The model’s dial
The Einheits-Chronometer
Orel Magidish (assisted by Claude Opus 5.5) – October 2026
A chronometer built from its drawings. Wempe’s works drawings for the German standard marine chronometer give each hole’s place to a thousandth of a millimetre, and beside them a measurement its makers took. This essay reads them, and shows where the model on the other tab finds them agreeing, and where it doesn’t.
A companion to The Marine Chronometer, the essay on the Hamilton Model 21, which explains how a chronometer keeps time: this one links to it where an idea is the same. The drawings themselves are on the Deutsche Gesellschaft für Chronometrie’s site; every figure here is drawn from the model’s own numbers.
One standard chronometer
The Deutsches Einheits-Chronometer, the German standard chronometer, was designed by Gerhard D. Wempe’s chronometer works in Hamburg in the early 1940s. It was made by Wempe and by A. Lange & Söhne during the war, and by Wempe after it. It is a marine chronometer of the classic kind: a fusee and chain, a spring detent escapement, a cut bimetallic balance with a helical spring, in a brass bowl hung in gimbals in a wooden box.
The Deutsche Gesellschaft für Chronometrie’s page that publishes its drawings names Ernst William Meier and Friedrich Leutert as its geistige Väter, its intellectual fathers, and John Schwarzer and Herbert Müller as its chief designers, whose names are in the drawings’ title blocks. Several hundred drawings were made, and the wartime ones were revised from what series production taught, the balance staff and the escape wheel among them. The published set comes from Wempe, from private sources and from the Deutsches Uhrenmuseum Glashütte, which holds copies stamped “KOPIE” of the drawings Wempe supplied to Lange.
One date doesn’t fit. The escapement’s layout, “Hemmung für Feder-Chronometer, Kaliber 020” (sheet 05 04_1), is dated 03.02.40, which the page reads as 1941, since Wempe had no spring chronometer in 1940. The lever chronometer’s escapement sheet, 05 04_20, is dated 03.02.41.
Knirim and Schmiedchen (watch-wiki) list changes made at Glashütte: the bowl and gimbal ring in Bakelite, with a heavy weight in the bottom, from No. 5100; the balance lowered to 3.8 mm from No. 5111; the maintaining spring (Gegenfeder) redesigned at Lange from No. 5201. Lange No. 5177, of 1944, is photographed on dg-chrono.de with its papers.
The model on the other tab is the spring-detent version, the Feder-Chronometer, Kaliber 05, as its drawings draw it, each sheet as last revised. Where a sheet gives a figure, the model uses it; where it doesn’t, the size is measured on the drawing at its scale or on photographs of real ones, and marked as estimated. Each part’s card says which.
Why a ship carried a chronometer at all, in the Hamilton’s essay
Reading a works drawing
A works drawing is a sheet for the shop: one part, drawn to a scale, with every size it must be made to and how closely, its material, and who drew and checked it. Take the pillar, 05 08 00-1, one of the three that hold the plates apart. Herbert Müller drew it on 2 June 1940, at five times full size, in Rundmessing 8,5 Din 1758 Ms 58, round brass bar 8.5 mm across. Its body is Ø 8.1 −0.2. It stands 19.6 ±0.01 between its shoulders, and at each end a spigot Ø 4.5 f7, 2.8 long, goes into a plate. The sheet also gives the overall length, 25.3 −0.2, which the parts check: 19.6 + 2 × 2.8 = 25.2. The pillar is tapped M2.6 right through for the pillar screws, “Gewinde von beiden Seiten einschneiden”: cut the thread from both ends.
“f7” is a tolerance from the standard system of fits: for a shaft 4.5 across it means between 0.010 and 0.022 mm under size, 4.478 to 4.490. The plates’ holes for the spigots are reamed “4,5 +0,012”, 4.500 to 4.512. So a spigot always slides in, with 0.010 mm to 0.034 mm to spare across its hole: the pillars set where the top plate lies over the pillar plate, and so where each arbor’s upper pivot hole stands over its lower one.
The shoulders don’t sit on the plates’ faces but in seats, Ø 8.6 and 0.3 +0.03 deep, turned into each plate’s inner face. So the plates stand 19.6 − 2 × 0.3 = 19.0 mm apart, the “innerer Plattenabstand 19 mm” of the overall plan (sheet 001), and 18.93 to 19.01 at the limits.
The plates come with tables. The top plate’s sheet, 05 09 01, lists every hole by its number, with its place as two coordinates from hole 1, the centre wheel’s arbor, at the plate’s centre: N or S, and O (Ost, east) or W. Then how each is made: centre-punched (Körner) or punched to a size, drilled, then reamed or tapped. The pillars are holes 11, 39 and 49; hole 11 is “S 3,225 O 36,858”, punched 4.3 and reamed 4.5 +0.012. Worked out, the three lie on one circle 37.0 mm about the centre, at 95°, 210° and 350° from N toward O: 115°, 140° and 105° apart, not evenly spaced. A hole has the same number and the same coordinates on every sheet it appears on, the plates, the cocks and the bridges; read across them all, they agree to the last figure.
| No. | Benennung | N/S | O/W | Punched | Drilled | Reamed |
|---|
The frame cut through pillar 49 in the working model
The top plate from above in the working model
The fusee, measured by its makers
A mainspring pulls harder fully wound than run down, and the fusee, a grooved cone the chain winds onto, gives the train an even pull from it. Wempe’s drawings include a record of how even. Sheet 05 06_20, “Schneckenkurve / Kraftverlauf” (the fusee’s curve, the course of the force), signed by Herbert Müller on 24 July 1941, plots the force over the fusee’s turns: “Mittel aus 2 Messungen”, the mean of two measurements, with a lever 262 mm long (“Hebellänge 262 m/m”), running down (“Ablauf”). Over the whole wind it reads between 126.5 and 138.4 g; the sheet’s own mean is 131.5 g.
Where was the lever? The sheet works out its own sum, 131.5 × 262 / 19.6 = 1.758 kg at the fusee wheel’s pitch circle (d = 39.202), which takes no account of the fusee’s radius; and a pull as even as this over the barrel’s five turns is what a fusee hands its arbor, not what a barrel gives. So the model reads it as measured on the fusee arbor, with the spring’s, the chain’s and the fusee’s losses already in it. Through the going train, 3,085.7 to 1 from the fusee wheel to the escape pinion, at an efficiency the model takes as 0.780 (0.730–0.833 from its pivots and its meshes worked tooth by tooth), that is 82.1–89.8 µN·m at the escape wheel over the wind.
No sheet gives the mainspring’s length, only its strip, 15 × 0.37 mm, on the same sheet. The model solves the spring as the Hamilton’s is solved: an elastic strip coiled in the drawn barrel, its shape the one of least bending energy at each quarter turn of the wind, its coils pressing on each other, set up three quarters of a turn (“3/4 Umgang gespannt”) and wound through the drawn fusee. Its free curve, which no sheet gives either, is the Hamilton’s measured spring scaled to the length, its turns three quarters of the Hamilton’s. A strip 1,240 mm long gives the measured force’s shape to 2.25 % rms, and to 1.8 % with its coils’ friction counted (the force was measured running down, the friction in it): the hump a turn from run down, the dip in the middle, the rise to full wind.
The level agrees too. At 0.37 mm the solved strip pulls 9 % harder than the sheet measured: the measured force is 0.917 of its pull without friction. The losses between the spring and the fusee arbor, worked out from the parts (the coils rubbing, the barrel on its arbor, the chain’s links on their rivets, the fusee’s pivots), come to 2.0–14.6 %; the measurement asks for 1.1–13.2 %, and with the losses at their middle a strip 0.370 mm thick: the sheet’s own figure, its “0,37” stencilled, not a 7 written over another. Solved so, the strip is stressed 2.45 GPa fully wound, at a steel mainspring’s working limit, where a mainspring is meant to work. The free curve’s set decides the level more than anything the sheets leave open: with less set, which fits the shape best if the friction is left out, the same strip would pull 23 % harder than measured, past its steel. The coils’ friction takes most near full wind, where a spring with less set nears coil-bound, so counting it the shape asks for the set at which the level closes.
How a fusee evens out a mainspring, in the Hamilton’s essay
The fusee in the working model
The helical spring and Phillips’ curves
The balance swings against its hairspring, and the two together keep the time. A marine chronometer’s hairspring is a helix: a strip coiled into a cylinder, with a curve at each end that brings it in to the collet on the staff and out to the stud on the cock. Wempe’s sheets draw it. The strip (05 03_6, Spiralfaden) is 0.285 mm thick and 0.52 mm high, steel hardened and tempered dark blue (“dunkelblau anlassen”), coiled 12.8 mm across with its turns 0.73 mm apart. Its terminal curve (05 03_7, Spiralkurve) is drawn at ten times full size, with a table of its radius every 10°. These sheets carry the lever chronometer’s numbers, but the spring they draw, with its collet and stud, is this one’s, and the model takes it as shared by the two.
How long is it? 05 03_7 says “vorm biegen der Kurve auf 13 Umgänge −3,5 mm kürzen”: before bending the curve, shorten to 13 turns less 3.5 mm. Read as the words read, the whole strip is cut to thirteen turns of the coil as wound, less 3.5 mm, and both curves are bent from its ends. That leaves 11.80 turns of helix between them, 505.4 mm from pin to pin, and puts the two pinning points nearly in line, 7° apart, where 05 03_1 draws them 5° apart. The drawn coils lean a turn longer: 05 03_5 draws about 12.4 turns between the curves.
The curves matter because a plain coil pinned at its ends doesn’t open and close about its centre. As it winds and unwinds it pushes the staff sideways in its pivots, and the rate comes to depend on the swing. Phillips showed in 1861 what a terminal curve must do to keep the coil centred: the curve’s centroid, from where it leaves the coil to its pin, must lie on the perpendicular to the radius at that junction, and at R²/l from the centre, R being the coil’s radius and l the curve’s length.
05 03_7’s curve is drawn with a compass. Its first quarter, 0–90°, lies on a circle of radius 5.94 about a point 0.33 mm off the coil’s centre, and its run from 110° to the pin at 220° on a circle of radius 4.35 about a point 1.26 mm toward 90°, both within 22 µm of the table; the sheet marks small circles at those two points. Against Phillips, the drawn curve’s centroid lies on the perpendicular, 1.7° off it, so the first condition holds. But it lies 3.01 mm from the staff, where the second asks for 1.91: the curve’s first moment is 1.58 times what it should be. The curves that would meet both conditions, to the same pins and leaving in the same direction, are 28.4 mm and 28.4 mm long, against the drawn 20.5.
What the second condition is for is the force on the pivots. Under a couple, the drawn spring presses the staff sideways with 0.018 of the couple divided by the coil’s radius; ends that meet Phillips’ conditions leave none, and plain ends about 1. Each curve’s error lies along the tangent at its junction, so the errors of the two cancel when the helix between them is a whole number of turns (0.0025 at 11, 0.0023 at 12) and add at a half (0.030). The pins’ place sets the helix at 11.80 turns, and so the force.
The spring also stiffens a little as it winds, so against small swings the balance runs a few seconds a day fast; about the model’s swing that gain shrinks as the swing grows, and the drawn spring loses 0.66 s a day for each 10° more. Over the wind that works against the escapement’s own change of rate with the swing, and takes back only a little of it: with both, the model’s rate stays within −1.19 to +0.96 s a day over the 56 hours, where the escapement alone would move it −1.27 to +1.04 s.
The length also sets the period. The strip’s stiffness, with spring steel’s modulus, is 409 µN·m a radian; with the balance as the model weighs it, 2,501 g·mm² with its rollers, collet and a third of its spring, that is a period of 0.491 s, 1.7 % short of the 0.5 s the train counts (its seconds wheel turns once a minute, in 120 teeth of the escape wheel). Most of a larger gap closed when the masses’ height was read: 05 03_2’s near pair and Wempe 3955’s balance seen side on put them 6.46 mm high from the rim’s lower face, where the model had built them 5.4 from the far pair’s outline, and the period was then 0.468 s. What is left is small: 0.5 s wants a balance of 2,589 g·mm², 4 % more, or a strip 1.1 % thinner, inside what the strip and the steel’s modulus allow. Read as fourteen turns less 3.5 mm the spring would give 0.510 s, as fifteen 0.528 s. The model runs its balance at 0.5 s, as a maker would time it.
The hairspring and why it breathes, in the Hamilton’s essay
The hairspring in the working model
The bimetallic balance and the adjuster
Warmth weakens a steel spring. Its modulus falls as it warms, so a steel hairspring on a balance that didn’t change would lose about 9.5 seconds a day for every degree Celsius warmer. Carried from a cold port into warm water, it would be minutes out within days.
The answer is the balance itself. Its rim (05 03 01-1) is two metals, “Nickelstahl Sonderlegierung u. Messing”: nickel steel inside, brass outside, 32.32 mm across, cut through at 90° from the arm. Each quarter is a cantilever from the arm’s end. Brass grows more with heat than nickel steel, so as it warms each quarter curls inward, and carries the masses on it toward the staff. The balance’s moment of inertia falls, it swings faster, and that offsets the spring’s loss. How much depends on where the masses sit: near the cut they move most.
The masses stand 30° from the cuts’ line on Wempe 3955 and in the 6MX’s book; 05 03_2, a poor copy, scales 32°. The model works out the curl by Timoshenko’s formula for a bimetal strip (1925), each quarter a cantilever, every part of the balance weighed on the model’s solids. The sheet names the rim’s steel only as a “special alloy”; the model reads it as Guillaume’s nickel steel of 42–44 % nickel (an estimate). With that, at the drawn place the balance gives back 6.8 s a day of the spring’s 9.5, and the chronometer still loses 2.7 s a day for each degree warmer, where the 6MX’s technical conditions allow ±0.10. The masses would cancel it at 19° from the cuts’ line. Each degree along the rim moves the figure by 0.22, so the 6MX’s ±0.10 is ±0.4° of the masses’ place. Over the published ranges of the metals’ constants, the drawn place loses between 0.9 and 4.5 s a day a degree, and the place that cancels it lies between 14° and 26°. With a plain nickel steel no place on the rim would do: the masses as near the cuts as they fit would still lose 3.2.
So by the model’s middle constants the masses would have to move 11° toward the cuts, about 1.9 divisions of the rim’s graduation: 05 03 01-1 has “8 Körner in 60 Teilung”, which the model reads as eight punch marks of a sixty-part division, 6° apart. Setting them is the adjuster’s work: the chronometer is run hot and cold, and the masses moved along the rim until the rate holds, as the 6MX’s book describes for its balance. Wempe 3955’s masses, on a chronometer adjusted in its day, stand at 30°, so with its own metals the balance compensated there. What is open is not the masses’ place but the constants: the rim’s steel and the spring’s change of stiffness with heat, which the drawings don’t give.
A second error stays. A steel spring’s modulus falls faster as it warms, so a balance set right at two temperatures gains between them: Dent’s error, a few seconds a day at 15 °C for one right at 0 and 30 °C (Guillaume’s Nobel lecture, 1920), about 2.5 seconds in the middle of a 30 °C range (VintageWatchstraps’ account of the Neuchâtel trials). The model takes 2 to 3 seconds (an estimate). Over the 6MX’s trial at 4, 20 and 36 °C that alone gives a middle-temperature error S of +2.3 to +3.4 s, and a rim of brass on plain nickel steel leaves it at +2.2 to +3.4 s, outside the 6MX’s ±1.20. Guillaume’s integral balance cancels it with a nickel steel whose expansion curves the other way, and the 6MX’s book calls its balance integral. With Guillaume’s 42-44 % nickel steel as the National Bureau of Standards tabulated it (Circular 58, 1923) and Dent’s error at the low end of its range, 2 seconds, the model’s S is +0.8 to +0.9 s; the passports of five Soviet 6MXs tried to the same conditions state +0.80 ±0.17 s, where 3 seconds would give about +2, outside the trial’s ±1.20.
Then the trial. The 6MX’s technical conditions put each chronometer through five weeks at 36, 20, 4, 20 and 36 °C, and count the last five days of each (the limits are in the table below). The model runs it day by day: the balance’s and the spring’s change with temperature; the air, which takes over half the balance’s damping and is denser in the cold; the swing over each day’s wind, wound daily to “8” on the up-and-down dial; a scatter and drift from day to day, matched to the test cards of eight Soviet 6MXs, since no Wempe card is to hand; the barometer and the oil (the Hamilton’s estimates); and a rate regulated at 20 °C to within half a second a day (an estimate). Move the masses and the trial follows.
As adjusted, the model passes; with the masses at their drawn place and the model’s middle constants it fails by its temperature coefficient alone, and moving them along the rim is all it takes. The adjuster’s hand is the one estimate the trial turns on: the model takes the masses set within ±0.45° of their place (an estimate), which keeps the coefficient within ±0.101 s a day a degree.
How a cut bimetallic balance answers the spring’s loss, in the Hamilton’s essay
The balance in the working model
The escapement in time
The spring detent escapement locks the escape wheel on a stone set in the detent, a light arm on a spring. Once each oscillation, the discharge stone on the balance’s roller lifts the detent by its gold spring and lets a tooth go; the wheel turns, drives the balance through the impulse stone on the other roller, and the next tooth locks. On the swing back the discharge stone only bends the gold spring aside.
Wempe’s layout, 05 04_1, draws the escapement in angles. At rest the impulse stone’s face stands 28°10′ before the line of centres, the line through the balance’s and the wheel’s arbors: the sheet gives it as 6° on from where the impulse begins, 22°10′ before that line. Its other angles are the wheel’s: a tooth’s pitch of 24°, a drop of 1° onto the stone, 21° of impulse, 10.5° either side of the line of centres, and 2° on to the lock, which puts the locked tooth 36.5° round the wheel from the line; the locking stone itself scales 37.0°, 7.14 mm from the wheel’s centre, so the teeth lock 0.235 mm deep. One more angle stands on the balance’s arc beside those two, 5°, running on from where the impulse begins toward the line of centres: the tooth is let go with the impulse stone already 5° inside the teeth’s path, so the stone is surely there when the tooth comes. The layout then has the tooth on the stone at once, as if the wheel were there the moment it was let go.
A wheel let go from rest can’t do that. Solved from the parts’ shapes and masses (the escapement solver the Hamilton’s model uses, with the Wempe’s settings), the gold spring frees the tooth with the balance at −2.4° (the lock is 9.9° of swing, the let-off 1.9°), the impulse stone by then 5.0° inside the teeth’s path, as the sheet has it. The wheel starts from rest, with the seconds arbor behind it, on only 82.8 µN·m, and gathers speed slowly, while the stone moves on ahead of it. It strikes the stone 16.1° of swing after it was let go, the stone’s face 10.3° before the line of centres, and drives it to 21.8° past: the wheel is on the stone for 12.3° of its own turn, where the layout draws 21°. The strike costs 20 % of the impulse, and the unlocking takes 17 % of its work. Integrated in time, beat by beat, with the balance, the wheel and the train behind it and the detent moving as bodies through every contact, the tooth lands where the solver puts it, within the 1.5° that check allows.
Taken in time, two more things show that a layout can’t. The detent leaves the discharge stone at −1.8°, flung up off its square corner rather than riding down its bevelled end, and the energy it carries there is lost. And the wheel, let go from rest, has hardly moved by then, so the detent falls back onto the tooth it has just let go, at 13.3°, and rides it, and down its back, until it is home at 19.8°, its friction on the tooth holding back the wheel’s chase. Both cost swing, and the model’s escapement takes both in.
The model’s balance swings 207° twelve hours after winding. No Wempe source gives how quickly the balance loses its swing, so its free run-down time, 125 s, is worked out from its parts (the air on its masses and screws, the endstone’s friction, the spring’s own losses); a running 6MX’s balance, read frame by frame off a video, runs down in 118 s (105–129 s) at the model’s swing: a little more slowly than the model, which keeps the figure an earlier reading of it gave, well inside what its parts allow. Over the wind the model swings 155.1–165.8°. The 1954 Soviet instruction for checking chronometers gives 225–270° for chronometers in general: the model swings well under it. The late strike is why: let go where the sheet has it, the wheel reaches the stone only a few degrees before the line of centres and drives it through little more than half the drawn impulse. On the measured drive, no damping the balance’s parts allow brings the model to 225°. Either the Einheits-Chronometer swung less than that instruction’s chronometers, or something the model takes from its sources is off: the drive’s level (the fusee section’s open figure), what the wheel has to start at each beat, or the reading of the sheet’s 5°. It is open.
The spring detent, step by step, in the Hamilton’s essay
The escapement in the working model, seen through the top plate
Brass that hangs
A chronometer is adjusted to run dial up, so its case hangs in gimbals and stays level while the ship rolls: a ring pivoted in the box’s walls, and the case pivoted in the ring on the axis across (05 01, “Gehäuse Kardan”). Hung low, the case is a pendulum, and stays level as long as the ship rolls far more slowly than it swings.
The bowl’s sheet, R-05 01 00-1 of 6 March 1942, names its metal, “Ms-Sandguß”, a brass sand casting, and its weight: “Gewicht 350 gr.” The two don’t agree. The bowl’s drawn solid is 139.6 cm³, which in brass weighs 1,179 g; 350 g would be 2.51 g/cm³, which is no brass. In aluminium the bowl would weigh 377 g, in Bakelite 191 g, in hard Buna rubber 165 g. Weighed on the model’s solids, the case with the movement in it is 2,467 g and the gimbal ring 404 g: 2.87 kg in the gimbals. The 6MX’s conditions allow their whole chronometer 2.2 kg.
The drawn pivots don’t decide it. They lie about 5 mm under the seat on which the movement ring’s flange rests, with the movement and the heavy brass bezel over them, yet a light bowl would hang too: with a Bakelite bowl the case would weigh 1,479 g, its centre of gravity still 7.23 mm under the pivots. The sheet’s “Ms-Sandguß” names brass, and the model is brass. What the 350 g means is open; Knirim and Schmiedchen list a Bakelite bowl with a heavy weight in its bottom among the changes made at Glashütte from No. 5100.
The sheet also says how to level it: “Kloben verstellen bis Gleichgewicht erreicht ist”, move the brackets until it balances. As drawn, the case’s centre of gravity lies 14.27 mm under the gimbal centre and 0.88 mm to one side of the pivots’ line, so it would hang 3.5° out of level until the brackets were moved in their slots. Levelled, the case swings on its pivots with a period of 0.67 s, and the ring, carrying it, on its own 0.76 s: far quicker than a ship rolls, ten to thirty seconds for a merchant ship, and clear of the balance’s half second. The Hamilton’s case swings in 0.85 s and 0.92 s: the Wempe’s centre of gravity sits lower under its pivots, 14.3 mm under them against 12.8 mm.
Gimbals, and why a case kept level keeps time, in the Hamilton’s essay
The case in its gimbals in the working model
The reparation drawings and the 6MX
The drawings held at A. Lange & Söhne became the basis of the post-war Russian marine chronometers. Johannes Altmeppen, quoting Kurt Herkner, writes that two specialists from Moscow and an interpreter came to Glashütte in 1946/47, and the drawings were translated into Cyrillic. The 1st Moscow Watch Factory made the chronometer from them as the 6MX. Its movement has three pillars because the reparation drawings were of that type; the factory made about twenty changes, of which there are no drawings.
Altmeppen compares a Wempe with a 6MX and marks the differences. In the 6MX pillar 39 is in another place; the seconds and intermediate wheels run in jewels in settings, not in brass bearings; the barrel’s cap is underneath; the detent’s block is new, with an adjusting slide; the lower balance cock has another shape; the upper cock has no window, and the three small windows by the pillar screws are gone; the bowl and bezel are brass, not Buna; the dial’s feet are at 19, 39 and 59 minutes, not 20, 40 and 59; and there is no hand-setting from outside.
The 6MX’s technical conditions, TU 25-07.1533-84, give its movement as 85.0 mm across and 39.8 mm high, with a balance of 0.5 s, 56 hours on the wind indicator and 15 jewels, the whole at most 2.2 kg. They set the trial each chronometer had to pass: five periods of seven days, at 36, 20, 4, 20 and 36 °C, the last five days of each counted. That is the class II trial of the Soviet standard GOST 8916-58, with limits of the TU’s own between the standard’s two classes, and two more figures: the middle-temperature error S, and a figure of merit N.
| 6MX (TU) | GOST 8916-58, class I | class II | |
|---|---|---|---|
| E: the daily rates’ mean departure from their period’s mean, s | ±0.35 | ±0.20 | ±0.40 |
| R: the rate’s recovery, the last 36 °C period’s less the first’s, s | ±2.00 | ±1.50 | ±2.00 |
| D: the largest change between two days’ rates, s | ±2.30 | ±2.00 | ±2.50 |
| C: the temperature coefficient, s a day per °C | ±0.10 | ±0.10 | ±0.10 |
| S: the middle-temperature error, s | ±1.20 | — | — |
| The mean daily rate in any period, s | ±3.50 | ±3.00 | ±4.00 |
| N = 42E + 4.5R + 60C + 2.4S | ≤ 30 | — | — |
| The trial, periods of 7 days | 36, 20, 4, 20, 36 °C | 36, 28, 20, 12, 4, 12, 20, 28, 36 °C | 36, 20, 4, 20, 36 °C |
The copies are evidence, not the source. Where a 6MX or a post-war Wempe differs from the drawings, the model follows the drawings.
The model also builds one 6MX, as a third version beside the sheets as last revised and as first drawn (Variants, Version): Kirov No. 247 of February 1952, from Altmeppen’s photographs of it and of other 6MXs of the 1950s, each put on the sheets’ hole table. Building it settles most of his list. On its top plate pillar 39 stands at 190°, not the drawn 210°. The intermediate arbor’s pivots and the seconds arbor’s upper one run in jewels in flush gilt settings, which with the drawings’ twelve make the technical conditions’ fifteen. The escape wheel sits just under the top plate, where the sheets first drew it, not where they were revised to, on one long cone. The new detent block is held up to the top plate by a slotted stud, 2 mm from where the drawings put its screw, through a gilt slide in a notch at its corner, and a large screw in its side stands over the detent’s foot; how the slide adjusts it isn’t seen. The lower balance cock is cut back to a third of its depth except under its lug. But the three small windows by the pillar screws are not the auxiliary holes 3, 12 and 40: Kirov 247 has those, and what he meant isn’t found. Nor is the bowl the drawings’: where they draw a flange of 118 mm, it has a threaded rim of 110.5 mm, onto which a knurled bezel screws, its glass thick and standing proud, and on higher brackets it hangs 13 mm lower in its gimbals.
Kirov 247 also carries what the drawings don’t: an electrical seconds contact, for sending its seconds to a ship’s instruments, which the technical conditions list as a second version, with twenty jewels to the other’s fifteen. A wheel of sixty ratchet teeth, one of them left out, turns with the seconds arbor. On every other beat a tooth lifts a lever’s stone and the lever’s screw leaves a knurled disc, so the circuit opens for half a second each second; at the missing tooth it stays closed for a second and a half, which marks the minute. Two springs in the bowl carry the current to sockets marked + and − outside. Romanov’s textbook of 1975 draws the device and Altmeppen photographs it; the lever and its staff, which no photograph shows, follow the book.
The jewels change less than one might think. The friction the model’s sources give oiled steel in ruby and in brass is the same, so the train loses no less. What they give is a closer fit, the pivots’ shake down from up to 0.05 mm to 0.025, and so more room in placing the holes: at the seconds wheel’s mesh a maker must hold them within ±25 µm on the drawings’ bushes and ±35 µm on the 6MX’s jewels. Its escapement, at the first-drawn heights with the 1941 collar, swings 0.1° less than the drawings’ does, since the contact wheel on the seconds arbor adds to the train its escape wheel must start at each beat, 155–166° over the wind; its balance, lighter by the hub’s screws, runs 931.2 s a day fast until its timing weights are set. The model runs the 6MX on its own escapement; its rate stays the drawings’ to a hundredth of a second a day.
Kirov 247 travels in a box of orange lacquered wood in three tiers, 190 mm each way by the technical conditions, closed by two hasps, the factory’s diamond badge on its glazed lid and its number on a plate below, and that box in an outer one whose lid closes into a notch in its front.
See it on the other tab: every part can be picked out and named, with the sheet it comes from.
What the drawings leave open
A set of works drawings records what was meant, and here and there what was measured. Where its figures disagree with each other, or with what a model built from them does, the model says so rather than choosing quietly. These are the questions the drawings leave open, as the model’s README lists them, with the reading the model takes for each.
- The mainspring’s length. No sheet gives it, nor its free curve: solved in the drawn barrel and wound through the drawn fusee, 15 × 0.37 mm strip gives 05 06_20’s measured force, its shape and its level, at 1,240 mm to about ±50 mm. The model’s reading: the measured force drives it; a made Wempe’s strip is chosen on the bench by the balance’s free run-down. The fusee
- The swing. No source gives the Einheits-Chronometer’s. The model’s reading: 155.1–165.8° over the wind, under the 225–270° the 1954 instruction gives for chronometers in general; with the damping borne out by a running 6MX, the drive’s level is the likelier cause. The escapement
- The period. Read as 05 03_7’s words read, the hairspring gives the weighed balance 0.491 s, not the train’s 0.5 s; the drawn coils lean a turn longer, and the detent chronometer’s own curve sheet isn’t in the set. The model’s reading: the whole strip, 13 turns less 3.5 mm, the balance timed to 0.5 s as a maker would. The hairspring
- The metals’ constants. The rim’s nickel steel is named only as a special alloy, and the spring’s change of stiffness with heat isn’t given. The model’s reading: Guillaume’s steel and a published range for the spring; by them compensation lies near 19°, where Wempe 3955’s adjusted masses stand at 30°. Heat
- The detent’s preload. The banking is the block’s cross screw, its domed head bearing on the pipe at the stone (read on a 6MX’s block of the same form: likely); how hard the detent presses on it isn’t shown. The model’s reading: 0.80 g at the stone, an estimate between two limits the tools find: from 0.64 g a knock of 20 g across the plate lets no tooth past, and from 0.85 g the stone, falling home, catches the escaping tooth’s back. At it a held tooth’s stone, shaken by the gold spring’s return, rings a little over half the depth of lock and no tooth gets past; the 1954 instruction sets the lock by a banking.
- The gold spring’s return pass. Whether the gold spring’s release on the return swing throws the detent turns on its alloy and its free shape round the pipe. The model’s reading: 333 gold, its tip square; the horn 4.4 mm, 1436’s figure, which Wempe 3955’s detent bears out.
- Chain or band. 05 07 04 sizes the chain completely, and the German movements photographed run with chains; 05 07 03 draws a band in 1943, without its width, and a band came with Wempe 3955. The model’s reading: both, the chain by default (the band under Variants on the other tab), the chain run across between the drums. No photograph shows the crossing itself, hidden behind a pillar, but a fully wound Soviet copy seen from above has its chain’s last hook on the barrel where only a crossed chain leaves it. The chain in the working model
- The bowl’s 350 g. The brass casting the sheet names weighs 1,179 g. The model’s reading: brass, as the sheet’s material says (a lighter bowl would hang on the drawn pivots too). The gimbals
Settled since the first readings, from the sheets and photographs of real ones: the masses’ height and place, the horn’s length, the banking, and the mainspring’s 0.37.
The model’s README lists every estimate with the reading it rests on, and each part’s card on the other tab says whether its sizes are the sheet’s, scaled from it, or estimated.
Every check behind the figures in this essay, its method and its results, from the geometry against the drawing list to the escapement in time, the fusee, the trial and the swing of a running 6MX, is in the model’s report: The Wempe Einheits-Chronometer: does the model hold to its drawings, and would it keep time?
Found a mistake, in the essay or in the model, or something that differs from Wempe’s drawings or a real Einheits-Chronometer? : it opens as an email to me, with the section you are reading filled in.
Sources. Wempe’s works drawings, as published by the Deutsche Gesellschaft für Chronometrie (dg-chrono.de), © Johannes Altmeppen, by courtesy of Gerhard D. Wempe KG, 2017, with Altmeppen’s photographs of Wempe 3955 and Lange 5177 and his comparisons with the 6MX. Das Deutsche Einheits-Chronometer (Heel Verlag), with Johannes Altmeppen and Herbert Dittrich among its authors. Knirim and Schmiedchen on watch-wiki. Хронометры морские 6МХ. Технические условия, TU 25-07.1533-84, from the 1st Moscow Watch Factory; Instruction 246-54 for checking standard chronometers (1954); V. A. Shpolyanskii and A. M. Kuritskii, Spuskovye regulyatory priborov vremeni (Moscow 1963). For the physics: E. Phillips, on the balance spring’s terminal curves (1861); S. Timoshenko, “Analysis of bi-metal thermostats” (1925); C.-E. Guillaume, Nobel lecture (1920); VintageWatchstraps, “Temperature Compensation by Nickel Steels”. The working model’s sources are listed under “About this model” on the other tab.