the ends of the rubber sleeve. Everything can still be taken apart again. However, to eliminate any remaining rotation-play between rod and core, a drop of cyanoacrylate can be dropped into the nail-hole. The rod is then permanently fixed in the main core-component, but sleeves can still be exchanged.

Using the above construction a set of new, replacement Blickensderfer platens was made:

From left to right
 - an original rock-hard platen,
 - new TPU Shore 98
 - new TPU Shore 95
 - new TPU Shore 90
 - new saturated cork

Note that even though the TPU rubber platens look credible (from a distance), they require a lot of work to remove or lessen printing artefacts. There are of course the lines from the stacked layers of material, with 0.15 mm pitch this is not too bad and easy to mitigate.

Flexing is however a bigger issue. Because the material is rubbery, the part starts to flex under the printing-head drag as it gets taller. This means that towards one end the surface becomes less defined - it is rough, wavey. Another hard-to-avoid artefact is a bulge from the z-step of the printer, this will need to be filed away. (TPU does not like filing.)

After the artefacts have been mitigated a surface layer of black paint is applied plus a fine sanding to finish. This way a reasonably smooth, cylindrical surface can be achieved. It is nowhere near the quality of a professionally re-surfaced platen, but good enough for a typing surface -and testing.

Exchanging the platen on a Blickensderfer is unfortunately a bit involved. The left carriage bracket has to be taken off and the linespacing mechanism on the right-side of the platen also has to be removed. There are also washers to be placed back in the right order. (Note that one washer between right-flange of the platen and the carriage bracket.)

Testing the new platens, the results varied.

The cork platen worked, but only when taking care and typing very lightly. A press that is only a little too hard would cause adjacent characters to smudge the paper. It is too soft for typing, even on a Blickensderfer (as others have noted too). Making a platen with cork probably is possible, but will require extra infusing of the cork and stiffening of the outer surface. 

The Shore 98 platen is almost as hard as the original. It may be TPU, but there is hardly any trace of resilience left. Finger-nail test fail. Also in writing with it on the machine there is no difference with the original rock-hard rubber platen.

The Shore 95 platen feels very stiff, but definitely is resilient. The Blickensderfer types quite well with the 95 platen, a noticable difference with the hard, original platen!

Also with the more legible typeface of a wheel 407 it types quite well. And less worry about the wear on an original typewheel, because the surface does conform. I.e. it passes the finger-nail test.

The Shore 90 platen felt very rubbery resilient, yet still firm. It is clearly softer than the 95 platen, just going by finger-nail measuring. The typing on this platen however was much less distinct than on the Shore 95 surface. That was not expected, this nicely rubbery surface was expected to be best for writing with the Blickensderfer.

Quickly swapping back to the Shore 95 TPU platen, this looks the most promising sleeve material hardness.

Outcome of the testing, answers to the questions!;
  - yes, new rubber does make a difference for typing. 
  - Shore 95 TPU is a good hardness.

There are still many details to be worked out on e.g. exact diameter of the sleeve, thickness of the sleeve itself and method of surface finishing. Things can be further optimized, e.g. the clunky nail-locking could perhaps be re-designed with captive steel nuts to take the torque. Another new platen knob with proper M4 thread perhaps to be made. And perhaps invest in importing a UNF 8-36 threaded rod all the way across the ocean :-)

Regardless, already a better typing Blickensderfer! 

Blickensderfer tweaking and starting on platens

The Blickensderfer 5 typewriter is a brilliant machine. It also has an almost endless set of adjustments to make it work. One of the issues that remained was an imperfect printing of the 'z' character. This I'd attributed to some remaining tuning of the typewheel rotation-stops or something wrong with my new typewheels.

Not so.

When the letters 'z' or 'j' do not print well (i.e. the most-rotated characters), then the ink-roller needs to be adjusted down. Imperfect or faint printing was caused by the ink-roller being too high; the roller had already passed over the wheel when the rotation to the correct letter had not been complete. I.e. the 'z' as a largest-rotation character would not get properly inked. 

The ink-roller arm is fortunately easily moved - one screw (the obvious one in above photo) holds the bracket and it can be moved up or down to find a spot where it all works best.

Similarly the forked bracket that holds down the paper has an adjustment; a screw on the front-face of the right 'leg' of the top-casting. When the fork is adjusted so that it hovers horizontally just above the sheet, it works wonders in keeping the sheet in place to get clean printing! (Otherwise the paper can lift and characters get smudged a bit until the sheet is held under the rear bail.)

These adjustments were found during first testing of alternative, new platens for the Blickensderfer. The original is rock-hard, as they almost always are. Don't yet dare to send it off for re-covering professionally, so first trying to cobble-up some test-platens to see if it can make a difference. And to find out what hardness a Blickensderfer likes.

This above line was typed using a fairly crudely made cork platen. The cork sourced from cutting up a 'placemat' and wound round a newly made core. The cork itself is much too soft, so first spray-lacquered. The cork absorbs the first few lacquer aplications, adding strength and stiffness. Then a layer of PVA 'painted' over the surface to add more stiffness and close the surface, to fill remaining 'potholes'. The whole surface then lightly sanded with very fine grit. This brought the cork-platen into the workable range of hardness, and not too grippy.

It works!, but still on the soft side of the practical working window. More test platens to be constructed and more methods of cobbling together a resilient platen surface to be tried (without a lathe :)

Project! :-D

Marchant Pony B calculator typeplate reproduction

The Pony B calculator was missing the distinctive Marchant brass nameplate.  As this is a very visible part of the 'look' of a Marchant of that period, a reproduction was attempted.

Even though not an exact replica, a good lookalike nameplate could be drawn. The recreated design then to be implemented in relief. This should of course be brass (etching?), but noit having the chemicals and kit for etching some simpler methods were tried.

The Old English typeface is not an exact match for the originals. From browsing Marchant advertisiments of the period and some other publications however, the company itself was also not that strict on the exact typeface for the wordmark.

A while ago, had tried pressing soft paperboard into a mold to create a textured surface. This had worked well to re-create the missing sides of the pouch for a Golden Gem adding machine.

With the image flipped mirror-image, a mold was 3D printed. Half mm paperboard was soaked in hot water and then pressed against the mold. A clamp and bits of wood to distribute the force.

Left for several hours to allow the paperboard time to dry via the sides. Then opening the mold, there is a fairly crisp impression of the mold. It even showed the 3D printing artefacts; the diagoinal lines of the filament-path.

Experimenting with the 3D printing parameters/orientation and some sanding of the molds, the artefacts can be reduced. The cards are immediately fixated with spray-lacquer when they are taken out of the mold, dried. The paperboard is best placed on a curved surface (jam-jar) held in palce with masking tape. The lacquer permeates the paperboard, making a surprisingly stiff and strong part. 

After first fixating, the parts are cut to size and the four riveting holes are punched out. As final make-believe step, they are given a brassy finish.

The results with this method were very crisp, however the relief fairly shallow. The photographs of genuine Marchant machines show that the lettering is quite deep. The paperboard is also prone to warping.

To try a different method to get deeper lettering, also direct 3D printing of the typeplate was tried. The part is  printed flat and then placed over a curved surface (that jam-jar). With a hairdryer they are then softened and settled in the curve needed. A few variations were tested, with different printing parameters and thicknesses. The part then finished with a brassy finish.

These plastic direct-printed typeplates looked very similar to the originals in pictures found online.

A couple of these plates were made - these use almost no material, are very fast prints - and finished with different hues of 'brass'.

Some judging next to the genuinely brass dealer-plate to select a colour that is not too jarring on the machine.

Then fitted on the cover. Not with rivets, but with tiny hex-nuts and domed-bolts.

Very bright. It won't tarnish like the dealer-plate, so later it may be swapped for a darker fake-brass to stay in-range with the dealer-plate.

Another possible method could be to press thin brass sheet (possibly with paperboard backing) into the mold - that would give the correct material. Unsure if that'd give the desired depth of lettering, even with proper glowing of the brass. Perhaps something for a next project.

The Pony B may not work all that great, but it is a 'decorative' machine - on display for now :-)