Marchant Pony transfer levers - adjusting

The Marchant Pony calculator did basically work after re-assembly, but not very well. It was noisy and 'rough' to operate, although it is hard to tell what these machines would've been like when new. Setting numbers was fine, but adding to the result-register required quite some force and especially the carries would sometimes block the machine. I.e. small calculations worked, but anything with several carries or larger numbers blocked - that for sure is not what it would've been when new.

Adjusting the drum position relative to the carriage already helped a lot with the required force - there are no dowel-pins on this machine to lock the side-brackets, so these can be adjusted a mm or so. Distance adjusted to create the smallest pressure-angle of pinwheel to the intermediate gears, whilst still no parts clashing that shouldn't.

The carries were likely blocked by the transfer levers requiring too high a force to push out. Feeling by hand, some were very hard to move.

These levers are held with a spring-loaded pin against a rod through their slot, toggled between out- and in-position. An extra tooth on a numeral wheel pushes it out, a cam on the drum pushes it back in. In the transfer-lever's out position, it causes an extra count on the next-up numeral wheel by the drum.

Taking the carriage off the machine, the force needed to push out and back in every transfer lever was measured. This confirmed that these forces were indeed very high and also varied a lot per position!

Pushing over 900 gram-force via that tooth by a single pin of the pinwheel is of itself worrying - risk of damage to that single pin. The variation in forces is likely caused by the springs and pins of a position being too stiff and/or rough.

The hole for the spring-loaded pin is (well..., should be) on the centre-line of the lever and centred in the slot, as per drawing. Differences between in- and out-forces then from an irregularly shaped pin, or maybe small deviations of a part. This calculator had very few of the original pins (or springs) left; most are 'crafted' replacements made when it was completely rebuilt sometime in the past. Whatever the reasons; the forces were too high.

With the force-meter at hand to test a position, all transfer-levers were re-assembled one-by-one. I.e. for every position the spring and also the little beveled pin trimmed to bring the forces somewhere between 200 and 400 gram-force. Lower forces to push out would be even nicer, but it became more difficult to make the levers 'snap' to their end-positions when pushed over the centre of their toggle.

After the re-fixing of the side-walls with improved drum-distance and tuning of the transfer-levers the calculator is noticeably lighter to operate. As bonus it's a bit less noisy too. The machine is now capable of doing zero minus one with a ripple-carry over 13 positions - still a challenge, but it can be done. Carries no longer block the machine! All that having been said; it is not as smooth and light-running as e.g. a 1930s Odhner or Thales. This of course may simply be wear of this particular specimen and not its design.

During some more testing, one position started to fail doing a carry. This was caused by the lever not 'flipping' fully to the out position. By pushing in a rod from the right, some 'minimally invasive surgery' was possible on this single position. The offending lever was rotated off the rod to get at the spring and pin. Then re-assembled with a stronger bevel filed to the pin of position 10 - the bevel helps it 'flip' to the end-positions when toggled.

The Marchant Pony B Special calculator; now with adjusted carriage slightly less noisy and capable of 13 carries in a row :-)

Sighting of an old Adler on a shelf

During a trip last week, did a stop-off at a local thrift shop. On a shelf were a well-used Hermes standard and an old Adler. That Adler was surprising; pre-war machines rarely show up in thrifts stores anymore.

The decal on the cover was in great condition with cover-nuts present as well as original spools.

 

This may actually be a pre-pre-war machine; the serial number was I think 4-digits and that'd make it a pre-1914 specimen. (Serial number 6639 ? -should've taken a better picture.)

The drawband was disconnected (still present) and one keylever had been mistreated (bent?), but overall the machine looked complete and rustfree. The nickel should polish up fine, and the wide ribbon looked usable too.

Overall a decent machine with lots of potential to be a very fine ancient typewriter; at 45 Euro an entirely reasonable asking price too. (Left on their shelf :)

Marchant Pony B calculator coming together

Now that the main functional blocks of the calculator, i.e. the carriage, the setting-drum and the carriage-shifting are all re-built, the whole machine was put together again. A 'test fitting' , with all the covers on too.

The carriage cover had the 18th position blocked off, an old 'fix-up' of a severely damaged machine:

This blocking-off had been very neatly done with a soldered-on bit of brass. Now de-soldered and the hole cleaned-up:

The covers of Marchant machine 70049 are all stamped with '49', i.e. the last two digits of the serial number. For now, only the carriage back-cover was swapped-out with the part from the donor-machine (serial number unknown). The original of machine '49 was really very mangled - oddly so.

Several bits were also tackled or transplanted - the decimal-pointers and bars were taken from the donor-machine -and of course the wooden baseplate.

When putting the cover on the carriage, the counter register did not align properly with the openings - the left digits drop below the viewing holes.

Some checking found that this was not due to a wrong assembly or the carriage itself being bent, but simply because the holes in the cover are misaligned! The cover part is manufactured over many separate stamping steps, so there is a risk of the plate not being fully seated against the alignment pegs on one of these steps. A small misalignment in the tool this time (i.e. back in 1920) resulted in a slight misalignment in the counter-holes. Not major, but still unexpected that this got past final-inspection - these were very expensive machines! A Pony B Special sold new for well over 300 dollar in 1920.

For the whole 'typical look' of the Marchant, a quick-fix reproduction nameplate was mounted on the top-cover and a bit of felt was cut to size to put between baseboard and machine. The long-lost factory original felt probably was black or dark purple/blue, but green is nice too.

To try make the calculator work smoother and as a test, the right sidewall here swapped out with the donor-machine's sidewall (that does not have the brass weld repair) - it really is a luxury to have so many spare parts to choose from! :-)

There still remain many small items to be worked out, but this Marchant Pony B Special calculator basically works again. It does not work very well yet, but it's a start :)

Agelist for Underwood frame numbers (left-foot numbers re-visited)

Spurred on by extra datapoints of Underwood typewriters left-foot numbers on The Database (Thank you James!), and now looking only at the Underwood 5 samples, there are enough points to reasonably map a relation.  [Updated graph; showing 24 Underwood 5 numbers.] 

It is not quite a linear fit (the portfolio mix of Underwood was evolving), but a polynomial fit gives a pretty good correlation. Using that fitted relation and the serial numbers from the agelist, then an 'auxiliary' agelist based on frame numbers can be created:

year frame number

1905    75,000

1906   115,000

1907    154,000

1908   206,000

1909   260,000

1910   329,000

1911      410,000

1912   515,000

1913   639,000

1914   769,000

1915   883,000

1916   998,000

1917 1,141,000

1918 1,287,000

1919 1,422,000

1920 1,601,000

1921 1,812,000

1922 1,948,000

1923 2,130,000

1924 2,323,000

1925 2,520,000

1926 2,768,000

1927 3,005,000

1928 3,214,000

1929 3,386,000

This agelist from frame numbers (a.k.a. the front-left foot numbers) probably has a margin of about a quarter-year either way. Note that after 1926 the first digit of the frame number can be missing, but the full number likely still stamped on the carriage side.

This 'auxiliary' agelist very likely can also give a year-estimate for Underwood typewriters that fall in other serial number ranges than the 5, but were built from the same frame-casting and assembly line.

There's more information still to be had from more data - e.g. there are hints that 1916 saw a significant, sudden shift in the Underwood portfolio -or factory set-up. Also there are hints on batch-sizes, and production buffering was not completely first-in first-out either.

I.e. as more frame-numbers become known, the data will improve and more information can be extracted. 

But already possible is a rough "front-left foot agelist"  :-)

[Update: volume estimates.]

After 1929 the data becomes hazy; maybe because of portfolio shifts or because of the market upheaval from the October stock market crash. After October the sales of expensive business equipment such as typewriters (or Elliot Fisher bookkeeping machines!) probably collapsed.

Nevertheless, between about 1905 and 1929 there seems a robust set of numbers; both for Underwood 5 serials and for all Underwood frames (on that typewriter-frame line, at least). This means an estimate can be made of the volumes and volume-share of the ubiquitous Underwood 5 for the company:

This shows the fluctuations of the overall sales, and also that the share of Underwood 5 (and 4) machines is more than 90% of machines in 1905, but by the end of the 1920s has dropped to about 60% of sales. It has become a less dominant product for the company - typewriter sales have become more diversified; the business market perhaps becoming more receptive to special features such as wider (and more costly) carriages.

Marchant carriage shifting mechanism repair

The carriage shifting mechanism of the early Marchant calculators is very recognisable, it gives these machines a distinct profile. This mechanism is subject of a 1916 Marchant patent and is quite exposed on the front of the machine - putting it at increased risk of damage.

One of the levers was indeed broken on the target machine, the broken-off piece fortunately still held inside the mechanism.

To try fixing this, the lever was taken out. Fortunately the rod that holds it all in place was freely movable, so slid over to the left and the parts taken out (hold the spring as you slide the rod, to prevent it shooting off into the distance).

The right shifting lever was indeed broken, and actually had an old repair that had failed!

The original steel lever arm (rocker arm) had originally broken where it's weakened by the hole for the connecting rod.

To repair it, some time long ago a brass plate had been soldered to the side and two drilled/pins were added. The solder connection of the brass to the rocker arm had however failed and the part again broke.

After very thorough cleaning of the part, the old repair was re-soldered. The pins could now also be embedded in solder and extra care was taken (flux, much flux) that the solder reached the entire contacting surface of the brass plate with the lever. Hopefully the solder will hold, to be able to keep the original levers on this machine; they are stamped with the last digits of the serial number - parts of machine 70049.

The whole carriage shifter design seems a bit 'out of character' with the rest of the calculator. Even though it works well and is easy to use, it looks (overly?) complicated in its parts design. Even the lever is assembled from three separate parts; three stampings (needing multiple tools!) and a turned rod. Below shown as loose parts; the two different (!) rocker arms and the keypad - these need assembling and 'riveting' together.

That's not even mentioning the complex locking-pin, the swaying gear assembly and complicated casting details of the base. Quite a few parts to be assembled, with a few very tricky springs to insert as well. Here in in the factory; assembly of the full-size Marchant A, not the Pony:

It does however work very well, is surprisingly easy to use and is an (the) original Marchant-specific part of the calculator.

Comparing to the donor-machine, there are again multiple small differences also in this mechanism. These machines are probably very close together in time; the donor machine (shown on right) will date to late 1919 (November?) and the target machine (shown on left) probably is an early 1920 machine.

The levers are different; an extra spring-hook and a changed order of plating the part. Different torsion springs, even the base-casting is different!

Also the locking spring-plate of 1919 (left) is entirely different from the 1920 item (right). The pin from the donor machine was filed to fit the target machine; to replace the pin with the broken-off carriage-release tab.

All put together on the target machine and the carriage shifts again as it should! :)

Pinwheel drum of a Marchant Pony B - cleaning and its manufacturing

The pinwheel drum of the donor-machine was blocked - that is, no column would operate all pin-positions, pins were rusted solid in retracted position. Part as a rehearsal for overhauling the drum of the target-machine, partially to see if this drum could be better and partially just out of curiosity; taken apart for a cleaning and overhaul.

To take the drum out of a pinwheel calculator, one of the sidewalls has to be removed (i.e. a 'bracket', in Marchant-parlance). Then the drum itself can be taken out and apart. First step for that is to remove the pinned gear at the counter-side (the left in image below). Second is to remove the holding nut (red arrow).

The biggest challenge in cleaning this drum was the loosening of the holding-nut. This needed heat (soldering iron) and even some carefully-aimed hammerblows to start moving. (Clamp shaft in vise with protection, eg copper, don't apply any force to the disks!) The taper-pin of the counter-gear on the other hand came out easily. With the counter-gear and holding nut removed, all the sections of the drum simply slide off. Then a collection of 9 pinwheel disks, 3 carry-segments, clearing-arm, clreaing-arm-ring, a collar, one holding-nut, counter-gear, locking-comb, comb-pushpin, comb-spring and of course the main shaft:

The pinwheel disks are held together by two screws and a cover-plate. The carry-pins in disks 2 to 9 are peened in-place and cannot be removed. A brass disk, operating-ring, detent_spring, 9 pins, 2 screws and a retaining plate:

All disks (or dials) are different, specific to the angles of the specific dial digit-position. The timing of the pins and especially the tens-carry is staggered over the drum. The dimensions of the parts for every dial are thus different. Angles are given in the setting of the dividing-table of the slot milling machine.

The diameter of the drum is divided in 37 positions. Central 9 positions are needed for the setting pins, slightly staggered from disk-to-disk -to spread the load of turning the numeral wheels over the turn of the crank. Outside of this centre setting-pin 'field', there are the 10-carry pins spiraling out from dial 1. These carry-pins need one position offset from disk-to-disk to allow a wheel to do a full carry before the next engages.

The steel cams that reset the ten-carry levers are most clearly different; shown in this below drawing - these cams are all stamped with their position number.

The position-specific parts of a dial are all marked with their position number (an obvious and very useful thing to have in manufacturing) - below for example dial number 7:

Dial parts were cleaned individually: steelwool to de-rust the plate and metal-polish for the pins and ring. A very small amount of light, clean mineral oil (sewing machine oil) in assembly and some vaseline rubbed on the retaining plate against rust. When put together again, all pins move - although perhaps not quite as smoothly yet as they would have when new. 

Taking apart the drum also revealed markings from the manufacturing process.

The drum of the donor-machine is marked at the outside of dial 1 with a date and initials. This suggests that employee RA assembled this drum on October 15, 1919 (a Wednesday). 

Several -not all- of the pinwheels (or dials) are marked with initials TH scratched on the retaining plate. Individual disks were assembled separately, with employee TH scratching their initials into every disk they made on the obvious surface to do so. 

Dials were first assembled and stacked per dial-number onto a stake on a wooden plate. There are even photographs of this in the Marchant factory around that time. A store of assembled dials in a 'vault' cabinet from a 1918 article and stacks on a workbench in a cropped detail of a 1920 photo captioned "service department" (?):

Another thing found out when comparing the two drums (and trying to exchange) was that the large driving-gear is not in a pre-determined position to the drum. That was surprising. It is a press-fit onto the drum-shaft that has the keyway to orient all the dials and parts. The gear teeth position appears to have been random, but at a first assembly of a machine the crank-handle shaft would be drilled for the handle. Thus the two shafts are matched in the angle of gears, they work on that specific combination. This was concluded from the gears on the drums of the donor and target machine not being at the same tooth-position. 

The small angular-offset if mixing parts between machines is probably not too noticeable for a regular Pony, but for a Special with the check-dials it means that the check-dial gears won't mesh perfectly centred with the gear on ste setting-dials:

Effect of all this is that replacing a crank-shaft may mean also needing to exchange the drum-shaft. They are matched sets.

Anyhow, continuing the drum re-assembly in 2026. All the clean parts laid out and slid onto the shaft:

And put together again.

And then mounted into the target-machine - both drum and crankshaft from the donor-machine.

The target-machine original drum actually worked relatively fine after a little oil and a work-out. This donor-drum may in the end not be kept in the machine, it does however enable using the straight crankshaft. A restoration choice to be made later; straight shaft, but not the original for machine 70049 - an authenticity question...

The set of drum and shaft from 70049 now out of the machine, to get the same deep-cleaning and rebuild - no scratched-in date, but a stamped A-number instead. Maybe there are some hidden markings on the internals yet to be found :)

Carriage rebuilding of Marchant Pony B

The carriage of both Marchant Pony calculators were partially seized - multiple positions completely unmovable. Additionally, the 18th position of the result-register of the restoration-machine was missing (!), and mangled wingnuts.

At some moment in its past, this machine was radically repaired - repaired from a 'catastrophic event', e.g. being dropped, being hammered. That event is probably also when the main crank was bent, levers broken and the 18th a wheel lost (how??). Not just the numeral-wheel, also its ten-carry lever plus gear was taken out. A small piece of brass was soldered in its viewing aperture in the cover. This was then neatly painted black, making it a 17-digit machine. 

As a start, the target-carriage was completely emptied of all parts - the usual stubborn screws and hardened old oil, but also the counter numeral-wheels were jammed solid between the carriage sidewalls. In the end, an empty and clean carriage; a brass casting with a simple steel strip screwed to the bottom as the sliding surface.

No pictures of rebuilding the counter register - this was hard! to do and was assembled and taken apart at least ten times! One surprising feature and difference with the donor-machine were brass plates or shims between all the numeral wheels. 

These very thin brass plates were all somewhat mangled, probably from brute-force clearing attempts or perhaps a botched assembly. When this shim is not aligned with the main rod, the clearing-pins would 'notch' the brass as the rod is pushed in - this makes extra thickness and pressure on the stack - causing it to be blocked. It is doubtful that all positions of the counter would have worked well after the old repair.

These shims also made it impossible to exchange some wheels with the donor. To accomodate for the brass shim, the wheels are about 6.75 mm wide, whereas on the donor all wheels are 7 mm wide. Well, they really are just under 6.985 mm. This is an American machine; the pitch of the columns is not 7 mm as they would be on a Continental machine, but on the drawings is given as 0.275". Inches and fractions...

(The making of many parts of the Marchant Pony are shown with good explanations, drawings and photographs of the actual tooling in a great 1919 book on Punches & Dies. The book also shows several stamped parts for the Noiseless Typewriter and some Smith Premier typewriter parts. The book can be found in full on The Archive.)

From both donor and target machine, a glut of numeral-wheels. A total of 35 specimens in various states of wear - cleaned and to be sorted on quality. 

Again small differences, the target machine wheels are about 24.5 mm diameter, the donor-wheels are 24.0 mm on average. Probably simply batch-to-batch variation.

Many wheels - a 35 digit register; that'd be a 116-bit register - the actual 18-digit Pony B is already not bad as a ~60-bits computing device :-)

Cleaning, fitting and filing of the the tens-carry levers, then on their rod and screwed in-place. The left-most levers have extra prongs to trigger the overflow-bell (carriage here seen from behind).

After these levers are fitted, the springs and tiny plungers have to be cleaned and fitted to every lever as a rod is pushed in from the left to lock things in-place with an intermediate gear next to every lever.

These are the little spring-loaded pins that keep the tens-carry levers in the out- or in-position. The parts of the target machines are not original - these springs are offcuts of a different size, and the plungers are rounded bits of varying length. Good replacement/repair effort, but not original. The circled three are original Marchant pins with a bevel.

As the rod is pushed in one lever at a time, the springs and pins are selected to give an even, reasonable force needed to flip the lever between positions. Some spring shortened, lengths of pin and springs matched. (Below picture is of a test-fitting, gear-wheel for column 17 needs to be added still.)

Also here the 18th wheel was given a test-fitting - to check if there was something wrong in the carriage-frame, causing the 18th position to be removed. No problems, all works fine. Likely that one numeral wheel or other part was broken, then leaving off the 18th position is the way to solve a lack of parts. The repairer decades ago won't have had a donor machine :)

After the gears-rod is fully inserted and all the intermediate-gears plus all carry-levers are in place, the register can be filled. Starting from the right one position at a time.

To test, first placed on the machine with the first two digits only. The holes vsible on the register main-rod shows that several clearing pins are likely replacements, only few original pins remained. So many pins needing replacement does suggest mis-use; excessive force on the clearing wingnut is needed to shear off these pins - or forcing wheels (perhaps breaking a wheel). The scoring of the numeral wheels also hints at brute-force attempts to move blocked numeral-wheels. This machine really needed to be seriously rebuilt!

The springs for the overshoot-prevention rockers need to be placed with care. When simply pressing everything in place, springs don't enter their pocket and will be mangled and the rocker will be very stiff (blocking?). Several springs were indeed mangled - now replaced with good springs from the donor. 

When everything is in place, the front rack-bar can be screwed in-place. The entire carriage is now clean and functional - well, mostly functional. 

The clearing-flange of the counter is mounted upside-down; that vane should be at the bottom. With the new wing-nut and the pins being what they are, I simply cannot find a position where all wheels clear and no wheels are fouling when zero. In comparison, the 18-digit result register was easy to get to work.

The gear surfaces that slide over the detents are still rough (hard to polish); excercising the digits hopefully will make things smoother. For now the result is one basically functional carriage of a Marchant Pony B calculator :-)