Re-keying a Comptometer model J

Another, second Comptometer got a full new keyboard. 

This one is a common model J that was bought as a parts machine - it was severely rusted (all keystems rusted red) and half the columns didn't work. 

But it had good white keys, so these were all harvested to refurbish the Supertotalizer. Having taken its keys, couldn't quite bring myself to 'bin it' - so made an attempt to get it working again.

As often with these machines, deep cleaning of some rods and levers internally got it all working again. All keystems taken out and refreshed (steelwool!) and remaining green keys removed made it look much better - then it needed a full set of 90 new keytops.

The very labour-intensive method for making keys as used before on a model H was too much for this J. Besides, the keys on a model J should not be fully opaque as on a model H. Especially the white cellulose-acetate keys of the J are a bit semi-translucent. So, no expensive resin-printing and lacquering+polishing, but a simple FDM print for this machine.

New 3D models were made, this time suitable for less-hassle FDM printing. Keytop legends were copied from some remaining good specimens. A modern ' hobby printer' can then make pretty good parts, even with a regular 0.4 mm nozzle. The STL files can be downloaded from Printables, with extra hints on print settings.

To do some simple visual smoothing, the keys were given 3 or 4 layers of clear varnish. That also stops the lettering-paint creeping between layers when filling the lettering.

Latex wall-paint again was used, filling the lettering (well, numbering...) - letting it dry, then any spillover on the keytop can be wiped off with a damp cloth. Hold the cloth taut, flat - then it will only soften and take off the paint on the key surface and not draw it out of the recessed lettering.

The white and the green plastic is a bit 'bright' still, but the documents of the time do state that the Comptometer had green-white keys, not green-cream! 

Several columns were by the way printed using a raft, which isn't really needed - no rafts gives a cleaner base. Nevertheless; result was a full set of 10 columns of new, reproduction keys in PLA plastic.

Fitted on the machine and comparing side-by-side with aged originals on a 1930s white-green Comptometer, the keys do look very bright - yet this might be what these machines looked like when new.

The new 3D models and method of manufacture really lower the cost for making new keys for a Comptometer. In case one or two keys need replacing, a printed part can of course be painted to be a good colour-match with the rest of the original keys. In case a full keyboard is replaced, this is less critical and ease of manufacture becomes key. (pun!)

Result is that we now have yet another Comptometer; a very common model J - with uncommonly bright keys :-)

Deciphering a hay tedder - Meccano double puzzle

This one image with a parts list is the instructions for a Hooi Keer Machine, or a hay tedder. Here shown in a 1924 Dutch edition of the Meccano Book 1 of instructions. This model first appeared in 1916 and was replaced by an improved, updated model in 1928.

In style it's a typical early Meccano model; very spidery and 'schematic'. And of course a very period subject, modern horse-drawn farm machinery. Not really knowing offhand in 2025 what a real 1916 hay tedder looked like makes such a small model even more of a puzzle.

It turns out that the right-most (in picture) pulleys are not wheels, but merely end-pieces to hold the axle in. It's a two-wheel contraption (as most tedders of the period were), where one wheel drives a rotary 'rake'-wheel on the right-most rod via a driving band (red). When moving the model the rotary 'rake' really does turn faster than the wheel, actually gives a good illustration of the working principle of a (rotary) hay tedder.

The two bars could suggest a push-cart, but the seat on top I think makes it clear this was meant as a model of a horse-drawn machine.

After 100+ years these small relaxing puzzles have gained an extra dimension; not only how-to-construct, but also 'what-is-it?' :-)

A 1975 vintage Olympia CD43

Unexpected thrift-store find!

In suitably coloured plastic, a 1975 Olympia CD43.

Spotted by our youngest, who scans the bins of 'button-boxes' for anything cool - this rated cool enough, certainly for the few Euro asking price. Batteries included - and it works :)

The vacuum fluorescent (VFD) display glows crisply green - in a bezel that is a little higher than the keyboard.

This small, simple four-function calculator was manufactured for Olympia in Japan. Noting that although it is a basic calculator with the four functions, in 1975 this was an advanced and relatively expensive item. It works on two AA batteries, but also has a power-jack for an external power-supply.

The arrival of low-cost electronic calculators in 1975 - 1976 spelled the rapid end of mechanical calculators and sliderules alike. This Olympia is reportedly based on the NEC µPD941C integrated circuit; a dual-inline 28 pin package.

This Olympia electronic calculator is in a way a direct descendent of the first commercially succesfull pinwheel calculator, the Odhner of the late 19th century. Setting up a German manufacturing / licensing of the Odhner patents was the basis of the Brunsviga line of calculators in the 1890s and Brunsviga was absorbed by Olympia in the late 1950s.

Showing how it compares; the usual simple division of 355

by 113

gives the answer immediately:

That really shows how much the little electronic calculator was superior to mechanical four-species machines (such as e.g. this Olympia RT4).

This thrift store find came complete with its vinyl protective sleeve.

That sleeve and it probably being kept in a drawer for likely 40-odd years away from direct sunlight is what probably helped keep the beige plastic relatively bright. It will continue to be kept by our youngest out of sunlight in its sleeve :)

Updated typewheels for the Blickensderfer typewriter - Italic English

New, improved 3D models of English Italic typewheels - catalogue number 440 (scientific) and 497 (universal). The STL's for both are available for download on Printables.

From using 3D-printed typewheels on the Blickensderfer and gaining more experience from printing these designs, several small improvements were made to the model. These Italic typewheels should now print more reliably (on the 3D printer) and also print more reliably (on the typewriter).

New felt under the Blickensderfer 7 base

The Blickensderfer 7 typewriter is screwed to an oak base-board. On the bottom of this base-board now again a thin sheet of ochre wool-felt.

This new sheet was cut to size from a sheet of 1mm thick wool-felt and glued under the base.  For glue, used a glue-stick ("Pritt"); i.e. mostly starch - so not too historically inappropriate.

It may have been a later modification by the owner, but I suspect this style of case originally had felt. Similar to e.g. the wooden base-boards of a Hammond; Hammonds have green felt, this Blickensderfer had ochre. Perhaps many/most Blickensderfer cases lost their felt over the past century or more; haven't been able to find a lot of online evidence of others with felt. Only one spotted so far, with a dark blue felt on a dark oak case. (The bottom of the base is of course not the most obvious angle to photograph a Blickensderfer typewriter, so that won't help either :)

This particular base still had some remnants of felt attached; these remaining areas perhaps had been re-glued some time in the past with a better glue than the original. The remaining patches were very hard to remove, whereas the rest of the base-board showed no signs of any glue ever having been there. The original glue probably dried-out, lost all adhesive power.

Whatever the case is; these remnants allowed an attempt at color-matching:

Close enough - another small step in the re-furbishing of the Blickensderfer 7. (Does anyone else know of a Blickensderfer case with felt under the base-board? And what colour?)

Repairing fault-handling of the Burroughs Portable adding machine

With the subtract-function fixed and columns 7 & 8 re-activated, this Burroughs Portable performed all calculating functions flawlessly. What however did not work at all, was the catching of faulty, incorrect input conditions. (Because this doesn't impact the normal functioning of the machine, this a 'malfunction' that can remain completely unnoticed. The machine did not dis-engage and simply completed a cycle, with possibly incorrect results.)

This now also repaired and re-activated.

Looking back, this functionality was probably broken from a combination of wear to parts and from parts getting bent by handling without its case/base. And made worse from my injudicious handling of the keyboard-module during cleaning - learnt for a next time that the exposed brackets are 'soft' and critical to timing.

Most adding machines have interlocks to prevent wrong operation that could cause incorrect results or damage to the machine. E.g. the keyboard is blocked during a cycle or it's made impossible to reverse half-way through a cycle. The Burroughs Portable has plenty of such safety interlocks. It however also has two fault conditions that are not blocked, but that are prevented by dis-engaging the handle and in that way prevent completing a cycle.

The two conditions that should trigger the handle dis-engaging are trying to start a cycle with:

• Depressed total or sub-total and a value set on the keyboard.
• Depressed total or sub-total and the machine set to subtraction.

Reading user instructions confirmed that the machine should catch these fault conditionsa and then reset. (Scans of these original instructions very kindly provided by Mike Hancock who created the very informative Burroughs Info website.) 

Reading further in the very extensive US patent 1,853,050 by Mr Horton covering the Portable gave more background on these two exceptions, how they are caught and what should then happen. The two conditions have their own specific 'trigger' that both then lead into activating the same dis-engage & reset mechanism.

To start with, the mechanism to dis-engage the handle from the mechanism is shown in Fig. 3A of the patent (here below). This is a detail of the left-side of the machine shown in Fig. 3 that is mostly dealing with the electric drive arrangement. The axle actuated by the handle (91, orange) is inside another hollow axle that drives the machine. (Actually, it pulls the machine via springs with the dashpot limiting the speed; a lot of thought & experience went into this design.)

To give the short (...) summary: The handle drives the rod 91 that has mounted on it the hook 89, this pulls along the machine's mechanism by contact-surface 87 onto hook 83 that is fitted to the machine's mechanism. At the same time, the rod 191 moves rearward (locks the keyboard among other things), the stud 194 moving in the slot 210 then rotates the extra hook 212 out of the way. That's normal operation. When an exception is detected, the rod 191 does not move all the way rearward and the hook 212 is left in the path of face 86. This then smartly pulls hook 83 away from 89 and the handle is dis-engaged from the mechanism.

In an actual machine, without electric motor, the relevant parts can be seen much clearer.

This dis-engaging should happen after the keys-reset lever is dropped behind its position on rod 191 on the right side of the machine AND it should happen before the pawl of the full-cycle mechanism clicks in its first stop (red arrow) on its cam 76 inside the machine. Then the mechanism will jump back to its rest-position and clear all keys. Simply moving the lever back to its start position will re-engage and the machine is ready again.

Probably due to wear, the dis-engaging happened AFTER the first full-cycle stop clicked in - that meant that a fault-condition would truly disable the machine. It would need opening up and manually resetting, continuing a cycle or forcing a clearing via manipulating by hand rod 191. (This wear and fault may have been a reason long ago already for the machine to be 'retired' and may also be why the baseplate of the machine was loose. It was not fitted to the mechanism, the machine sitting loose on top of the 'pan' so easy access to the parts that need twiddling.)

To compensate wear and move the timing of the dis-engaging before the full-cycle stop, a shim was added onto face 212. A bit of a kludge and too large, but this way the 3mm brass clamps itself on the back of the pawl. Fixated with cyanoacrylate, it only needs to keep it 'still' and the glue does not have to take any force. It also makes it reversible; applying heat (soldering iron) will disintegrate the glue.

The tip of the brass 'shim' was filed down to about 0.5 mm to slightly advance the timing. Another tweak was adding a bush around stud 85 (not visible itself, it's peened into the oval covering of part 87), this pushes out the hook 86 a bit further and makes it easier to dis-engage.

Brass is really too soft for this kind of function, but great for filing to shape when experimenting. In any case it's a 'kludge' and should not see much action - it is exception handling for faulty use.

The more complex side of the mechanism to detect the wrong setting is on the right side; that's where all the control mechanism is. That's where the push-bar is that is pushed forward a smidgeon further if total and a key are pressed at the same time. The keys detection goes via the U-section bail at the top of the keyboard to that push-bar.

That bail (in green circle) is the stop for a push-bar 185 that is pressed downward by the total or sub-total keys, changing its stop-position. This push-bar has an opening around the rod 191, thus limiting the travel of 191 and thus triggering the dis-engaging. (All this took a while to figure out...) With the dis-engaging mechanism again working, this fault condition triggered correct again.

The dis-engaging when a totals is depressed and the machine is in subtract still did not work - the handle did not dis-engage and a cycle was completed (with possibly wrong results).

The detecting of subtract-mode and a total key is a different mechanism to trigger the same dis-engaging mechanism. To make it interesting; the actual machine is different from the patent drawing in just this detail! Other than this subtract-lever detail, the machine is exactly as shown and described in the patent:

Instead of the forked extension to the front of the subtract-lever around rod 191, there is a separate part slotted over rods 191 and 122 with an upward forked-extension that interacts with a stud on the subtract-lever. During a cycle, the rod 191 is swiveled to the rear, raising this upward extension into the path of the subtract-lever's stud; this prevents operating the lever during a cycle - a safety interlock.

This part also plays a role in the detection of the fault-condition - that is done by preventing the subtract-lever reaching its end-position. When a totals is depressed this pushes forward a tab 223A about a sixteenth of an inch, and this tab blocks the subtract lever from reaching its end-position.

This sixteenth of an inch (well, approximately) should make the difference between the stud of the subtract lever sliding past the upward-fork or catching and stopping it. This no longer worked, probably because I accidentally (clumsily) bent the rod 191 assembly during cleaning of the keyboard module - the brackets may all be fairly heavy-gauge, but are soft. To repair this, applied another kludge; an angle-piece shaped from a steel nail was added to the side of the forked-extension. This now makes it catch on the stud when a totals key is depressed in subtract-mode - and move free when no totals key is depressed.

The left image shows it clearing the stud, the right image shows it catching on the stud (and triggering the dis-engagement of the handle, the machine jumping back to start and clearing all keys).

Overall, detecting these 'fails' and finding out how they were supposed work (and then figuring out how to repair) gave extra insight into the whole working of the Burroughs Portable adding machine. With its fairly modular build-up it is not the most difficult or complex of adding machines, but still (or because of that) contains a remarkable amount of ingenuity!

The patent is perhaps a bit unusual, in that a single patent document contains the full description of a whole machine - more often there are multiple patents documenting different features (inventions) of a machine. Absent a full repair or service manual for the Burroughs Portable of the 1920s, this patent is a pretty good resource. To give an impression of the patent; it contains a full 20 pages of detailed diagrams, followed by 28 pages of terse technical explanation.

To note though that there are some detail-differences with actual machines (e.g. the subtract-lever) and it was not written for a repairman. It e.g. glosses over some details that will be important when servicing. For example, what it states is true; the keyboard can be lifted off after removing two screws and sliding it rearward. But you also must unhook two springs and use a finger to keep rod 191 pulled all the way forward to take it off without breaking anything! (I've taken the keyboard off about a dozen times by now; that little spring that holds the reset-push-bar is easy to miss.)

These Burroughs Portable adding machines are pretty common and not all that sought after today. The reason these are common is I think not only because Burroughs had a great sales organisation, but also because it was a very good, low-cost and fit-for-purpose design - it sold in large quantities:

Office Appliances, November 1926

Advanced digital computing technology of the 1920s!

For the mechanically/technically interested, these are absolutely worthwhile machines - and repairable too :-)

Type cleaning on the Burroughs Portable adding machine

Just like typewriters, the type on an adding machine can get clogged-up with old ink or dirt from the ribbon. The Burroughs, like most adding machines, has a rising-segment printing mechanism; the trick is to first get good access to the type for cleaning.

On the Burroughs it is relatively easy to remove the platen, by taking off the top-assembly. Removing 3 or 4 screws should allow this assembly of platen and ribbon-spools to lift off. To then make the segments pop-up, enter 9's in all columns and stop mid-cycle.

Before pulling the handle, it's a good idea to set the machine to non-printing; without a platen present the printing hammers would only try to knock the segments out of their guides. I.e. slide the little tab to the left:

Then pull the handle forward, but not all the way. The full-stroke mechanism has several 'clicks' that lock the handle and prevent it going back before completing a full stroke. Leave the handle at the last  'click' before completing a stroke. This will leave the handle in the forward position and the type segments raised to their full extent.

The housing doesn't need to come off, but could be easily lifted off after removing the two screws in the front.

Cleaning of type is exactly the same as with typewriter slugs; a sharp needle to pick-out the caked-up areas and a brush to clean the type. A rag to prevent the dirt dropping out of the type falling into the mechanism.

After cleaning, the stroke of the lever can be completed - and the segments drop back to their rest position again. The machine is then re-assembled, i.e the platen and spools module placed back - followed by a quick test of all numbers on all columns. The rock-hard platen is now probably the limiting factor for print-quality - nevertheless, all clean and readable :)