Designing new type wheels for the Blickenserfer typewriter

It is possible today to create new type wheel designs for the Blickensderfer typewriter! With readily available software tools and common, low-cost 3D printing technology, functional type wheels (typewheels, typing elements) can be created.

To get started with creating own type wheels with a layout and typeface of choice, install the OpenSCAD modelling software. For better design-control and a much faster rendering, get the 'nightly build' version. Enable the text-metrics function in preferences and keep the fast 'manifold' rendering engine as default. (Much faster than the latest stable-release!)

Then get the set of design-files for new Blickensderfer type wheels. The zip file contains the main design script, a wheel layouts-file and several example type wheel configuration files. When all files are in the same folder (directory), it should all work with OpenSCAD.

The complexity of the 3D model is kept in the file "Typewheel-design-12.scad" that is used by the configuration files. This file should never have to be edited (or even opened). Note that the OpenSCAD language is a functional language, like Haskell. There are e.g. parameters, but not really variables - overall it's a bit different from 'regular' procedural or imperative languages. Because the type wheel design is completely defined by the parameters in a configuration file, this can fortunately all be ignored.

Opening one of the example configuration files in the OpenSCAD application, a preview of the type wheel is shown next to the opened text-file. 

OpenSCAD has a preview and a render for a 3D model. The preview (F5) is fast, but not geometrically 'sound' for generating a 3D printable STL file. It is meant for a preview when editing and configuring a design script. To generate a correct, 3D printable geometry, the model is rendered (F6). This rendering does all the correct calculations and will be much slower. The model is programmed so that characters show readable in preview, but correctly mirrored and shaped in render. The preview of a typewheel generally takes ~0.2 second, but a render may take up to a minute (depending on complexity of the typeface).

Naturally, the font used in the design must be available on the system. For the re-creation of an Italic wheel catalogue number 440, the italic variant of the TT2020 font was chosen. This is a widely available free font that can be downloaded and installed. Even for this straightforward type wheel, the caret (circumflex) is taken from another font - e.g. from Courier, Courier New or Courier 10 Pitch - also needs to be installed. For the catalogue-number, ideally Arial Rounded is available on the system. 

When all fonts are present and the preview is correct, then the render-function (F6) will generate a proper 3D model. This can be exported to an STL file to be printed and finished as described in an earlier post. Fitted with a newly-printed type wheel 440 R (Repro) from this configuration file, the Blickensderfer types in Italic.

The 3D model in the script is designed and optimised for FDM printing. Printed with a fine nozzle of 0.2 mm and a small layer-height of less than 0.1 mm, the result will be good enough to take the ink and make a credible imprint on paper.

The configuration files are relatively small, but still contain many options and parameters for tweaking the type wheel.

The showBoundary option toggles visual guides around every column of characters that indicate the space they have when typed on the Blickensderfer (with 10 characters per Inch). This is useful when configuring, shows when characters would overlap as typed.

The showNozzleComp option toggles the extra distortion of characters to compensate for the accuracy-limitation of an FDM printer.

This distortion adds extra width to character verticals to adjust for the diameter of the deposited filament. The ~0.22 diameter of the extruded 'string' limits the sharpness of outer radii, so there is a horizontal adjustment to the characters to compensate for this radius 'drop away'. It can be useful to show this distortion in preview, to see if e.g. details of characters are 'swamped' by this when rendered.

The layout can be chosen from one of the presets defined in the layouts file.

This file is very limited so far, but of course freely extendable to match the keyboard of any Blickensderfer. The arrangement in the file is identical to that shown in the catalogue; traversing the keyboard from top-left-half to mid-left via bottom-row to mid-right to top-right. Alternatively, the layout can be chosen from the custom set defined in the configuration file itself. This can be useful to make wheel-specific changes like e.g. using the letter 'O' instead of the numeral '0' for a particular font or to e.g. include symbols like € or ㋡ on a wheel.

Different styles of type wheel are supported, chosen with the wheelStyle parameter.

There are many parameters to modify the font. There is horizontal scaling, overall fattening (or thinning) and horizontal fattening only (to adjust for scaling). For selected characters, an extra scaling can be applied. Similarly, an alternative font can be selected for a sub-set of characters.

One of the peculiarities of the Blickensderfer is the caret character. This is rare in font-files, so the circumflex accent substitutes for this. Being a diacritic, this needs moving down onto the baseline; hence the caretDrop parameter.

In general, a clean fixed-width font will not need too many tweaks to work well. To however use a proportional font on the monospaced typewriter, many tweaks may be needed to get all characters to fit and not look too out-of-place. The preview shows the result of modifications, but needs to be manually refreshed after making edits to the file. Even a monospaced font may need extensive tweaks to get it just right, see for example the wheel 407 created with Courier New.

Making a typeface perfect for a typewheel really should be done by editing glyphs in the font-file - for example to create a "Pf." as a single character for a German typewheel. Nevertheless, the various tweaks possible in the configuration should allow for already a lot of corrections.

Most of the parameters have an explanatory comment, or at least a comment that was meant to be helpful in understanding what it does :)

The use of a fine nozzle of 0.2 mm enables printing with reasonably fine detail. It is good enough to re-create type wheels that perform very similar to original type wheels. A 0.2 mm nozzle is however small, and most FDM printers use a 0.4 mm nozzle as standard. Though less precise in printing, with a 0.4 mm nozzle it is still possible to create functional type wheels. The elements shown below were printed with a 0.4 mm nozzle.

Whilst less suitable for typefaces with fine detail, the resulting wheels are entirely functional.

The type wheel 3D-model was created especially for 3D printing on FDM-technology printers, the most common and hassle-free low-cost type of 3D printer. In principle, setting the nozzleDiam parameter to zero will create a 'clean' file that would also be suitable for printing on e.g. a resin (DLP) printer. The overhang-angles that work for FDM will certainly be good enough for DLP printing. However, because a resin-printed part is more brittle and fragile the current model may be too delicate in resin to survive actual typing. Additional changes to wall thicknesses may be essential to be able to 3D print a usable wheel in resin. These can all be tweaked in the design-file of course, the whole model is set-up parametric so it should not be too difficult to adapt.

 

An important caveat is that the critical parameters of baseline and platen-centre are best-guess estimates only! And these guesses are based on only a single typewriter! The values are 'weird', but seem to produce identical result to that from an original wheel. Please do not hesitate to share/comment/add extra experience or inputs for these key parameters.

Net result is that at least this Blickensderfer 7 now has a variety of typefaces to type with. In keeping with the originals from The Blickensderfer Mfg. Co. these new wheels are stored in wooden containers (small round boxes, widely sold as wedding-ring or jewelry boxes) with a re-created lookalike label.

With these files as starting point, it should hopefully be feasible to create more new type wheel designs for the Blickensderfer! 

More typing Blick's :-)

Messing-about making new Blickensderfer ink rolls

The original ink rolls of my Blickensderfer were hard and dried out. Whilst it is written on The Internet that these can be revived, decided instead to have a go at making new ink rolls. It is also written on The Internet that gun-cleaning felts of 7 mm diameter are a good substitute.

The local gun shop only had 6.5 mm in stock - so used those instead of 7 mm. (The original hardened rolls I have, are now between 7 and 8 mm 'diameter' and not quite round anymore. Given the close spacing to the platen and needing to clear the Paper Guide, I'd have expected 1/4" as likely diameter. But maybe the originals were 5/16" or some such fractional dimension. Until an original factory drawing or purchasing specification turns up, it'll probably remain a bit of a guess.)

Gun-barrel cleaning felts are too wide to fit the bracket on the machine, space between the bracket fork is around 9 mm. The hole in the pads is a bit 'vague' too, so started by soaking some of the felts in hot water. Left in the water for ~3 minutes for the wool to be properly wet.

The fully hydrated felt taken out and cut to length (or width) with a sharp knife. Wool (=hair =keratin) should be much easier to cut when fully hydrated. In retrospect, might also try simply cutting with large scissors.

After cutting to ~8mm width, the cylinders were pushed over a 2 mm knitting needle and left to dry. This makes a nice, clean 2 mm central hole in the felt cylinders.

Original ink rolls are a dense felt on a central brass tube. These new felt rolls already work fine on the machine without a central brass tube. The main thing is that the felt should not be too wide. If it doesn't touch the sides of the machine's bracket, it will rotate fine. As extra, a bit of glue (water-resistant type) can be spread round the inside of the hole and let fully set - this will prevent the hole from 'collapsing' and getting stuck on the pin.

However, in the interest of historical authenticity decided to add a brass centre. Acquired some brass tube of 2 mm diameter and 0.25 wall thickness - i.e. has ~1.5 mm internal diameter to fit nicely round the ~1.4 mm pin of the Ink Frame (i.e. the fork bracket that holds the ink roll).

To cut lengths of brass tube, a jig was made by gluing coffee-stirrer sticks to a scrap bit of wood. A channel to hold the tube and a stop-strip to easily set the length to saw off. 

This makes it relatively easy to quickly saw off multiple short tubes all with a length of just-under 9 mm. After sawing, of course file the ends clean and remove any burr.

Then inserted the tubes into the dried felt cylinders. Experimented with different glues; regular hobby-glue, PVA glue and even cyanoacrylate. Tweezers for the felt and the tube pushed over a toothpick to keep fingers well away. None of the options were ideal.

The felt cylinders themselves are perhaps not very 'robust' and are difficult to really fix on the tube. Another thing is that the felt pads swell when wetted - also when later wetted with ink. They increase in size and lose their shape a bit too. After some experimenting, it turned out to be helpful to apply some glue (PVA or cyanoacrylate) to the hole-inside and also over the side-walls of a felt-pad. After this has hardened, these 'infused areas' help keep the cylinder in shape. The only surface that needs to remain 'clean' is of course the outer cylinder-surface for inking the type wheel. (The better solution would of course be to get very dense felt.)

Tried inking and typing with both modern water-based stamp inks and an oil-based metal-stamp ink. The results were very clear; even apart from any concerns about ink attacking original vulcanite type wheels, the metal-stamp ink is unsuitable. It tends to 'splatter' during typing (even though it is very viscous). 

The water-based inks worked reasonably well, but probably have less colorant per volume than the original rolls ink. Re-inking an ink-roll a few times after it has 'dried-out' fixes that, effectively building up a denser ink.

To apply the ink, several drops were placed in a dish and the blank roll placed in the ink. To get the best saturation of the roll, it can be left an hour or two to let the ink slowly be pulled deep in the felt. Of course, rolling it over the ink will quickly apply ink to the outer layer of the roll.

From Blickensderfer literature of the period, it seems that the user would put a fresh ink roll on the machine fairly frequently. Ink rolls were sold in glass vials, 6 of them for 40 cents. As glass is a nice way to store inked rolls against drying out, some new 12 x 75 mm glass tubes with cork stoppers were found online to replicate these '6-packs'. (Re-inking a roll on the machine I cannot imagine as being a practical idea - mucking about with ink drops right over the typewriter itself.)

The ink rolls made with this method from the gun cleaning pads look decent; a good approximation of the original. They are however much less dense than the original, in fact they are fairly 'fluffy' and do not keep their shape well. The plus-side is that they're cheap and plentiful - easily made from relatively low-cost materials.

There are still functional improvements to be made; the descenders are not always inked properly. This may be due to the diameter (shouldn't matter, should it?), the fluffy surface or simply a mis-adjustment of the ink-roll arm on this particular typewriter. Or of course, might be that the type wheels need improving still. More to be found out.

The perfect way to make new ink-rolls is yet to be found. Nevertheless, these are quite usable already!

Manufacturing process for a new Blickensderfer type wheel 407

How to make a new type wheel for the Blickensderfer; in this case a reproduction 'Small Roman' type wheel catalogue number 407, codename 'Table'.

To start with, take a 3D-model of the type wheel. This 3D-model is designed for 3D printing in PLA material on an FDM printer. The overhang-angles and outer contour are designed for the FDM process with a small nozzle (0.2 mm) and best printed with a fine layer-height of ~0.080 mm. The model does not require any supports.

The 3D model can be downloaded here.

After printing, the wheel needs some finishing to be usable - the top 'stub' that holds the spring-clip is far too weak to withstand any use and the important surfaces will need fine-tuning. The tools needed for this are a round file (small, ~3mm diameter), cyanoacrylate glue, a toothpick and a bit of scrap card.

To strengthen the stub on top, place a drop of cyanoacrylate on the toothpick and bring it to where the stub meets the top-plane. Cyanoacrylate will fill remaining gaps between the deposited filament layers. Use the toothpick to bring the drop all-around the stub, also in the corners of the spring-clip notch. Use the scrap-card to absorb any excess and wipe clean/flat the cyanoacrylate over the top face of the wheel. Then give the part a few hours, say 10, to let the cyanoacrylate set - do not disturb during hardening.

Any stringing or z-seam artefacts on the bearing-surfaces in the central hole can then be made smooth with the round file. The mounting-rod on the typewriter is ~ 3.2 mm diameter, the type wheel should fit on this rod without any friction and ideally without play. Any friction when rotating on the rod will cause the Blickensderfer to 'jam' or cause the typing-head to not come back up completely after a keypress.

Any cyanoacrylate that got into the central hole can also be filed away after setting. In case a hole turns out to be too large, an even application of some cyanoacrylate with the toothpick round the inner-surface can again reduce the diameter a fraction of a mm. A thin application of cyanoacrylate on the inner-surface of the central hole may be good anyways - it gives it a very hard and low-friction surface compared to plain PLA.

In case the square take-up hole is too tight, it can be filed wider. Note that the actual dimension of a printed part can vary slightly per individual 3D printer and also per filament used (and the condition, how well dried etc.) 

The spring-clip that fixes the wheel in-position can be formed from spring-steel wire of 0.020 gauge (approximately 0.5mm diameter). E.g. plain-steel guitar-string of 0.020 gauge is a good source of a practical length of suitable wire.

A few nails driven in a scrap bit of wood serves as a jig to create several clips. As always, taking care with springy wire that has sharp ends. Small pliers help to bend a new clip into shape.

The spring-clip is simply snapped on the stub - and the new type wheel is ready to be used on the Blickensderfer! 

The machine can now be typed on without fear of damaging an original vulcanite type wheel, even with rock-hard platen. An extra backing-sheet is a good idea anyways, also helps with the quality of the impression.

* Note that when actually using a Blickensderfer, it quickly becomes obvious that type wheels get covered in ink - especially the top where the ink-roller is pushed to when the type hits the paper. This makes swapping typefaces without getting ink all over your fingers tricky. Best to use a patch of tissue or waxed-paper to handle a type wheel when putting it on or taking it off the typewriter -and to wrap it in when stored in its wooden box.

* Note that on original vulcanite type wheels, the central tube is brass. It is a brass tube of ~4 mm with a ~3.4 mm (?) hole that is the wheel's bearing/mounting on the machine. This reproduction is all-plastic, but of course a reproduction could be designed to take a brass tube with a machined slot for the holding-clip.

* Note that the alignment of characters is from their position on the type wheel, so should be good. However, any play of the type-wheel on the shaft or with the take-up vane will result in alignment deviations. Also a loose or badly-fitting spring-clip can cause alignment irregularities.

* Note that the FDM printing process is not as high-resolution as a DLP resin-printing process would be, but it is good enough for typing and the material is much stronger and impact-resistant than a resin print. Another benefit is that FDM prints are much easier/cheaper to make - less hassle with unpleasant chemicals and this part should take about 4.6 grammes of PLA. For FDM process, the PLA material is a good one - compared to e.g. ABS or PET it is easier and usually gives a superior surface finish.

* Note that the baselines of the three rows of text do not seem to have an immediately obvious spacing. It is not like e.g. the Mignon that is spaced in tenths of an Inch (i.e. 2/10ths between rows). Comparing test-wheels with the typing from original wheels, the three rows baselines were set at experimentally derived distances from the top-face of the wheel. The current values seemed to work fine, but may not be correct or optimal for all Blickensderfer machines.

Excavating (or destroying?) a layer of history on a machine

This particular Underwood 5 was bought locally, originally meant as a reference machine to help with the restoration of an older 1920 Underwood 5.

This machine was not only a newer 1928 version, it also showed evidence that it was refurbished and put to use in a later, very specific time-period in the British Isles. In archeological terms; it had acquired a 'layer of deposits'.

Taken off the machine and replaced with 'nicer' specimens. The knobs were replaced with Underwood knobs - albeit later, larger-diameter pattern. The green ruler and especially the wrong-pattern knob on the Underwood were 'jarring'.

The modifications are in principle all reversible; the removed parts are kept. But realistically, this machine viewed as a historical artefact has entered a new phase; 21st century collectable (though not rare). 

Sticking with the 'archeological' viewpoint - this removal of a layer has now, hereby been documented :-)

One mystery part - mildly disconcerting

After taking it apart and putting it back together, one spring that won't fit anywhere. 

This Blickensderfer 7 was also one key-lever spring short. The odd thing is, that the remaining spring will not work as a key-lever spring; it's the wrong shape and simply won't fit. In the end, a new replacement key-lever spring was formed out of piano-wire. This new spring is a great fit, in fact can't tell it apart from the originals, and all 28 levers are held up properly. But the left-over spring is a worry - no idea where it came from.

All 28 key-levers have their spring to keep them in the up position. In the middle of the machine then the two springs that work the left and right rotation-bars, these springs have a proper loop to go on studs. And finally there is one thicker spring to go on the universal bar.

Couldn't fit the mystery spring on any key-lever, just couldn't make it fit. 

Maybe later I'll discover how it can fit a special key-lever that needs it - maybe it simply was a 'wrong' replacement during servicing long ago. Or a 'whim' at assembly of the machine, even longer ago.

The other thing is that this Blickensderfer works flawlessly without this mystery spring. 

mildly disconcerting

:)

Swapping-out cellulose acetate keys on a Comptometer

To start, five columns of white keys for a Comptometer.

These are genuine, old keys. Not new reproductions, but relatively good originals. They do have some yellowing and have started to deform, but still very serviceable. These keys were harvested from a ~1930 Comptometer Model J with severe rust and -oddly- loss of many of the green keys.

 

Usually it is the white keys that degrade instead of the green. That's because the white keys did not have enough colorant in them to act as antacid to halt the decay of the cellulose acetate.

This very common 'key-rot' on Comptometer Model J machines was introduced by changing the keys material from cellulose-nitrate to cellulose-acetate. It would have seemed a sensible idea to change to this new material - just as film moved from cellulose-nitrate to cellulose-acetate (i.e. safety-film; not as flammable as the nitrate!). 

This change could have been fine. However, the unfortunate (in hindsight) choice was to also change from the completely opaque white nitrate keys to a slightly translucent white. This likely looked great and 'gem-like' when new and perhaps made the keys a bit cheaper, but with too little zinc-oxide the cellulose-acetate mass remained slightly acid. Then the acetate starts to degrade, adding more acetic acid (vinegar) to stimulate further decay. The acetic acid can even 'sweat' out of the key and then mingles with the (wax?) lettering, making it look like the lettering just oozes off the key. It also makes them sticky (yuck).

The green keys on Model J machines are fully opaque; i.e. seem to have much more pigment and these are generally fine. It is quite unusual for green keys to also have crumbled. The white keys all being fine on this machine suggests that these were already a replacement set, probably fitted sometime in the 1940s.

The replacing of keys requires removal of the entire key-stem. Removing keys-with-stem from a Comptometer in the usual manner. Then using a slotted wood-block and hammer to tap the key from the stem, as it is held in a vise. 

Replaced the keys one column at a time - not removing all stems from a column at once, to make re-fitting a bit easier. (Otherwise it can be tricky to ensure all the levers and rods go at their correct side of the stem.)

The right-most column with the new keys, the contrast with the raisin-like old keys is noticeable. After replacing the keytops on all the white (well, cream) columns, the Comptometer Supertotaliser looks much better.

At then end of this procedure and having done another machine earlier; we now have a jumble box of Comptometer keys in varying states of degradation. From almost ivory white to deep-brown.

Two of the green keys on the machine (circled in red) are already shrunk more than the rest; so a hunt through the box to find a potential replacement. There actually is a quite a bit of color variation in the green; the best quality 4-key is a distinctly different shade of green (dotted circle). For now, all green keys were left on the machine In case the keys really do fail in the future, there are some ok-replacements (circled white) in the box. 

As a final small step, the large engraved numbers '6' were removed from the typeplate. Or rather, the rusty keytop-donor Model J had exactly the same version of typeplate, so these were exchanged.

(Note that replacing typeplates on these machines carries a risk - the screws go into a threaded washer that is lodged in the cork-lining of the top-plate. If this washer should be pushed loose or lets go, it drops away into the mechanism and all keys need to be removed to take-off the top-plate to recover the washer and be able to re-fit the typeplate.)