Marchant Pony B calculator continued tinkering

After refurbishing a Model B Comptometer and playing with the Blickensderfer, returning to the Marchant Pony B Special calculator. The Pony B still had issues.

The reason for originally buying a Marchant Pony was to experience what these were like, it was local and they're rare hereabouts. They've been described as very light, and was curious how an American-made pinwheel machine would be. Unfortunately the machine was seriously damaged, and only after getting a parts-machine was it possible to try a repair.

Though it now looks complete and mostly worked, it still occasionally blocked. It'd block hard - not just more resistance, but no way that a turn could be completed, the only way out is to reverse the handle. (That would explain why on machines the full-turn pawl is missing or the back-plate removed, not statistically significant and all that, but from looking round online this seems common on these Marchants.)

The first hypothesis was that one of the short-ish pins in a carry-lever would sometimes jam. To test/fix this, the pins of suspected levers were replaced with long pins. From having a donor-machine, there's an assortment of springs and pins to choose from :)

By driving in a rod from the right pushing the gears axle, only one lever can be carefully released (hold something over hole, to prevent pin shooting out and get lost inside the carriage). Then the lever with replaced spring and pin can be moved back in-line and the axle pushed back in place.

That did not solve the problem of randomly blocking on a carry.

Observing the blocked machine with covers removed, it turned out that the carry-lever was actually stopped by the adjacent dial of the drum. 

The replacement drum-axle is of slightly different dimensions than the original and the drum is a few tenths of a mm too far to the right, relative to the carriage position. These Marchant calculators are truly a feast of interdependent adjustments. Finicky.

After trying to fix it with washers, ultimately swapped the original drum-axle back in. (Washers fixed the blocking, but created new issues with the clearing-lever. Plus the donor-axle had a broken full-turn gear.) 

Replacing the main axle and keeping the donor-handle meant that the gears of drum and handle-axle were now off. To mitigate that, decided to 'cheat' and change the handle position on the rod. The pin is now merely decorative and the fixing is with cyanoacrylate. Of course, the drum needed to be taken out (and apart) and thus the right sideplate needed to come off and because the locating-pins are missing the whole machine then needs tweaking to run freely.

 

Having the covers off, had another look at the interlock of carriage-clearing. This does not work. Looking at the pin that is pushed out of the carriage-frame when clearing, it is hard to see how it could have worked when new.

This is the original carriage, all parts are for this machine. Yet clearing simply doesn't push out the pin anywhere far enough to reach the dimples in the extra horizontal rod on the machine base. Can't see anything that could be adjusted, this is entirely from dimensions of the parts. We'll accept this interlock doesn't work - just take care to not clear and turn at the same time :)

Because the blocking problem is solved, now daring to fit the full-turn or anti-reverse mechanism. Trying to move the parts of the 1919 donor-machine over to the 1920 machine it turns out that the screw-threads are different! Both the lock-screw and the shoulder-bolt have different threads! (Why?? I'd now really like to talk with the Marchant factory design office of 1919.)

By shortening the shoulder-bolt, a lot of careful filing of threads and an extra washer, the full-turn lever is in-place.

After this extra round of fixing (attempts) the calculator still 'mostly works'. It is ok on addition and reasonably light in operation, but it is not smooth. Some carries in subtraction need work. It also is fairly noisy; the full-turn lever makes a ratchety clinkety-click as it dances over its gear. This maybe was another reason to remove this part in the past, it adds an annoying tinkle.

The clearing-nut for the setting register started to work - partially. It also is a mystifying mechanism - can't make sense of it.

The check-wheels are geared with the setting-dials via intermediate gears. When the handle is pulled out, the intermediate gears are shifted to the left to allow the drum to rotate and check wheels to remain static. The check wheels are held by a spring-loaded ratchet in a valid 'digit' position, nevertheless an extra hard-lock is engaged when the lever is pulled. In below photo the slotted bar is holding the intermediate gears, positively blocking any rotation.

When the handle is back in the rest position, the intermediate gears are back in engagement with the setting dials and the gear teeth now run through the slots in the locking-beam - below photo.

Adjusting the positions of all these shifting axles is finicky, any wear or deforming can block things. The scratches on one check wheel suggest that it had blocked and a violent attempt was made to get it moving again. (Forcing a wheel was a bad idea - wiggling the setting level can already help. Turning the clearing nut while at the same time pulling-out the main crank will unblock the lock-bar and set all check wheels to zero. That is also the way to re-sync the drum and chek wheels if they ever get out of sync.)

The locking-bar arrangement seems a bit over-complicated and superfluous - the machine should work fine also without this extra lock, the sprung ratchets will hold the digit. Another thing that is perplexing me still is that during clearing the lock-bar is shifted to the left as well - and then the intermediate gears need a nudge to the left too. The intermediate gears however also need to remain in engagement with the setting dials. This adds to the complexity of finding all the right adjustments, so many more ways for the mechanism to fail and block! 

With the machine at-rest, the intermediate gear should engage the setting dial, but must not foul the path of the pins. I.e. must be in-line, see below photo:

When clearing the drum with the wingnut via the check wheels, the intermediate gear must shift slightly to the left - just enough to remain clear of the left-shifting lock bar. But the gear must also remain in engagement with the setting dial. See below photo:

Frankly, the reasons for this complexity is not clear to me - plus that the way it is engineered makes the design 'brittle'.

After many attempts of adjusting, it now works - mostly. The column 7 and 9 do not clear, the pins on the axle must have been sheared off (again evidence that violence was applied in the past to a blocked check register). This whole mechanism is fixed with pinned collars - have not been able to remove any of the steel pins in the brass collars. Drilling out is an option, but so far that's deemed too much effort for this 'rough' machine.

The calculator assembled back together again after this round of explorations. Has to be admitted that it still looks better than that it actually works.

Even though this Marchant is not a representative experience of what these were originally like, it has given a much better understanding of how these calculators work. Also of how the Marchant factory operated and how their design evolved - the impression frankly is of two design cultures. There is the solid 'core' of the original Odher machine essentially unchanged, and then bolted on are the convoluted constructions of e.g. the carriage movement or the check wheel register.

During the 1917 to 1922-ish period Marchant would have sold every calculator they made so it may not have mattered all that much, but the impression is that they were not at 'world class' level of engineering competence for calculators.  In 1920 not yet, anyways.

This machine purchase is definitely delivering on the goal of experiencing the Marchant Pony! Oh and the Pony B still has issues :)

Blickensderfer Blackletter (blue fingers)

Starting with typewheel 407...

....and then swapped to a Blackletter wheel.

To make it easier to see how the ink is spread over a typewheel, swapping in a white typewheel:

It needs 'wetting' for the first characters, but then starts to be properly inked and type:

On a white wheel the blue ink clearly shows up.

The pattern of ink on the wheel nicely illustrates the Blickensderfer scientific layout; the most inked are the most frequently used letters that have the smallest rotation of the wheel.

This Blickensderfer 7 now back to the black wheel - it perhaps makes sense that Blickensderfer wheels were black, even if only for appearances sake :)

Model B - final step, all assembled

With all the bits cleaned, it's fairly quick and straightforward to put the Comptometer model B mechanism into the case again.

The photo exagerates the whiteness of the keys, they are still fairly grey in reality. The machine however really does look much better than it did:

It still shows signs of its age, with old repairs and old (and new!) refurbishments.

The '3' keys of columns 3 and 4 are not original, but are 1920s replacements. These keys may yet be replaced by new reproductions that are closer match for the composite keys of a model B. These celluloid replacements are however also 'original' and likely a century old themsevles - they are of Felt & Tarrant manufacture and an example of normal repairs that would have happened over the machine's lifetime. 

(The likely reason for these replacements could be seen on the remaining black 3-key. The composite material has ground stone (mica?) mixed with the resin; the black 3-key showed that cracks were starting where porous particles were embedded. It looked like the batch of 3-keys had too-large particles (sawdust?) embedded that weakened the part.)

Also the typeplate is a later, probably 1930s, replacement. It shows a 1921 patent number (for features not on this machine) and looks to be chrome, not nickel. It also shows this is a French machine, exported and sold in France.

Below found image of Comptometer 30680 shows what the original plate would have been, with a 1904 patent as most recent. 

These plates were however already often replaced for export machines, or swapped by the importer, with a localised typelate. Dutch machines often got a neutral plate without patent numbers, with text either in English or in Dutch. In the collection of Comptometers here, one ~1921 model H had just such a plate.

Because the French patent 528,226 of 1921 actually does describe the improved clearing mechanism of the H and would be correct for it, the plates were swapped out between the two machines. (No longer 'original state' now, but the B's plate wasn't original already. Fitting new/different plates and panels was routine during refurbishment of Comptometers, judging the configurations of several observed machines.)

This particular Compptometer was used; the holes for the keystems do have wear. Especially the colum 2 had a lot of 9's entered.

Placing the model B (left) next to a model C-light (right) shows some of the changes made, their development path of Felt & Tarrant. The keys were of course changed from composite to celluloid with a slightly different design and different front panel mounting.

Internally the very noisy B clearing mechanism was completely re-designed in the C (light) and more oiling holes were added to the case (lots of them).

Also an unexpected, small difference is in the subtraction cut-off tabs. On the B they are bent to the right, whereas on the C they are bent to the left. This probably make sense, an improvement - it visually links the tab more obviously to the column it is blocking the carry of.

On a model B the serial number is on the front panel between columns 4 and 5. This model B has serial number 30639 that places manufacture around 1908.

Thinking about that, 1908 is actually quite a long time ago.

Ladies at the Longchamp Hippodrome, Paris, 1908

Looking at it again, it is neat to be able to own and operate such a machine, calculating fine at well over a century old!

Model B - next step, cleaning the mechanism

This Comptometer Model B mechanism had collected an amazing amount of dust and debris. Amazing, because there are not many openings for dust to enter the case of a Comptometer.

It will have taken more than a century, but very nicely settled in dust-towers on some of the rods. Also a lot of debris had settled on the shelf of the bottom slats. This is pretty common, probably dirt that enters via the keystems and perhaps bits of cork lining.

Using a brush, most of the dust and debris was carefully removed. (Unfortunately not quite carefully enough - one small torsion spring fell out of the mechanism! Jolly tricky to get it back into the carry/suppression levers arrangement. Note for next time; brush even more carefully!)

With dust removed, there is some tarnishing, but almost no rust. The carry-suppression of column 8 did not work. Some gentle bending of the 'leg' of the supression-tab was able to fix that.

All the keys had of course been removed. This is fairly simple, unhooking the spring allows the 'wishbone-lever' in the bottom to tilt. This then releases the curved bottom tip of a keystem, allowing it to be lifted out. The keys were kept per column in numbered bags, i.e. all column 1 keys in bag 1.

All were cleaned in the usual way for these. That means, first fine steelwool to remove dirt and rust from the stem. Then wash in lukewarm water with a bit of dishwashing liquid to get rid of steelwool particles. And then a stiff (nail)brush to clean the keys. The lukewarm water will soften the dirt, acculumated grime of decades (century). The picture above flatters the cleaned keys perhaps a bit; they are still grey. But even if not half-as-white as celluloid keys, they are better and no longer dirty brown.

The missing white of several zeroes on the number wheels were re-touched with a light brown latex paint. Starting with cream and adding brown, yellow and black ink until it's a reasonable match.

Compared to later models with the controlled key feature, this mechanism is very open and empty on top. Noticeable is the solid actuating bar that pushes down all the rocker-arms forthe (noisy!) clearing.

A few levers needed to be lifted back onto a notch, but everything seems to be working as it should. Also visible on the frames are bits of cork where the metal had fused to the lining. Ready to be fitted with its top-plate again to fit keys.

Model B - first step, case refinishing

Lovely, decorative scrolls.

Those are the distinctive 'floral' scrolls on the side of the shoebox-style Comptometers. Taking a break from the Marchant Pony B, moved to tackling a Comptometer Model B.

Older Comptometers are becoming rather hard to find, but last year managed to get a Model B for a reasonable outlay. The reason it was reasonable, was that it looked rather bad. It seemed mostly complete, but some 'wrong'  keys, covered in black paint and -judging from clearing handle- possibly blocked. 

Now taking the acquisition out of storage, proceeded to open it up. The clearing handle was stuck solid on its shaft, only heat-cycling (soldering-iron) finally made it budge. As can happen with Comptometers, the mechanism was glued/melted firmly to the case-lining. That took time and force to break it free. After those challenges, it all came apart fairly easily. 

With some columns already taken out in the above image, it's obvious that the machine was painted black without taking off the keys. The dark area of the keys is the original, heavily corroded/patina copper finish. The black paint also was 'irregular' with many specks/spots - that could be rust underneath, but proved to be simply dirt and dust embedded in the paint.

Screws had also been painted-over.

Did not want to leave the case this way; so first removed the old, black paint. Paint can be removed in several ways, e.g. lye or acetone or specialised paint-stripper formulas. With older paints, a least-aggressive method that often works is simply hot water. Much safer than other methods and no problem then to leave the cork+linoleum lining in-place.

In this case hot water worked brilliantly. Place the panel in a container, e.g. a large baking-tray. Then pour boiling water onto it and leave it in the hot bath for about 10 - 15 minutes. The black paint could then be peeled off, revealing the state of the machine at the moment it was decided to cover it in black paint. Some areas still serviceable, but all corners and the top-front panel had been rusty.

Another benefit of the boiling hot water treatment is that the cork-linoleum lining turns soft and malleable. That allowed it to be flattened again, by letting it all cool off with weights (bricks) on the lining.

Fine water-proof sandpaper was used to slightly roughen-up the surface for adhesion and also smoothen rusted areas, then all panels were re-finished in a metallic copper-brown.

Glare makes it hard to get an impression from the above image, but attempted to get the colour within the variations that occur over time with Comptometers. Took a few panels on a trip to the auto-parts store, to select a close, matching metallic lacquer.

In the end, settled on using two spraycans. The Motip colour 51530 is very close, but has too much of a reddish hue. Spraying on a coat of the darker colour 51200 then followed by a coat of 51530 on the still-wet lacquer results in a decent, mix colour with less of the reddish overtones. The slight variations in colour over a panel that result from this 'manual-mixing' were found to be fine, similar variations are found on original panels (from variation in thickness of the original lacquer and from usage too).

The re-finished front-panel here on top of a Model J - it of course varies with different lighting, but overall the lacquer is a credible match.

The decimal-pointers on the front-panel were by the way not taken off, but masked with tape. Between applying coats of paint, the pointers were shifted to different positions. Net-result is that there is lacquer also underneath all the pointer positions (but fewer layers, of course).

After cleaning all the screws, the box was re-assembled and looks much better. All ready for the mechanism to be tackled next.

Platen hardness and type imperfections

With a spare platen salvaged from a wrecked Nr. 5 it was possible to try a variation of platen hardness. The Underwood Nr.5 already had its platen professionally re-covered a few years ago with relatively soft rubber - too soft for typewriter use really, but chosen anyways and types fine. The salvaged platen still had relatively good rubber, properly hard yet resilient. The newly-covered soft platen probably is around Shore 85 and the hard salvaged platen likely Shore 95 or thereabouts.

To start with, a few lines were typed on the machine with its soft rubber and a regular new ribbon; the type widely sold with a relatively coarse weave. (Though not quite as bad as the ribbons sold today for adding machines and/or cash registers - those are coarse and heavily inked too.)

Swapping in the hard-rubber platen and comparing the same lines, it becomes visible that the soft rubber cushions the impact and thus also blurs the type. The hard rubber creates a blacker and crisper imprint. The sound of the typing also is noticably crisper with a hard platen; more a sharp tick than a thud. The overall sound-levels are not all that different, more a change in character of the sound rather than a change in loudness.

The harder platen is less forgiving for any alignment issues of the type. Where the misaligned 'L' typeslug still prints the whole character on the soft platen (the type digs deeper into the rubber / the rubber accomodates the error), the hard platen only prints the top of the character. 

Swapping out the ribbon for a better one with proper inking and a fine weave, this perhaps makes the characters still a bit more crisp. Alignment issues of course show up very clearly too. That's the third test-paragraph on the test-sheet (no backing sheet).

Some forming and re-attaching of the 'L' slug mostly corrected its angle towards the platen, in the fourth test-paragraph it prints better. Not evenly, but at least a complete character. It is still too high and slanted to the left, but correcting that requires specialist bending and/or peening tools for typewriter repair (which we do not have).

From this brief dabbling, optimal hardness for a typewriter platen is probably at least Shore 90. If accepting a more blurry imprint, then a softer platen of less than Shore 90 (but likely (much) harder than Shore 80) can make a machine more forgiving for misaligment and change the sound a little.

(Comparing also with a cork platen; the cork definitely is on the softer end of the scale. Not quite as blurred as the soft-rubber, but certainly less crisp than a proper hard-rubber platen. This was a very hard cork platen, likely dating from the 1940s, and not comparable to cork-sheets available today from craft stores. The plus-side of cork for a platen is of course that it does not harden over time the way that natural rubber can do and will have made sense e.g. when rubber was hard to come by (England in the 1940s); but there is a trade-off with print quality.)

Net outcome, an Underwood 5 that types very decently; with a choice between crisp writing with sharp/loud typing sound and a slightly less crisp with more of a thud/loud typing sound. Swapping platens on an Underwood 5 is not quite as tool-less and easy as on some machines, but with a single screwdriver can be done in a few minutes.

-and also extra understanding of platen hardness and its effect on typing quality.

March 2026 safari - A single Swiss machine from Chemnitz

This morning's visit to the local thrift store, shiny chrome and gloss black amongst the bric-a-brac!

Viewing a bit closer, a Continental Standard in very nice condition - with an extra carriage attachment. Probably for typing cards?

With a QWERTZ keyboard and in French 'MAJUSCULES' on the shift keys.

Its origin is clear from the dealer label; this machine was sold in Lausanne, Switzerland.

Patents on the back (missing / screw-holes for optional tabulator parts?).

Its serial number 477466 dates this Continental to 1931. 

Over irs 95 years it travelled from Chemnitz via Lausanne all the way over to Friesland, now for sale on a table in a local thrift store - and recorded for The Database too :)