The Wempe Einheits-Chronometer model · verification · 12 November 2026 · the Hamilton's: would a chronometer made to it run?, the model against real Model 21s

The Wempe Einheits-Chronometer: does the model hold to its drawings, and would it keep time?

The model of the German standard marine chronometer of the 1940s is built from Wempe's works drawings, each sheet as last revised. Two questions follow from that. Does the model hold to its drawings: is every part a closed solid where its sheets put it, clear of its neighbours, counted against the drawing list, every figure read and every guess said to be one? And would a chronometer made to those drawings keep time: does the escapement run, does the fusee even the spring, does the energy budget close, do the teeth and springs hold, does it stay in rate from cold to hot and at sea, and does it pass the trial its Soviet copy, the 6MX, was made to pass? Every figure below comes from a script in the repository, run on the model's own numbers, and is shown as the script prints it: ok a check that holds, ~~ one the script reports but doesn't fail on (an open question, or the strong end of a range only), !! a failure.

Approach

The Wempe's checks are the Hamilton's, pointed at the second model (--model wempe), with tools of their own where the two chronometers differ: the Wempe's detent carries a gold passing spring on its own horn, its fusee's force was measured by its makers on the drawing itself, its balance is a cut bimetallic one with movable masses, and its hairspring is a helix with drawn terminal curves. The tools run on the same files the page draws from (js/escapement.js, js/escphys.js, the shared makeEsc), so what they check is what the page shows.

#QuestionMethodScript
1Is every part a sound solid, clear of its neighbours, and on the drawing list?Closed-mesh tests, a 0.05 mm voxel interference check through the escapement's phases and the wind, a screw and loose-end audit, the Spread view's clearance, every part against bom.jsonsolids.py, fine.py, audit.py, exploded.py, bom.py
2What do the drawings say, and what had to be guessed?The sheets transcribed figure by figure with how sure each reading is; the readings corrected on the sheets; the estimates listedReferences/wempe/, the model's README
3Does the escapement run, and does it run in time as the averaged model says?makeEsc solved from the layout drawing; the escapement integrated beat by beat; the detent as a frame of beam elementswempe_escapement.js, dynamics.js, tripping.js
4Does the drawn fusee even the mainspring as its makers measured?The spring solved for its shape at every quarter turn, fitted in length and set to 05 06_20's forcewempe_mainspring.py
5Is there energy enough, from spring to balance?Every loss from the parts, the balance's damping from its parts; the swing the drive can holdwempe_physics.js
6Do the gears mesh, and how exactly must the holes be placed?Tooth-by-tooth contact on the page's own outlines; the train in time; each mesh's depthing windowgearing.js
7Will the loaded parts hold, fully wound?Each part's stress against its material's published rangewempe_strength.js
8Does the drawn hairspring keep the balance free of side forces, and isochronous?Phillips' conditions on 05 03_7's curve; the spring at large swings as an elastica and as beam elementswempe_hairspring.js
9Does the bimetallic balance compensate, and where must the masses sit?The rim's quarters as bimetal cantilevers, every part weighed; the secondary errorwempe_thermal.js, wempe_thermal_frame.js
10Does the case hang level and stay so at sea?The case and ring weighed on the model, levelled as 05 01 says, run in a seawaymassprops.py, gimbals.js
11Would it pass the 6MX's trial, and what would a navigator carry?The TU's 35-day trial and a 60-day voyage, Monte Carlo over the rate's partswempe_errorbudget.js
12Would a Wempe made to its drawings work, and how is it tested?The unsized parts' tolerances from the escapement's swing over the wind; the bench's tests with their pass figures; 2,000 movements made to tolerances, as made and after the benchwempe_tolerances.js, wempe_bench.js, wempe_montecarlo.js
13How does the escapement compare with a period text's rules?Each proportion against Watch and Clock Escapements (1904), Ch. III; reported, never judgedkeystone.js
14Is the balance's damping right?A running 6MX's swing read frame by frame off a video, fitted with the Wempe's balance equationwempe_swing.py, wempe_startup.js

Units: millimetres and micrometres (µm), newtons, seconds a day (s/d, + gaining). The swing is the balance's amplitude, each way. The model runs its balance at the train's half second; the drawn hairspring with the weighed balance gives a little less (section 8), which a maker times out. A rate of 4 s of time is a mile of longitude at the equator. Each tool's whole output is under its checks ("its whole output"), as it printed it when this page was built.

1. Geometry: does the model hold to its drawings?

The model is a set of closed solids, as a CAD model is, so it can be checked as one. Five browser checks open the Wempe's page itself and measure the parts as built. Solids: every mesh closed (no open edge), consistently wound and with a positive volume, as loaded and in a section with Moving parts only. Fine interference: every part voxelized at 0.05 mm in states (the escapement through its unlocking, impulse and return; the train at many phases; the fusee from full wind through winding to run down), every pair of parts that overlaps reported; only the contacts listed below, each with its reason and a size it may not exceed, may overlap. The screw audit: screws overlapping, screws with nothing under their heads, arbor and pin ends sitting in nothing, faces lying on each other, parts touching nothing, in the movement and in the case and box, against the leftovers expected. Exploded: the Spread view lifts each part on its own path; no two may pass through each other. The parts list: every line of bom.json, transcribed from the drawing list, counted on the model's pieces, and its relations measured on the built meshes.

The contacts that are meant

Every overlap fine.py finds is one of these, each a joint the drawings make that a turned or extruded solid can't show cut: a screw in a thread not cut across a turned body, a pressed fit drawn at its fit's middle, the chain in its groove. The measured column is the largest overlap over all the states, against the size it may reach.

The audit's expected leftovers

The parts list

2. The sources: what the drawings say, and what was guessed

The model's evidence is, in order: the works drawings (sheets of Wempe's, as published by the Deutsche Gesellschaft für Chronometrie, each read as last revised); the coloured section of the whole movement; photographs of real German movements (Wempe 3955 above all); and, for what only the Soviet copy shows, the 6MX's own documents. None of them is reproduced here: they are other people's work, cited and linked from the model's page. The model's README says for every figure which sheet it is from and how sure the reading is; below is that record in numbers.

The sheets

Figure 1. Each group's sheets by date: when each was drawn (dot) and, where the list gives one, last revised (ring). The model builds every sheet as last revised; 05 03_3.2, a later production staff of 1965 (not built), lies off the scale at the right; the lever chronometer's sheets, used only to cross-check, are left out. From References/wempe/README.md's catalogue (wempe_essaydata.js).

The transcription

Each group's sheets were transcribed into a table of every figure, with its tolerance or fit and how sure the reading is (sure: legible, likely: legible with one reading taken over another, unsure: faint or partly hidden). Where a number mattered, the sheet itself was cropped and enlarged, and the readings were corrected there (below).

Figure 2. The figures read in each transcription's tables of dimensions, by how sure the reading is. References/wempe/transcriptions/.

Readings corrected against the sheets

Checked on the sheets while building, where the transcription read a figure or a feature wrongly ( readings; the transcriptions are left as written):

The readings corrected

The estimates

What no sheet gives is estimated, and every estimate is listed in the model's README with the reading it rests on ( in all). Grouped by the phase that built it:

GroupEstimates
The frame (P1a)41
From the parts audit (3 November 2026)3
The train (P1b)11
The fusee and barrel (P1c)12
The escapement (P1d)15
The energy budget (phase 2b)5
The escapement in time (phase 2a)9
The escapement in time (phase 2d)5
The transmission (16 October 2026), tools/wempe_transmission.js8
The hairspring and temperature (phase 2c)5
The dial, motion work and hands (P1e)12
The case and box (P1f)12
The case, gimbals and the trial (phase 2d)5
The Soviet 6MX (the Variants' third version, Kirov No. 247; The Soviet 6MX above). Measured on Altmeppen's photographs, each fitted on the sheets' hole table; estimated or chosen where they don't show17
Left out, small1
In all161
The frame (P1a): 41
  • The key's paddle in brass (9 November 2026): the default key's paddle brass with its body, as The Mariner's Chronometer, Fig. 2.35's bare brass body; 51 00 11 gives the paddle no material and no photograph read for it settles its colour.
  • The fusee's pins (28 October 2026, the second drawing check's batch 5): the Vorreiber's pin (05 06 00-5, no sheet) Ø 1.16 in its holes reamed together Ø 1.15, its size reconciled from the hub's 0.8, the arbor's 1.1 taper-reamed and 05 06_1's 1.2 (the arbor, above), its 1:50 taper not drawn, its 8.3 length scaled on the coloured section (±0.15); the chain's pin (05 06 00-6, no sheet) pressed Ø 0.567 in the ribs' 0.555; the fusee wheel's bore 4.55, 0.04-0.06 clear on the 1:70 seat, which neither its "4 +0,1" nor 05 06_1's straight 4,45 H7/g6 can be. The dust plate 1.0 from the square's root as 05 06_1 figures it, its hole at 3.110 there (within its H7); at the fits' middles it would sit 1.3 off, and the coloured section draws it on the root.
  • The navigator's voyage (Navigating by it, wempe_navigation.js): the route, from 45° N in port to 15° N by day 20 where the box reaches 85 °F, is estimated (no source gives one; the latitude only scales miles of longitude by its cosine, and the equator's figures are given beside it). Bowditch's probable error of a fix (±2 miles) is taken as a circular normal whose median radius is 2 miles (σ 1.70 a side); Bowditch gives the figure, not its distribution.
  • The detent's knock margin: since 19 October 2026's physics pass, with the estimated 0.8 g preload, a 3 ms blow across the plate of 25.6 g lets a tooth go on the frame as drawn (22.6 g at 0.8 times, 28.7 g at 1.2 times; the knock's floor under 0.64 g, below the margin's 0.78 g, which the bench takes). Before it (18 October 2026), with the estimated 0.63 g preload, a 3 ms blow across the plate of 20.4 g lets a tooth go on the frame as drawn (20.0-20.1 g before the horn was traced) (tripping.js --model wempe, its least-knock search refined from six halvings to ten: the coarse one read 22 g). The check asks 20 g or more, so the margin is a hair at the estimate. The model keeps the estimate; the bench takes the knock's floor, 0.61 g, as a second least beside the margin's 0.51 (the user's choice), its window 0.61-1.55 g aimed near 0.84 g, where the 3 ms blow wants about 25 g (the line through the frame's runs carried on past 1.2 times: estimated).
  • The key holder (case.js, Schluesselhalter, Schluesselhalter.F; 18 October 2026): the box's plan, 127-556. A-1 (51-00_1, 1:2, the Boden-Chronometer's arrangement, a lever chronometer's, borrowed for the Einheits box; scaled 1.24-1.25 px/mm by the box's 185 and the gimbal ring's Ø 146.6), draws it: a triangle in the back right corner, its legs 22.3 along the two lines it is drawn against (2-4 mm inside the box's inner faces as the box's centre places them, taken as the walls), and on its long edge a round fitting, the circle fitted on 8 rim points (0.4 px rms): Ø 19.5 ±1, its centre 10.35 from each of those lines (13.4-14.5 from the inner faces by the box's centre) and 1.1 inside the long edge, with two dots on the box's diagonal 11.9 apart (about Ø 2.4). Lange VEB 2285's box (drouot_33246921_langeVEB2285_3, from above) shows the same arrangement: a wooden corner shelf and the key's body in a round black cup standing over its front edge. Built (the user's reading, 18 October 2026): the key's axis 14 ±1 from each wall (16.5 until then, estimated, and 21 before 16 October 2026; 13.5 would leave the paddle 0.24 from the walls). The shelf's front edge (the owner, 24 October 2026) is a concave arc following the gimbal ring, as Wempe 6161's (antiquorum_208-4_wempe6161) and Wempe 3044's (milivault_529241_wempe3044_screen) boxes show it, drawn about the box's centre at r 82.5 so the Ø 19.5 cup stands wholly on the wood, 1 inside it (legs 59.8 along the walls), the key's tube through a hole r 6, 10 thick; until then the plan's straight edge 1.1 beyond the key, which left the cup half over nothing and the key's tube in an open notch. Estimated: the cup's height 5 and black finish (Lange 2285's; the drawing gives neither), its Ø 11 bore (the key's cone seats on its top edge, 65 over the floor, so the 1943 key's paddle stands 20.7 over the box's top as before: on the wartime Wempe, rmg_L5393_wempe_wartime_001, 24 ±6, 20 ±8 read apart); the screws 14 apart with heads Ø 2.2 on Ø 1.2 shanks, both into the shelf (the dots' 11.9 and Ø 2.4 would put the heads over the key's bore, so they are built as far out as the reading's ±1 allows), the shelf's thickness, its front arc's radius and centre (3044's shelf scales about 45-50 along the walls, ±15 %, against the model's 59.8); that the drawn corner lines are the walls. The paddle (Ø 34.5 × 3, facing the box's middle) clears the walls and the closed lid's by 0.74 at its widest, the 1940 key's pear by 0.57 (measured on the page). The 6MX's shelf (legs 65, its hole at the same place) stands 15-30 under the rim with the paddle above (moMvyx8eXfo 3:06, a 2022 box, empty; 12_Kirow_DEC_422_Kasten_1); Wempe 5377's, in a later case, under it. Of the box's wood.
  • The page's ship motion (The page's time, ship motion and reports): the sea itself (14° of roll each way, pitch 0.55 of it, the axes 5 m below the box) is the Hamilton's illustrative motion; the gimbals' damping ζ 0.05 is the Hamilton's estimate; the time the transit lock takes to level the gimbals (about a second) and the twist that restarts a balance stopped to set the hands (the Twist button's 0.6 of the running swing) are estimated. The gimbals' periods are the model's own (weighed).
  • As first drawn (phases E1, E2, above): the 1941 barrel's seats (their mouths Ø 39.25 and 39.2: 020701-3's 1.5-deep seat with its 5° undercut taken as faced 0.3 to 0207's flush 1.2 when the ring goes from 19 to 18, the lid's 3° seat 1.0 deep, each disc's edge turned to its seat's cone: 28 October 2026) and its boss's cone (020701, likely); the 1941 lid's oil sink Ø 7.4 (0207 draws it undimensioned: scaled against the bottom's dimensioned Ø 8, at the bottom's 10°) and its prising notch's size and place (020701's plan, scaled 2.5 × 1.9, half-round; opposite the Steg as that plan is turned onto 0207's, both seen from the bridge's side: likely); the 1941 arbor's journals Ø 4.48 in the discs' Ø 4.55 and its pivots 2.88 and 3.88 (the 1948 drum's fit and the frame's reamed holes: 020704-1 and 0207 draw the journals Ø 5 ±0.02 and the pivots Ø 3 and Ø 4, 0207 the discs' bores Ø 5, while 020701 draws them Ø 4.5 as 05 07 01-1 and 05 07 01-2 keep: the 1941 sheets disagree among themselves and none can be shown to rule, so the 1941 arbor's diameters are an estimate, not 020704-1's: the second drawing check, 4-M6); the 1941 drum's chain slot's place along the axis (0207 draws it undimensioned, its middle 4.6-4.7 from the bridge-side face; cut where the laid chain's barb comes, 4.04 from it: 0.6 toward the bridge, 4-M13); the maintaining ratchet's tips Ø 37.45 (05 06 02-1's "war 37,5∅", its tolerance not kept: the revised −0.1 taken); the maintaining spring's width past 270° to its square end; the 1941 staff's turned form (its sheet a blank: the ends, the collet's and rollers' seats as 05 04 02's); the collar's neck drawn flush with the arm (riveted); the escape arbor's cone to its upper pivot and the collet's cone top (scaled); the escape jewels' settings for the 1941 arbor's ends; the stone's pipe from 1.95 to 3.65 under the plate; the block's arm, cove and ledge (scaled on 05 04_1, its width and place from the photographs); the banking screw's size (Wempe 3955's photograph) and its near end in the arm's face; the Exploded view's offsets for the 1941 parts.
  • The pillar plate's hand-setting cut: the window's place across the cut, centred on the line through 18 (Wempe 3955's 0.2 off it, within its fit), and the inner face's 12 +0.2 drawn 12.1 on both ends alike; the end arcs' bow, read on the scan at 2-3 px; the bush in 18 (no number on 05 08 02, its bore and sink from Wempe 3955's photograph). Its bore across the plate and the rim's notch not built (How each frame part was measured).
  • The pillar plate's thickness, 3.35: its sheet's "3,35 ±0,02 stark" is struck through with a replacement that's no longer legible.
  • Screw heads and lengths (no WN sheets in the set), measured on the coloured section and Wempe 3955: pillar screws (WN-57; the dial side's unlabelled) head Ø 6.4 × 3.0, shank 7.7; movement ring screws Ø 6.2 × 3.05, 6.3 (2.25 high until 17 October 2026); barrel bridge screws WN-55 Ø 6.4 × 3.0, 6.2; the click's screw WN-22 head Ø 4.5 × 2.1, shoulder Ø 3.19 × 2.61 (a close running fit in the click's 3.2 +0.03, 0.025 clear: estimated; Ø 3.1 × 2.65, scaled on the coloured section, until 28 October 2026: the second drawing check's 8-M6, 8-M8), thread 3.2; the fusee bush's WN-34 Ø 2.1 × 0.45, 3.2 (ending 0.4 short of the recess; 3.6 until 17 October 2026); the upper cock's screw WN-55 Ø 6.4 × 3.0, 6.2 (blued, as Wempe 3955's); the lower cock's WN-53 Ø 5.8 × 1.7, 5.0; the intermediate bridge's WN-52 Ø 4.2 × 1.9, 4.7. the balance's setting screws (upper WN-55, lower WN-18) Ø 2.5, the upper 1.3 high, the lower 1.4 (Wempe 3955 gives the upper ones' Ø 2.5); the stud's screw WN-11 Ø 2.3 × 1.6, through the cock to 0.6 under it; the hub's screws WN-10 countersunk Ø 1.8 flush in the arm's top (17 October 2026). Their WN numbers are the coloured section's labels where they could be read (the detent foot's screw WN-13, the block's WN-15, the wire spring's stud WN-27: likely); the ring screws' WN-56 on the general arrangement's print (below). The section's print in Klassik Uhren 3/2006 (31 October 2026, below: Sources) reads the labels the coloured scan left open: the upper setting screws' "Schraube WN-55" (sure) and the lower ones' "Schraube WN-18" (sure: its leader ends in the head under the lower setting), and "Schraube WN-56" on the movement ring's screw (its leader's dot in that head, on the print and on the coloured scan alike: sure), the dial side's pillar screw unlabelled; renamed so (Closed).
  • Hole 4's oil sink on the top plate's outer face, Ø 4.3 spherical (its depth 0.95, two cones), from Wempe 3955's photograph (no sheet draws it; the second drawing check's 9-M11; also under "From Wempe 3955's photographs").
  • The escape settings' screws' bearing: the WN-es heads (Ø 2.6 at 3.525) lap the upper endstone setting's rim by 0.07, the WN-ec (Ø 2.6 at 3.525 on the 37° line) the lower cap's by 0.075; no sheet sizes the heads or the settings, and Wempe 3955's top plate shows its heads (2.5 across, 6.94 apart) lapping a Ø 4.5 setting by about 0.03 (9-M14): a sliver, as on the real one.
  • The pressed pieces' interference: the bushes WN-300, WN-301, WN-302, 05 08 02-2 and WN-hs, the seconds jewel's setting 05 05 04-2, the motion work's studs, the stop pin 05 08 01-5, the ring's pin WN-202, the dial pin and the plate's pin WN-201 are pressed (their sheets hatch or call them so) but no sheet sizes an interference: each drawn 0.01 over its hole on the diameter (line to line or a few µm loose until 28 October 2026: 8-M11, 9-M7). The plates' pillar seats (0,3 +0,03) and 26's lip (0,5 +0,05) stay at their least: at their middles the plates would stand 18.97 apart and every arbor's endshake change (9-M10).
  • The stud's screw hole in the upper cock (05 08 03-1 draws none): M1.4 tapped through the arm's thin end, 4.1 along from 23 toward the foot and 4.2 across toward O, from the lever chronometer's same cock 127-556.04 U30, its only figures (the owner's choice, 28 October 2026, O-8); Wempe 3955 reads 4.22-4.33 along and 4.25-4.36 across (2_Wempe_3955_A.jpg, ±0.15), which can't tell 4.2 from 4.35. The stud built off-centre about it as 05 03_9's plan draws it (scaled: likely). The balance's setting screws are named WN-55.ck after the coloured section's label "WN-5?", whose last digit (5 or 6) its 5-6 px digits can't settle (the second drawing check's 8-M14: one WN number for the upper cock's M2.6 and these M1.2 is unlikely, so one label differs; undecidable). The section's print in Klassik Uhren 3/2006 reads it "WN-55" (sure, 31 October 2026): the name stands.
  • The threads' form (26 October 2026): the sheets give each thread's size (M2.6, M2, M1.7, M1.4, M1.2), not its form or fit, and no photograph in the set shows a loose screw's thread. Every screw's shank is drawn as a right-hand ISO metric thread at its size's coarse pitch (ISO 261/262, DIN 13; M1.7 0.35, M2.6 0.45: KIT.isoP), ISO 68-1's basic profile (60°, the depth 0.541 p, flats p/8 and p/4), at 6g's maximum material (its major diameter less es = 15 + 11p µm: an allowance of 7.5 + 5.5p µm on the radius), run out to its root over the pitch before its 45° tip. A size the sources leave estimated keeps its estimate (the case's M3-M6, the 6MX's contact's). The tapped holes of the frame (the plates', bridges', cocks' and bushes' tables, the pillars' ends, the detent block's M1.7 and the wire spring's stud's hole 5) carry the inside thread at 6H's minimum (the basic profile), running out over a pitch to the drilled hole at each face, so screw and hole are the allowance apart radially all round; each screw is turned on its axis to lie on its hole's helix (KIT.matchThreads: 43 screws in all the versions). Other tapped holes (the movement ring's, the dial screws' in its flange, the fusee's, the balance's masses', the case's and the 6MX's contact's) stay plain bores at the thread's size and 0.02, the screws in them threaded. Standard practice, not read off a Wempe: sure of the sizes, estimated in form and fit.
  • The 6MX's dial screws' sinks (7 November 2026): Kirov No. 247's and No. 2973's dials show a dark ring round each screw's head, its inner edge the head (Ø 1.64-1.88, the slot across it) and its outer edge onto the paint Ø 2.21-2.43 (6mx-audit/gamma.md, dhead.py, each dial at its arbors' scale: likely). Its form isn't seen: the 6MX's sink is drawn as the drawings' Ø 1.8 opening in a 45° chamfer to Ø 2.3 at the face, 0.25 deep (estimated: the more conservative of a chamfer and a wider bore); the drawings' sink (05 02 00-4) is unchanged. In the same audit the 6MX box's badge was centred in its tier (Kirov 247 and 2973) and the outer box's catch drawn at its own reading's Ø 26 (Kirov 247's single oblique frame reads about 29: doubtful).
  • Steady pins' length, 3.0 (1.7 in the part, 1.3 out): U30 draws the upper cock's so; the others are taken the same. The cocks' pins carry no WN number on their own sheets; the coloured section labels the upper cock's "Stellstift 4 WN-203" (the second drawing check's 8-M6), and the lower cock's are taken the same.
  • The movement ring: the lip's depth under the seat (2.0) and so where the cone starts; the M2.6 holes at the thread's size through the band and the flange to the cone (the sheet sizes only the M2.6; tapped that far or drilled clear above the thread isn't drawn), with a 5 µm step 0.3 into the flange where the cut starts; the plate's pin 5.35 long with its top 0.1 under the plate's face (the coloured section draws it 7.6, which would cross the cone); its material brass ("——" on the sheet; the coloured section colours it yellow, GUB 12255's is gilt; weighed at brass's 8.45) and its finish (nickel, as Wempe 3955's looks).
  • The pillars' countersinks (Ø 2.6 at the face, 90° to the tap drill: 2.59 scaled; the thread cut through them) and relief grooves (out to r 3.25, 0.35 deep, half-round): drawn, not figured.
  • The balance opening's tongue in the top plate, 2.5 wide ("2,5": sure): its axis toward 113.5°, passing 0.85 S of the opening's centre, its round end r 9.0 from it (likely: ±4° on the axis, ±0.3 on the end). 05 08_1b traced (the opening's circle fitted, 9.81 px/mm, the sheet's N up by holes 51 and 26): its sides 115.5° and 111.8° over their length, the end's arc round a centre at 119.7°, r 7.74 (its far point r 9.06), the sides' outer halves alone 118.4-120.5° (the drawn mouth widens toward the opening); "50°" is from the pocket's dashed S edge to the tongue's N side, which the pocket's edge meets. Wempe 3955's top face (12_Wempe_3955_obere_Platte_oben.jpg, 11.43 px/mm on 8 holes): axis about 110°, 1.1 S of the centre, 2.85 wide, its end r 9.2 (8.6 read again by eye). Until 18 October 2026 its root 50° round from the pocket's line and its end 8.6 toward 112.5°: the axis toward 93°, 2.4 S of the centre, its end r 8.54 (estimated). The pocket's S edge, the tangent to the opening, runs from 51 to the tongue's N side; the crescent it would leave beyond the tongue's S side isn't dashed on the sheet and isn't cut.
  • The intermediate bridge's outline, 45.5 × 10.0 with round ends, 3.8 thick, its screws' sinks Ø 4.7 × 2.0, its pins' holes 2.5 deep: scaled on 05 08 04, the sinks' Ø on Wempe 3955's dial side (3_Wempe_3955_uZ.jpg, 8.6 px/mm on holes 3, 12, 40, 30, 18 and 26: 4.7 ±0.1, likely; Ø 4.4 until 18 October 2026), their depth estimated. The same photograph measures the outline 45.7 ±0.3 × 9.75 ±0.15 with semicircular ends: the form confirmed, the sheet's sizes kept. Its section measured at 6.9 px/mm by its 3.8 (likely): the bush's hole Ø 4.3 with a 45° mouth from the plate side, Ø 5.7 at the face, to the bush 0.7 in; the seconds jewel's setting the bridge's full thickness less 0.2 at the plate side, Ø 5.2 outside and Ø 3.7 inside, the jewel 1.75 in, a straight recess Ø 3.7 under it to the dial face (until 15 October 2026 both sunk Ø 6.2 from the plate side, 0.7 and 1.7 deep, and the setting closed under the jewel to Ø 1.1).
  • The click: its base 2.85 under the hole (the Ø 5,7 circle tangent to it; the scan draws 2.79-2.81), the centre of its back's arc (fitted), its beak's front face, and its turn on the bridge (23.3° from N, ±3°, on Wempe 3955).
  • The upper cock's relief: the 45° step runs from r 12 on the full underside in to r 11.2 on the relieved one (26 October 2026: the "12 r"'s extension line comes down from the full underside's corner on both Schnitt A-B and U30, U30 at 49 px/mm puts that corner at r 11.98 and the step's top about 11.35, and the plan's hidden arcs read about 11.5 and 12.3; until then 12 to 12.8, which side inferred). Hole 20 (Ø 2.8 at 10, out to r 11.4) then cuts the step's top 0.2, as U30's section draws its edge at the step's corner. Its two M1.2 holes are tapped (26 October 2026) at 3.525 from the bore, the middle of 3,5 +0,05, into the Ø 6 boss under the relieved arm (the "6 Ø" is a boss, not a sink: the second drawing check's 8-M6).
  • The dust bush's flat faces the upper cock (inferred from the cock's place, above), confirmed on Wempe 3955 (2_Wempe_3955_A.jpg: its normal about 256° ±5 against the model's 255°: sure).
  • The pillar's body Ø 8.1 −0.2 (05 08 00-1, read "8,1∅" with the ∅ written high; 8.06 scaled at 5:1: likely; Readings corrected, 05 08_11).
  • Rough readings recorded, not adopted (the survey of 16 October 2026, PLAN-wempe-estimates.md's appendices; the user's decision: record them, keep the model's values): the pillar plate 3.40 rather than 3.35 (the seconds arbor's 24.1 −0.1 of revision 24.3.48 and its jewel's 1.7 leave its endshake −0.05 to +0.05 at 3.35, 0 to 0.10 at 3.40; against it the coloured section draws the pillar plate about 0.1 thinner than the top plate); the intermediate bridge's steady pins about 4.5 long (1.3 proud, about 3.2 in blind holes about 3.2 deep, 05 08 04's section) against 3.0 and 2.5; the box rim's lining lip (Lange 1947's wall top in two steps, outer 7.4-8, inner strip 4.4-4.6, drouot_33810142_lange1947_2; Wempe 4585's open lid shows a rebate; its height unreadable); the barrel lid's screws WN-21, heads about Ø 2.2 × 0.75 against 2.6 × 0.5 (hidden in the model); the box's floor about 14.5 ±2 (51 00_1) against 12.5; the glass about Ø 116 ±2 (Wempe 6161's rim, about 4.6 wide) against 112.
  • The cork ring's sizes, Ø 16.1 × 3.86 on the neck: scaled on the coloured section, a pictorial drawing (likely; the bush's own figures scale to 0.1 on it). Its bore drawn 0.01 clear of the neck. Its density in the weighing, 0.24 g/cm³, natural cork's.
  • The bowl's thread, rebate and notch (24 October 2026): none is on the casting's sheet. The thread's pitch 0.75 read on a 6MX's rim (±0.05), its size from the bezel's traced bore (±0.2), its form the ISO basic profile (60°, 0.406 deep) and the clearance 0.03 estimated, right-hand (estimated); the rim's height the flange's 4.7, the rebate 0.05 clear of the flange, the run-outs 0.5 (estimated); the notch 1.8 wide (the pin's Ø 1.6, 0.1 clear each side) down to 0.8 over the seat, cut through the rim as on Kirov-Poljot 24020 (its depth there not seen: estimated); the radial pin 1.78 out of the flange so its end clears the bezel's thread's crests (the coloured section reads 2.0 ±0.2). The casting placed by the reading that its top is the rim round the flange, which puts the bottom where 05 06_01 gives it (49 from the seat) once faced 0.7; the cork ring drawn 3.70, squeezed, against the section's 3.86. Since 8 November 2026 the 6MX's own thread is pitch 1.0, read on Kirov 247's rim (0.9-1.2, likely: the 6MX case review); the drawings' bowl and bezel keep 0.75, from Kirov-Poljot 24020's, which the later 6MXs may have.
  • The bowl's winding-hole seat (15 October 2026; the 6MX's only since 24 October 2026, the drawn bowl's bottom meeting the cork: above): with the movement in its bowl as drawn (the bore 53 deep, R-05 01 00-1; the movement ring on the flange) the cork's top is 3.84 over the bottom round the winding hole, so on the drawn bowl it would seal nothing. The casting's sheet draws no hole and nothing machined. Two 6MX bowls have a raised seat round the hole inside: moMvyx8eXfo 3:35 (1080p, the bowl empty in hand), a flat-topped ring (that it stands, sure) 72 px across against the hole's 38.5 in the same frame, so Ø 15.3 by the hole's 8.2 (likely; the cork's 16.1 nearly), its top an ellipse 60 by 72 px, so seen 34° from the axis, its near wall 8.5 ±1.5 px at the bottom's scale, about 4.5 px/mm (the rim's 5.26 less the 53 mm farther), so 3.4 ±0.6 high (rough); DIxob2Rl4dc 5:09 (480p), a pale ring there, so perhaps a pressed bush. The Wempe's bowl is given that seat, Ø 15.3 round the hole's 8.2, drawn 3.84 high to the cork's top (estimated: inside the video's height; the cork's purpose). No photograph or video of a German bowl's inside was found (UhrenJust's Part 6, WCguXfdZ38E, puts the movement in dial down).
  • Finishes: the plates, cocks and bridges gilt as Wempe 3955's (the sheets give the metal, not the finish).
  • From Wempe 3955's photographs, not on the sheets (16 October 2026, the audit of 3.00.00): the maker's engraving on the top plate's outer face, "Gerhard D. / WEMPE / Hamburg", centred at O 30.6 N 8.6, its lines 8.7, 9.85 and 7.4 long and 1.2, 1.7 and 1.2 high (2_Wempe_3955_A.jpg at 7.74 px/mm; WEMPE drawn filled where it is outlined); "RGG" in a triangle stamped on the balance arm, 2.2 across, 11 out (15_Wempe_3955_Logo_plus.jpg, oblique: estimated); hole 4's spherical oil sink Ø 4.3 on the top plate's outer face (its depth 0.95 estimated, two cones); the detent block's steady-pin holes drilled through the top plate and countersunk Ø 1.4 on its outer face (blind 1.4 from inside until then); the dial screws' M1 holes through the movement ring's flange (2.5 deep until then); the hand setting's holes 14-18 and its square window 6.5, centred 30.9 out on the line through 18 (struck through on 05 08 02-1, but Wempe 3955's plate is drilled for it, each hole where the table puts it).
  • The build checks' figures (Building it): the guard band, 180-270° over the wind (185° until 19 October 2026's physics pass) (wempe_escapement.js's, 14° under the reference's swing: no source's; 193° until 15 October 2026); the mainspring's strip ±0.005, the detent's and gold springs' ±0.005, the hairspring's strip ±0.003 (boxed, no tolerance written) and its cut length ±0.3, the holes' places ±13 µm (the tables give none), the banking as first set ±0.02 and the preload by eye 0.8 ±0.3 g (0.63 before the physics pass), and as the bench leaves them ±0.005 and ±0.05 g, the endshake set again within 0.01 of its drawn figure; the endshake stacks' undrawn terms (the top plate ±0.02, the seconds jewel's depth 1.75 ±0.02, the escape arbor's endstones 22.84 ±0.02 (22.82 until 25 October 2026), the fusee bush's 0.75 ±0.02, the barrel's bearings 3.29 ±0.02); the endshake band 0.01-0.08 (free in the cold with an oil film, to the Hamilton manual's 0.003 in); the lock's shallow end, 0.232, from the frame's 116 µm taken the same at every depth (0.154 from its 77 µm before the physics pass); the preload's aim a quarter into its window; the banking's aim the lock window's middle, set again until it reads inside; the maintaining spring's 5-minute floor; the timing weights' reach (their size estimated, drawn against their heads); a drawn tolerance taken as made anywhere within its limits.
  • The escapement of 1943 (the Variants' fourth version; The escapement of 1943 above): the banking screw's dome rising 0.4 on its 1.1 cylinder (no photograph shows the rise), its length 8.2 under the head (the 6MX's loose block's), so its threaded end 2.9 past the upright's far face (no German photograph shows that side), its underside where the pipe rests (as the 6MX's). Its block's stud (29 October 2026): its height, 2.5 proud, and its M1.7 3.0 into the block the 6MX's (the plans show only its slotted face; DEC 422's S view puts its top about level with the pillar screws': possible); the near steady pin 1.2 into the block, the 6MX's; what clamps the block, no photograph.
  • Wempe 3955's set-up ratchet recess: 0.7 deep, its wall at 45° (only its two edges, Ø 8.7 and 7.3, are measured).
  • The 6MX's domed screw heads: 2.4 overall with a 0.8 dome (±0.3), read on oblique views against the model's head of the same width; their thread and shank the drawings'.
  • The 6MX's cork ring 2.4 thick (its top 1.6 under a 3.86 one's; ±0.3) and its waisted dust bush's flare taken as a cone; the 6MX's detent block stud screw 2.5 (±0.7) proud; its barrel arbor's square end ±0.4.
  • Wempe 3955's dust bush waist Ø 9.0 (±0.4) and shoulder height (±0.3), its collar 1.85 (±0.2); its fusee square's tip ±0.5.
  • The hammered finish's tile: its statistics measured (scale, spread, orientation), its pattern synthesised.
  • The hairspring's collet bored on its sheet's 1:70 to the 1946 staff's seat, Ø 1.647 at the arm's face to 1.604 (25 October 2026 the size, straight Ø 1.648; tapered 27 October 2026): its sheet, 05 03 02-2 (24.2.43), gives 1.6 H7 on 1:70, tapering to 1.55, the earlier staff's; on the 1946 staff's seat (05 04 02: 1.58 to 1.65 s6) such a bore would seat 3.15 higher, where 05 03_5's 13,4 and 0,75 put the collet on the arm. The collet is bored to the seat, as the impulse roller's 1.627-1.65 bore is, and sits 0.05 over the arm; a maker opens the collet to the staff (no effect on the rate).
  • The top plate's bushes for the intermediate and seconds arbors (WN-300 at 38 and 35, no WN sheet): bored 0.845 and 0.83, each 0.02 over its pivot, for 05 05_1's "Lagerstellen: Enger Laufsitz" (the centre arbor's WN-301 1.21 on its 1.19); 0.86 both until 25 October 2026 (the per-part audit). The change pinion's root Ø 5.4 (Wempe 3955, ±0.1) leaves 0.05 under the hour wheel's tips, and the up-and-down pinion's 0.035 under its wheel's (the sheets' own figures): depths a maker must watch; no check measures the motion work's depthing (gearing.js takes the going train).
  • The mechanisms in turn (25 October 2026, wempe_wear.js, wempe_overwind.js, wempe_motionwork.js, wempe_estimates.js): the wear coefficients (Rabinowicz's classes, the escapement's as dry non-metal on metal), the hardnesses' soft ends (333 gold hard rolled 150 HV), each contact's area, slide and the tolerance taken for it (the barrel drum's line 0.10, the chain's 1 % of its pitch, the fusee wheel's teeth 20 µm, the escape tooth's tip land 0.02 and its wedge 0.5, the balance pivot's flat end), the gold spring's loads (about 10 and 2 mN); a hand's torque carried on past the data to the Ø 34.5 paddle; the key's coil wire (a design value: Ø 0.68 or more, spring wire EN 10270-1 DH), its drag back 5 mN·m; the studs' shake 0.02; the cannon pinion's bench weight 30 mm along the hand; a hairspring strip drawn 6 % long; the TF range 80-200 s.
From the parts audit (3 November 2026): 3
  • The staff's lengths (25.6 overall, the pivots 6.9 and 6.1) are taken from 05 03_3.2. That sheet is the post-war Glashütte production's (12.3.65, no Wempe title block, tempered 210° against 05 04_4's 240°), not a revision of this chronometer. No Wempe sheet gives them, since 05 04_4's top is cut off.
  • The banking screw's head is Ø 3.15 (Lange 5177's, the smallest of the photographs' 3.15-3.6), to clear the detent with its hole centred.
  • The strap screws' head, Ø 8 × 2.0, is DIN 97's proportion for 4 (the standard isn't in the set).
The train (P1b): 11
  • the intermediate and seconds pinions' cut leaves: tips Ø 4.72 and 3.17, roots Ø 2.87 and 1.70 (likely), by the escape pinion's sheet, the only one that gives a pinion's cut figures beside its turned head (05 04_8: head 3.13, tip 2.97, root 1.6 for z 10, m 0.263, so d + 1.89 m, d + 1.28 m, d − 3.93 m): the Ø 5.0 and 3.4 scaled on the intermediate's and seconds' sheets (4.9-5.2 and 3.36-3.57) are their turned heads (d + 1.89 m: 4.93 and 3.34). Until 18 October 2026 the tips the heads' 5.0 and 3.4 and the roots 2.75 and 1.61 (the centre pinion's 2.13 m dedendum). The meshes' efficiency tooth by tooth rose with them (gearing.js: the train 0.780 against 0.749's rule, escphys.js's ETA unchanged: PLAN-wempe-estimates.md, B.5);
  • the pivots' finished lengths (3, the kompl. sheets', taken for the last-revised pinions too) and the centre and intermediate pivots' tapers (both read 1:90, the figures re-inked or unreadable; the centre's 1:30 until 25 October 2026); which end of the centre arbor's 1:30 journal is the larger (the collar's, where both sheets put the 2.2 and as the shank's 1:80 runs; 05 05 03-2's outline 1 px wider there, under the 2.2 px the taper would give) and WN-302's bore following it (0.026 clear);
  • the seconds jewel's depth drawn 1.75 (05 08 04's 1.7, no tolerance), so the seconds arbor has 0.05 endshake; the centre arbor drawn 19.03 (19 +0.05) and WN-302 0.1 under the face (0.1 +0.02), so 0.07;
  • the riveting collar above the centre wheel's seat (05 05 03-2's 1.6 −0.1 the seat alone, drawn 1.55, and the collar on to the framed 19: closed 16 October 2026); 05 05_1's layout's lesser figures beside the detail sheets that supersede them: its 17 to the centre wheel's top face (05 05 03-2's 17,2 +0,05 puts it 17.26), the top bushes' oil cups drawn about 0.9 deep at the bore against the estimated 1.2, a small chamfer where each bush meets the plate's inner face (not built: the plates' holes are batch 8's and 9's), and its unattached "5,5 Ø" at the seconds' top bush (nothing of that size on Wempe 3955's plate) (the second drawing check's 6-M12, F6);
  • every bush's bore (the pivot and 0.01-0.025: "Lagerstellen: Enger Laufsitz"), oil sink (spherical, Ø 4.55 × 1.2 in the top plate running out to the bush's rim with no flat: Wempe 3955's top plate, 12_Wempe_3955_obere_Platte_oben.jpg, 4.46-4.69 at holes 1, 38, 35, sure that no flat shows, likely on the size, the depth estimated; Ø 4.0 with a 0.3 flat until 18 October 2026; Ø 3.7 × 0.9 in the bridge; on WN-302 a spherical cup Ø 3.8 × 0.55, its form drawn on 05 08 02 and its size scaled in a schematic 2:1 section) and length beyond the plates' faces: the WN standard parts have no sheets in the set;
  • the seconds jewel Ø 3.6 × 1.1 (scaled on the coloured section, a pictorial section) with an olive hole Ø 0.78, its oil sink on the dial side Ø 1.7 × 0.6 (the section's, ±0.1), and its setting's form (a tube the bridge's thickness, burnished over the stone, open under it in a straight recess Ø 3.7 to the dial face: 05 08 04's section, likely; a floor under it with the post's hole Ø 1.1, the coloured section's, until 15 October 2026);
  • the studs' diameter, Ø 1.5 (05 02_1 at 5:1 scales both 1.52, the pipes' bores 1.50-1.56 on 05 02 02 and 05 02 04; Ø 1.3 until 15 October 2026, scaled on 05 08 02 at 2:1), their domed ends (05 02_1: rising 0.3) and the second stud bush at hole 19 (drawn dashed in 05 08 02's Schnitt A-B, taken as another 05 08 02-2);
  • the wheels' riveting: each seat its sheet's length, standing past the hub as turned, before riveting (the rivet's spread into the hole isn't drawn; the centre arbor's collar with its groove); the seats and holes line to line at their nominal sizes (the n6/H7 presses' interference isn't drawn);
  • the cannon tube's punched dents (05 05 03's "Körner" over the tube's groove): three, 120° apart, each 20° round and 0.4 high, pressed into the shank's groove line to line (their number, size and place estimated: the sheet draws the mark, not the dents; the tube held by nothing until 28 October 2026, 6-M6);
  • the centre arbor's shank groove under the punch (05 05 03 of 1943, scaled: 0.24 deep, 0.63 high) and the crown's sphere (r 3.1, from the arc's 0.35 rise over the 2.86 the side view shows);
  • finishes: the wheels gilt and the arbors bright steel, as Wempe 3955's (8_…Räderwerk_plus.jpg).
The fusee and barrel (P1c): 12
  • the mainspring's length, 1,240 mm (not drawn anywhere): fitted to 05 06_20's measured force through the drawn fusee (Physics: energy), good to about ±50 mm (GUB 12255's barrel opened and let down, UG44DSWMv3o 0:26 and 5:56 at 19.7-20.8 px/mm on the 38.5 bore, holds about 13 coils on the wall at 0.40 and 3 free turns, about 1,530 (1,360-1,700, rough), a post-war movement whose spring may be a replacement; solved at 1,530 (set 0.6 and 0.7, 16 October 2026) it fits 05 06_20's curve three times worse, rms 3.3-3.9 % against 1.07 % at 1,240 mm and the same set, and doesn't raise the level, 0.753-0.802 against 0.802: kept); 1,386 until 15 October 2026, half the room between the core and the wall, the usual rule; its height drawn 14.8, not 15: the 1948 drum is 14.95 inside under its lid (0207's 15.8 ±0.1 space is the 1941 barrel's); its shape (two packs and a turn between);
  • the band's width, 0.78 (a post-war GUB's band, about 0.2 mm² across, gives about 0.8-0.9 at the sheet's thickness: rough, agreeing; The fusee: the band); its hooks' outlines, approximated on their figures; the chain's hooks' arcs between their figures (traced, the crown r 1.55, the barb's front), their eyes r 1 as the plates' lobes; the hooks 0.80 thick (05 07 04's side views: the chain's full width), their eyes thinned to 0.26 between the end links' outer plates out to r 1.1 about the rivet (where the step falls isn't drawn); the plates' sides between the lobes (no figure: traced on 05 07 04's plan views, the waist 1.41, 1.38-1.44 over 13, the sides arcs r 0.93 tangent to the lobes); the rivets' ends flush with the outer plates and polished bright (the edge view draws nothing proud of the framed 0.82; Wempe 3955's chain shows them lighter than the plates), their holes Ø 0.56 round the drawn 0.55;
  • the stack's 0.05 gaps between parts that rest on each other; the fusee's face at 13.6 of 05 06 04's 13.7 −0.1 (its old "13,6 +0,1" and the new band's common end, the coloured section's ratchet to 6.1 off the plate: with the 0.05 gaps kept, 13.65 would need 0.025); the Vorreiber's face a re-turned to 2.76 (its flange 0.86; its Ø 9.9 relief 0.425 deep under the re-turned face, the sheet's 0,4 +0,05 from face a as re-turned: cut after the re-turn. Until 28 October 2026 this line said the re-turn left it 0.06-0.31 deep, as made at 3.1 or 2.9, which the solid never was: the second drawing check's 5-M6), which the stack under that face leaves; the fusee body's large face re-turned 0.2, set by 05 06 04's "0,5 +0,05" to the click wheel (0.15-0.2 with the body's 12.55 and recess 2.7: the 0.5 taken), so the body 12.35 against the sheet's bracketed (12,3) and the flange's ribs 1.8 over both against its bracketed (1,6), which no single re-turn meets with the 0,5 (a reference carried over from before 1948); the face falling back 5° outside 05 06 04's Ø 33 +0.1 (its angle drawn at the flange's outer corner, one leg on the 33's extension line), its purpose inferred: the face bears on the maintaining ratchet's plate only inside Ø 33, the flange's rim, the chain's pin and the first rib clear of it and of the maintaining detent's tip, which stands 0.15 past the ratchet's face just outside the flange; the maintaining detent at 6.2, the least of its 6.3 ±0.1 (05 05 05 of 1941, not revised with the fusee in 1948; the coloured section draws it at 6.0), so its tip covers 0.85 of the ratchet's teeth (0.75 at 6.3);
  • the fusee wheel's tube Ø 6.5 g6 (05 06 00-3 writes "6,6 +0,1" over a struck figure and draws the tube 6.31-6.44; kept 6.5 g6 by 05 06_1's 6,5 H7/g6 under the ratchet's 6,5 H7, 05 06 02-1 as drawn 29.11.40 and unchanged by its revision of 11.3.48: a 6.6 tube can't enter that bore) and bore 4.55 (its "4 +0,1" can't pass the arbor's seat; 05 06_1's 4,45 H7/g6 is a straight seat's fit, and on 05 06 04-1's 1:70 seat, 4.43-4.48 under the wheel, a 4,45 H7 bore would wedge: the 4.55 an estimate, 0.04-0.06 clear on the radius, 4-5 times H7/g6's; the second drawing check's 5-M10 and 5-M15); the fusee body's length as turned 12.55 (12.6 −0.1); the 1,2 taken as every turn's pitch (it is drawn once, at the first land) and the first turn's rise out of the flange's slot (a cubic, level at the hook); the slot's deepening under the chain's pin read as a slitting cutter's (its depth scaled on section A-B), the pin 0.17 rad behind the groove's start (where along the flange the groove starts isn't drawn);
  • the chain's pin's angle along the flange (0.17 rad behind the groove's start; its radius, 1 under the ribs' tops, is the sheet's) and the hooks' throats round it;
  • the stop work (05 06 03, no sheet): the bar's slide at full wind, 3.36 (the chain's, wempe_stopwork.py) against C-D's figured 3,2 (the lower end's way between its two drawn positions; the positions themselves scale 3.54-3.66, so the drawing isn't to scale there and the figure rules): 0.16 over (the second drawing check, 3-M15, likely). Not corrected, because C-D's figures can't all hold on the model's fusee: the 22,5 bar with the "2" (its upper far corner r 12.5 running) puts the lower point at r 12.36 running, where the "1,76" puts it at 12.26; at full wind the chain on the top turn (05 06 04-2's groove, figured) holds the lower end at r 9.32, the "1,2"'s 9.3, so the slide is 3.36 with the drawn bar and 3.26 with the lower point at the 1,76 (the upper corner then 0.1 past the "2"); 3.2 itself would want the chain's contact on the top turn about 0.06 further out than the figured groove puts it, or the upper corner 0.16 past the "2". Which of the four figures gives way isn't settled by any sheet; the stop's function doesn't depend on it (full wind is where the stop pin meets the bar, which the chain's push sets the same either way); the cap, 3.6, is the drawn positions scaled. Also the bar's thinner upper part, 1.05 from u 0.8 (its thickness from the coloured section's 0.70 of the middle (9.5 px of 13.5), its start so its full thickness stays inside the small end's recess across its 3.6, since the slot is 1,1 deep from the recess out on that side; u 3.4 until 31 October 2026); the spring's height, 1.0, a design value (no source draws its section: the coloured section leaves the recess between the bar's end and its wall uncoloured, where it colours steel blue; the 5:1 fusee sections 05 06_1, 03-1 and 03-2 draw the earlier stop, a post at the small end; no 05 06 03 sheet is in the set; no photograph or video found opens a fusee: the GUB restoration, ZEc91vwppVk and WCguXfdZ38E, keeps its cover plate on). Anything from 0.9 (its pin's Ø 0.5 hole through it, 0.2 of strip either side) to 1.5 (under the recess's 1.6, 0.1 off its floor) works: its stress is the same at any height, 750 MPa wound against hardened spring steel's 1,200-1,800, and its push, linear in the height, is 0.054-0.12 N half way over that range, 37 times or more the 0.5 g bar's sliding on the cover plate (wempe_strength.js); the bench's stop work test (wempe_bench.js) has a maker check the bar springs back and the stop holds full wind. Its thickness, 0.10, worked from its bending (the stop work), and its shape between the two drawn; its pin's height, mid-spring; the bar 0.02 off the chain where it presses it; the stop pin's fit and its outer end flush with the plate (its length is scaled off the coloured section, 1.85 ±0.1: the stop work; anything from 1.24, to bear on the bar's whole thin end, to about 6, short of the chain on the fusee's fourth turn, would work: a strong hand's 1.06 N·m on the key, Swain 1970 as overwind.js takes it, puts 76 N on it 0.7 out, 134 MPa);
  • the maintaining detent's arbor (05 05 05-2, no sheet: Ø 2.25, 18.95 between shoulders, pivots Ø 0.88 × 3.1; its shoulders' 0.25 × 45° chamfers and its pivots' domed ends, rising 0.35 at the lower shoulder's end and 0.25 at the other, scaled on 05 05_11, likely; its seat tapered 1:70 as the detent's hole (05 05_12's "Kegel 1:70") over the detent and the 1.2 it was knocked along, the body straight either side at the taper's ends, 2.233 and 2.279: 28 October 2026); its wire spring (Feder 05 05 06-1, no sheet: its hole is the top plate's 5, "Drahtfeder", M1.7) and its stud, measured on the north-side photographs (4_Wempe_3955_N, 4_Kirow_247_N, 4_Lange_5177_N, scaled by the hole table's spacing of holes 3, 4, 5 and 49) and from above (2_*_A), checked on the coloured section (9.37 px/mm by the plates' 19): the stud a plain slotted cheese head Ø 3.0 × 1.9 (2.6-3.5 and 1.8-2.0 on the three, the section's 3.4 × 1.87: likely ±0.3, ±0.15), its slot 0.45 × 0.7 (likely) turned 10° as Wempe 3955's (it varies), steel (Wempe 3955's is brass coloured), its shank 0.4 out under the plate with its end rounded (the section: likely), the wire in a hole 1.5 deep in that end (estimated); the wire steel Ø 0.25 (0.24-0.26: likely ±0.05), to 1.8 over the pillar plate (the section; the photographs 1.2-3.3: likely ±0.4), drawn upright across the detent (the section; the photographs lean it 1.8° and 2.9° toward the arbor and 4.9° away, a set given by hand) and bent out by the detent's back, which stands 0.32 over its axis at hole 5: 0.31-0.51 mm at the arm's upper corner as the detent rides the teeth, a cantilever 12.3 long, so with steel's 206 GPa it presses the detent with 0.020-0.033 N (its root stressed 160-270 MPa);
  • the maintaining spring's pins drawn Ø 1.2 in their 1.2 holes (020605: 1.25) (its swell round its pin traced on 05 06 01-1 since 15 October 2026: 3.4 at 288° and 4 at its end were estimated); its outer edge's run into the eye, an arc tangent to the ring 60° from the eye (the sheet's line comes in over about that, within its own ±0.08: the angle estimated; 27 October 2026); the click springs' holes on the strip's middle (05 06 02-2's 0,8 −0,05 from its edge at its lower limit);
  • the clicks' screws (WN-21: put in from under the ratchet, the shoulder Ø 1.19; its M1.7 an ISO thread at 6g in the plate's tapped hole, 27 October 2026), the click wheel's screws (WN-19, heads turned to the holes' 1:8; their M2 an ISO thread in the body's tapped holes, and the cover plate's WN-20's M1.7 in its, both tapped since 28 October 2026, the second drawing check's 5-M11 and S-6: plain until then), the cover plate's screw (WN-20, countersunk, 3.5 into the body's 6); the Vorreiber's pin (05 06 00-5) Ø 1.16 × 8.3 in the hub's and the arbor's holes reamed together, Ø 1.15 (the sheets' 0.8, 1.1 and 1.2 reconciled: the arbor, above), its taper not drawn; the chain's pin (05 06 00-6) pressed, Ø 0.567 in the ribs' 0.555 (its own figure isn't in the set; 0.0025 clear until 28 October 2026); the tapped holes drawn flat at the drill's depth (the M2s at 6.25, the middle of 6 +0.5), and the M2s' angle on the body read off C-D (likely ±2°); the click springs' rivets' heads (05 06 02-4, "Niet": domed, 0.2 high as 05 06 02 draws them, Ø 1.0 estimated) and their formers' tangencies 15° apart (02-2, likely ±2°) and the tip's place where the 27.0 ends, 113.6° from the hole against the figured 110° (likely) and the 1.7 sink and 2.2 opening of the clicks' holes on the faces first-angle projection puts them on;
  • the hook block and the Steg (built as their sheets draw them on 18 October 2026: the barrel, above): the rivet's place on the block, 1.4 from its far end (0207's plan, scaled: likely ±0.5); the strip's end on the block 0.3 short of its far end, its last 5.7 lying on it (the hook's form and the strip's hole aren't drawn); the Steg's ends flush with the bottom's and the lid's outer faces (0207's section: likely); the 1941 lid's and bottom's slots for it (2.025 wide, open to the discs' rims to r 18.35 as 05 07 01-2 of 7.4.41 draws the lid's 9 o'clock notch, since 28 October 2026: closed 2.025 × 0.55 before, estimated); the Steg's 0.5 in the 1948 drum's 0.8 slot (05 07 01-1's revision b widened the slot, "war 0,5": the coloured section draws the Steg 0.45-0.55 thick, so the slot's width doesn't make the Steg 0.8: the second drawing check, 4-M12, closed); and the drawn spring's bends under the Steg and the block's ends (over 0.25-0.4 rad, as the solve's states step: tools/wempe_mainspring.py). The block's length is 0207's dimensioned 6 (the later sheets scale 5.1-7.3). The chain's slot in the drum (cut at the fusee's assembly, 0207: 2.5 × 0.85 through the wall) cut where the laid chain's hook puts its barb, and the band's where its hook does, one drum for each (since 25 October 2026). The 9 o'clock slot of 05 07 01-1 (revision a of 11.3.48: 2 +0.05 wide, "0,95 +0,15", "18,3 +0,02" struck; revision b of 20.4.48: "0,8 +0,05", was 0.5 +0.03) is the Steg's, through the drum's bottom at r 18.3-19.1 (sure: the section draws it through the bottom just inside the bore, the plan its outline from the bore's 0.15 in to 18.3); from 17 to 18 October 2026 it was cut through the wall from the rim and read, with the lid's 9 o'clock notch, as one slot for prising the lid out. The lid's 3 o'clock notch's use isn't drawn (prising the lid out: likely; since the drum's insides were turned to 0207, 25 October 2026, it lies at the chain's slot's angle, 1.8 from it along the axis);
  • the set-up ratchet's teeth face the way the frame's click takes them (README, Unresolved);
  • finishes: the fusee, its wheel, the maintaining ratchet, barrel and set-up ratchet gilt, as Wempe 3955's (8_…Räderwerk_plus.jpg, 7_…W.jpg); the clicks, detent and maintaining spring blued (their sheets' 295° and 310°); the click wheel and dust plate plain brass; the click springs bronze.
The escapement (P1d): 15
  • The locking stone's place, 7.14 from the wheel at 37.0° from the line of centres (0.235 deep): scaled on 05 04_1 against the tips' Ø 14.00 (likely, ±0.04 and ±0.3°; 7.2 and 36.5° until 19 October 2026).
  • The detent block's turn about hole 13, fitted so the stone lands there: the stone then sits 18.74 along from the block's screw row, 0.06 short of 05 04 01-4's 18.8 to its cross screw (18.73 with the blade centred, until 27 October 2026).
  • Which face of the block lies on the plate. The front view of 05 04 01-4 is taken as the face on the plate, seen from above as the frame's plans are: drawn the other way round, the base's end that the detent's foot is clamped to would face away from the detent. Its side view is read in first angle, as the sheet reads throughout (DIN's; 15 October 2026, sure): the upright at the free face, 3.5-6.5 under the plate, the banking screw 5.0 under it and the foot 1.3-3.3, as 05 04_1's elevation prints them (the upright 6.5 to its end, 2.8 high, the screw's axis 1.4 from its end) and Wempe 9166 shows them (since 27 October 2026: the escapement of 1943's block; until then read in third angle, the upright flush with the plate and the foot 3.2-5.2, to meet the misread rollers). The block's steady pins' holes in the top plate (not in its table: drilled at assembly) and the stud's screw's tapped hole in the cock (not on 05 08 03-1) are cut in frame.js.
  • The detent's heights: its foot 1.3-3.2 under the top plate (3.2-5.2 until 27 October 2026), the blade flush with the foot's lower edge (1436's right view), the pipe 1.9 long (1436's) from the blade's top edge, 2.35-4.25 under the plate (as first drawn 2.0 long, 2.30-4.30: 05 04_1's elevation dashes it from the blade's top line, 2.28, to 4.25, beside the arm set back 1.08; Wempe 3955's still 1.95-3.65, its arm flush with the base's edge as photographed, where the drawn pipe would run 0.16 into it: estimated), its side on the banking screw's head, and the stone standing out of it to 0.06 over the teeth (1.9-3.8 and, as first drawn, in 1943 and the 6MX, 1.95-3.65 until 28 October 2026, from 0.1 under the teeth: drawing check 2's 2-M1; until 15 October 2026 from 3.3 under the plate, 2.0 long). 1436's right view draws the pipe's band from the blade's "unten" edge 1.9 toward "oben", 1.0 past the blade's flush edge: built so.
  • The gold spring's tip 1.54 from the staff, traced on 05 04_1, and its bending point 6.15 from the stone, 05 04_1's dashed mark (6.10-6.23 by the plan's scales: likely; 6.7 until 28 October 2026, on the "0,04" figure's glyphs beside it, drawing check 2's 2-M3; the owner's choice, O-6, with the detent's spring's mark, 12.52, for the flexure); the gold spring's head on 05 04_1 drawn 8.40-10.77 from the stone, its screw at 9.57, against the model's 8.18-10.58 and 9.38, 0.19 nearer the stone (05 04_1 is the 1940 layout; 05 04 01-2 draws the part, not its place: within the layout's precision, recorded: 2-M10); its bow round the pipe traced on Wempe 3955's photograph (likely). The horn bent toward the block to carry the gold spring (its curve, 1436's "Anschrägen", traced on Wempe 3955's photograph since 18 October 2026: likely, ±0.1; 1436 draws the bend later, its length not to scale there); its length 4.4 is 1436's figure with Wempe 3955's photograph (likely; 3.5 until 15 October 2026).
  • The discharge stone's reach, 1.578 from the staff (1.563 until it was recomputed, 1.5754 with the blade built, until the gold spring's tip's slide about the offset flexure was taken, 27 October 2026), set so the detent lifts the stone 0.305, 0.07 past the lock's 0.235 (the sheet scales 1.7-1.75; 1.547 lifting 0.245 until the lock was re-laid, 19 October 2026), its end bevelled 0.3 so only its leading corner works the gold spring, 0.4 across; the discharge roller's outline (scaled: round at both ends, as built since P1d; 05 04_1's dashes at the stone's end can't be read square or round, and a square end where they end would put its corners at the gold spring's tip) and its material (brass, from Wempe 3955's: likely).
  • Which side of the impulse stone the hollow lies (behind its face, where the tooth comes: inferred from the motion; 05 04_1's plan draws it so, about 3.7 long and 2.0 on the tooth's side: likely). A GUB's roller lying loose (Ldzmn4kuKMA 3:53.5, the video review of 27 October 2026, References/wempe/VIDEOS.md) has the hollow's larger part in front of the stone's face, as here: verified, sure (its hollow 2.5 wide against 02 04 02's 3.5, its second hole and its long stone are its own).
  • The escape wheel's crossings' turn against the teeth: one under the locked tooth as 05 04_1 draws it (05 04_23 draws its crossing 17.9° behind a tooth's tip; a crossing's place against the teeth is set in the making, not by the form); the meaning of 05 04_1's 86°, 22°10', 6°, 5°, 20° in the wheel is not read (05 04_23's "6,0606∅" is a dashed circle about the centre through the crossing, purpose not written). The openings' arcs' centres 9.43 out (traced, not figured), the floor's step slanting 0.6° (traced), the recess Ø 11.62.
  • The escape wheel's collet's diameters, Ø 2.8 and 1.7, its step 1.6 over the wheel and its chamfers 0.15 (scaled on 05 04_7.2: likely). A GUB's collet side-on (Ldzmn4kuKMA 1:20) reads about Ø 2.4 ±0.15 with no step to 1.7 showing, scaled on its pinion's head and the wheel's tips (its wheel end hidden by the crossings): that clock's, not adopted.
  • The balance: the masses' height and place read on the sheet and photographs (likely: How the balance and escapement were measured), their chamfer 0.26 × 45° (the plan's ring; the side view's about 0.4); their screws (WN-12: M0.8, the drawn shank's Ø 0.84, its head Ø 1.6 sunk flush in a counterbore Ø 1.62 0.55 deep, from Wempe 3955; threaded through the outer jaw's tapped hole to the rim, as the 6MX's book's Fig. 5.62 cuts a mass; the hole's mouth on the slot's round wall spot-faced up to 0.02) and the tapped hole in each top for WN-12a (M1.2 2.0 deep, its top flat: 05 03_2's plan and the coloured section; left empty, the owner's choice; WN-12a was a nut sunk in the inner face until 18 October 2026), the timing weights' size in Wempe 3955 and the 6MX (Ø 3.8: the photographs ±0.4) and their cuts' depths and width, the timing screws' heads (Ø 1.65 × 1.3, scaled on a poor copy), the punch marks' size and height on the rim's face (the photographs'). The hub's screws WN-10 go in from the arm's top (the coloured section), M1, into the flange's tapped holes (05 04 02's "M1"); the arm's sinks for their heads are read from the coloured section's conical heads, flush: 05 03 01-1's "1,8 Ø H7" (its leader at the side hole) and its "0,4" (an arrow 4.4 from the axis, on no hole) don't attach to a countersink, and its plan rings the left hole only; built a 90° countersink to Ø 1.8, 0.4 deep, in both (possible, recorded: the second drawing check's 1-M10). The staff's seats at their fits' middles (s6, r6) where pressed, the impulse roller's 1:70 bore on its straight s6 seat drawn with its interference (0.009-0.021 on the radius) and the discharge roller's estimated bore at H7's middle (0.005), the hub and the collet line to line on the staff's 1:70; the rollers' and collet's places along the staff as before.
  • The hairspring's outer curve the same as the inner, 05 03_7's table, pinned in the stud's hole 3.48 from the staff (3.55 until 28 October 2026, when the stud's screw went to U30's 4,2 across and the stud was built off-centre as 05 03_9 draws it: the owner's choice, O-8; likely); the collet's pin (the same Spiralstift 05 03 02-4) half round and brass as the stud's, its flat 0.005 off the strip in the collet's Ø 0.8 (the strip's end straightened along the hole, as at the stud), 0.45 proud of each side (its form estimated from the stud's; a steel rod Ø 0.22 loose in the hole until 27 October 2026, found by seatcheck.py). The stud (05 03 02-3) has no slot: its plan's dashed lines, 1.87 apart, are the upright's own sides under the bar, and its "1,2 −0,1" runs from the bar's middle to its edge at the square end (read as a slot 1.2 wide until 27 October 2026, drawn 2.0 deep and open at the end). The spring's outer end goes into the sheet's Ø 0.7 hole, drilled along the bar when the spring is fitted, 0.36 from the upright's middle toward the axis (the end view, sure to ±0.05) and 3.15 under the bar (05 03_5, 05 03_1: likely; 05 03_9 gives it no place), 1.35 from the upright's end: 3.55 from the staff, where 05 03_7's curve and the collet's 3.49 pin it, so both curves alike (likely; 0.72 from the end, 3.46 from the staff, until 28 October 2026). Its pin beside it in the hole, outside: half round, the hole's round less the spring's side, its flat 0.005 off the strip, brass, 0.45 proud of each face and domed over its last 0.2 (Ldzmn4kuKMA 4:32-4:34, a GUB's old spring with its stud: a brass end about as wide as the hole, rounded; its length and dome estimated, the face seen obliquely; the video review, 27 October 2026, found no frame that reads its length: both ends proud and rounded, and the GUB's bar hole sunk to Ø 2.2 or more at its top with its outer corner chamfered, that clock's). The drawn curve bows 0.18 across the stud's 1.8, so the drawn spring's last stretch is straightened along the hole; the helix is opened to the hole's height (above); the physics keeps 05 03_7's curve. The drawn spring's straightening is estimated: the maker forms the end to the hole.
  • The staff's ends from the later staff 05 03_3.2 (05 04 02 draws only the blank, 14.03 to the lower centre): its upper end 13.0 above the hub's flange and its lower 12.59 below (its 25,6 −0,02 and 12,6 −0,02, each at its middle: 25.59; 13.03 and 25.62, over the overall's limit, until 28 October 2026, 1-M6; 12.83 below until 27 October 2026), each 0.02 off its endstone (the upper jewels following it down 0.03), so the lower endstone's face lies 1.19 inside the lug's face (the coloured section: 1.06 ±0.11); its pivots' cones hollow on R 6 (its detail's "6" read as a radius, likely). The escape arbor's ends likewise, 0.02 off.
  • Every jewel, setting and endstone (no sheet): the stones' sizes and form scaled on the coloured section (a pictorial section, to scale where it was checked: likely, ±0.1; above, How the balance and escapement were measured), their holes 0.02 over the pivots (built so since 25 October 2026: until then the balance's 0.04 and the escape wheel's 0.03 over them), the oil gap between hole jewel and endstone 0.03 (at the escape wheel 0.07 above and 0.04 below: its endstones 0.03 off the arbor's ends for the 1958 book's 0.05-0.07 of endshake, the hole jewels left where they were, 25 October 2026; 0.06 and 0.03, the endshake 0.04, until then), settings to their seats, each holding its stone by a lip (the section draws them burnished over it); the escape wheel's upper hole jewel as last revised thinned to 0.4 and its lower to Ø 1.6 × 0.45 to fit what the sheets leave; as first drawn the escape wheel's lower endstone Ø 1.6 in its plug, the form scaled; the 6MX's chaton jewels and its contact's are left plain discs with olive holes (no section or photograph shows their insides); stones Ø 2.0 (1.8 and 1.6 for the escape wheel's) × 0.35, plain discs, and the escape wheel's lower hole jewel 0.2 off its endstone (a floor of its bush between them as first drawn), until 18 October 2026; the escape wheel's lower jewel in a bush in hole 26; the escape wheel's endstones in round gilt settings as Wempe 3955 shows them (the upper's spigot Ø 4.5 through the lip, flush outside, two brass screws from outside with heads Ø 2.6 × 0.3 under the balance's rim; the lower's Ø 4.6 flush in a seat 0.25 deep in the dial face with a boss 0.36 proud round its stone, Ø 2.2, two brass screws' heads Ø 2.6 lapping it: 2_Wempe_3955_A.jpg and 3_Wempe_3955_uZ.jpg on millimetre paper; a stadium cap 9.4 × 3 × 0.6 over the lower and a closed setting under the lip above until 16 October 2026); the cock's endstone seen through its setting, colourless (Wempe 3955's and Kirov 247's: sure; a Wempe's described as diamond, Kirov 247's faceted, drawn domed), its screws brass; the cock's hole jewel under it a ruby as the others, showing pink through the endstone (the user's choice, 18 October 2026; 3955's photograph reads grey there, so it may be colourless: unsure); the other screws' heads Ø 2 × 0.6 and their WN numbers where the coloured section's labels couldn't be matched (the block's two, the escape settings', the detent foot's).
  • The discharge stone's end bevelled 0.3 back from its acting corner (the code's DS, makeEsc's BEV, drawn since 15 October 2026): 05 04_1 measured again scales 0.15-0.2 (±0.12, its dashes faint at 24.4 px/mm, the "1°" figures beside the corner); 0.15 would move the swing 0.003° and the escapement's own rate 0.01 s a day, and no contact (only the acting corner reaches the gold spring's tip), so 0.3 is kept until a better copy of the sheet settles it.
  • The page's escapement (The page's escapement, above): the balance's beat drawn from makeEsc's pushes with the air's and pivots' loss spread as a viscous one (estimated: its share through the beat; its total is TF's); each contact makeEsc solves exactly drawn 0.01 mm apart, as the layout kept them; the gold spring's bend drawn as a turn about its bending point where makeEsc moves its tip bodily (they agree at its square end, where the stone meets it).
The energy budget (phase 2b): 5
  • the mainspring's free curve: the Hamilton's measured spring (KLUwI2UUCMQ 32:18) scaled to the length, its turns ×0.75 (the set fitted with the length to 05 06_20's shape with the coils' friction counted, where the level also closes with the drawn strip: Physics: energy; ×0.6, the frictionless fit, until 26 October 2026); its eye the Hamilton's (2.2-9.4 mm along the strip from its inner end, 5.2 high; cut in the drawn strip since 17 October 2026, the hook's face 0.05 off its end toward the tip; the hook's ramp, 0.475, stands 0.1 into the next coil, which a real spring bends over it and the drawn spiral doesn't); its modulus 207 GPa (200-210); no film between packed coils;
  • the level: the solved strip at 0.37 gives 25 % more torque than 05 06_20 measures; the budget and the strengths take the solved shape at the measured level, the losses at their middle (×0.883);
  • the losses to the fusee arbor: the spring's own up to what mainsprings measure (ICMFF9, as the Hamilton's), the coils' mu 0.08-0.16, the pivots' and the chain's rivets' mu 0.10-0.15, the fusee arbor's pivots counted or not (05 06_20's rig isn't drawn);
  • the meshes' friction, mu 0.10-0.20 (steel leaves on brass teeth, run dry), and the pivots' shake for the holes' tolerance (0.02-0.05, the drawn bush bores; the WN sheets aren't in the set);
  • the strengths' ranges (strength.js's: brass, hardened steel, a mainspring's limit), Lewis's Y 0.2, each pivot carrying its arbor's whole load, the chain's plates 0.26 thick (drawn: three plates in its 0.82), its hooks 0.80 (the strip 0.9 finished to the chain's width), the band 0.78 wide, the stop pin's and stop bar's steel (no sheet).
The escapement in time (phase 2a): 9
  • the balance's free run-down TF 111 s, a running 6MX's measured on video at the model's own swing (111.1 s, 102-125, at 194.3° since 19 October 2026's physics pass; 108 s, 107.8 s, 96-121, at 199.4° before: References/wempe/VIDEOS.md, "Measured"; a 6MX's balance, not the Wempe's; taken since 15 October 2026, the user's choice), inside its parts' damping (81-208 s: the air's drag coefficients, the endstone's effective friction taken from the Hamilton's measured, the spring's own Q; their middle, 125 s, was taken until then); so the swing, 188.5-199.5° over the wind;
  • the train's efficiency 0.780, the geometric mean of 0.730-0.833 from its pivots at published friction coefficients and its meshes worked tooth by tooth (gearing.js --model wempe at mu 0.10-0.20; 0.749 from the meshes' rule of thumb until 19 October 2026);
  • (the discharge stone's place on the staff, DSB, was an estimate here from 16 October 2026, leaving the gold spring's tip 5° after rest with the release at +2.6°, "the scan can't give it", and before that leaving the tip at rest; since 11 November 2026 it is 05 04_1's own figure, the 5°: Readings corrected)
  • the detent's preload, 0.8 g at the stone (just over the margin's floor with the deeper lock, at the 0.80 g cap above which the stone, landing with brass's restitution, lands on the escaping tooth's back during the impulse (Closed, 25 October 2026; The physics pass; 0.63 g, the Hamilton's after Op. 78, until 19 October 2026: no Wempe source gives one; a banking sets the lock, the 1954 instruction's п. 11), and its restitution on the banking (0.3); the steel detent spring's modulus 205 GPa; the gold spring's, 98 GPa (96-101: Au/Ag/Cu 333-333, the alloy Poljot 20027's passport gives the 6MX's; 05 04 01-2 says only "Goldblech 333": likely; 90 until 19 October 2026); the brass tooth on the sapphire stone mu 0.15;
  • the hairspring's modulus 206 GPa (200-210) and its working length as read (Physics: hairspring and temperature): the period 0.468 s against 0.5;
  • the densities of the rim's nickel steel (8.1) and the gold spring (11.6, 333-333 as above; 11.0 until 19 October 2026);
  • in makeEsc, the discharge stone's bevelled end as a straight bevel from the acting corner to the back corner, 0.3 (scaled), and the gold spring's tip as its strip's square end;
  • the maintaining spring's working (wempe_sustaining.js): its end at the wheel's pin held by the recess's wall as clamped (the ring's 32.3 -0.1 in the recess's 32.3 +0.05: no figure says it bears), the modulus 207 GPa (200-210), the detent taking the ratchet half a tooth back on average; so its ends' 0.27-0.58 mm in running and its 11-14 minutes; the twist's swing, about 130°;
  • 05 06_20's force scaled off its curve (±0.7 g), by the chain's turn on the groove from the large end: the page's turns from full wind taken off the stop's turn, 8.638 (escphys.js MF; fusee.js MW, the page warning if they part), its run down 8.70 turns on, 0.06 before the curve's start, taking its first value.
The escapement in time (phase 2d): 5
  • the stone's friction on the tooth it has let go (μ 0.15, as on the lock: 0.10-0.20 move the landing −9.5° to −9.1° and the swing 214.5-213.5°), its landing on it plastic, and makeEsc's quasi-static ride (RW);
  • the detent's own damping (Q 100) and its restitution on its banking (0.3), the Hamilton's;
  • the twist's swing, 0.6 of the running swing (128°);
  • in the frame (tripping.js --model wempe): the banking a single contact, the pipe's side on the banking screw's head at the stone (05 04 01-4's cross screw and the 6MX's photographs: likely; its head not drawn), its stiffness tripping.js's; the contacts' penalty stiffnesses and restitutions, tripping.js's (the return contact's matters: Physics: in time); the gold spring's tip square and unturning in its contact; the gold spring's fixed part rigid with the square up to its bending point (05 04_1); the screw WN-14's mass weighed, at its centre; the stone's lower half's height (to 0.06 under the teeth);
  • the train beyond the escape and seconds arbors in makeEsc's JW left out (0.03 g·mm², −0.08° of swing).
The transmission (16 October 2026), tools/wempe_transmission.js: 8
  • each mesh's stiffness, the teeth as cantilevers (the brass wheel's tooth and the steel leaf in series, halved for the contact: 1,620 N/mm a millimetre of face; traindyn.js's), and its damping (3 % of critical, oil);
  • the chain's stiffness, 120-700 N/mm (its plates' and rivets' give over the free span and the turns on each drum the tension reaches, μ 0.15 on the groove), 290 taken;
  • the click's face on the click wheel's tooth 10 kN/mm, the detent's 5 kN/mm, the escape tooth on the locking stone 2 kN/mm, the stop bar on its pin 6.7 kN/mm (its sizes, wempe_strength.js's), the maintaining spring banked 40 times its running stiffness; the contacts' damping 5 %, the springs' 2 % (the let-go's figures over the ranges each is varied in);
  • the pawls' springs' set, 0.3 mm (the clicks') and 0.5 mm (the detent's), and the click wheel's backs a straight rise over a pitch;
  • the coils' stick: the friction taken up over 0.05° of the drum;
  • the chain's mass, 8.6 g (2.9 mm³ of steel a pitch), the up-and-down wheel and hand's inertia (560 g·mm²) and the key's (2,200), none weighed;
  • the hand: its grip 1 N·m a radian, winding at a turn a second in half-turn strokes, a strong hand 1.06 N·m (Swain 1970), easing back over 0.15 s; the key's coil spring's drag slipping, 5 mN·m;
  • the cannon pinion's fit, 16.9 mN·m (the middle of its window: no sheet gives it).
The hairspring and temperature (phase 2c): 5
  • the reading of 05 03_7's "13 Umgänge −3,5 mm" as the whole strip, and its sheet as this chronometer's (likely: Physics: hairspring and temperature); the strip's ends past the pins (1.35 and 0.9 mm);
  • the drawn spring's isochronism (WESC.SET.HS −0.543 s a day per 10° about 207°) from its curves as drawn, both 05 03_7's table; the strip's own mass and the curves' rise left out;
  • the rim's nickel steel as Guillaume's 42-44 % ("Sonderlegierung" isn't named further): 7-8.5 × 10⁻⁶ a °C, E 145-160 GPa; brass 18.5-20 × 10⁻⁶, E 95-110; the carbon steel spring's thermoelastic coefficient −240 to −270 × 10⁻⁶ and expansion 11-12.5 × 10⁻⁶ (published ranges, named in the tool);
  • the masses clamping the rim over their slot (Ø 6.3) so it doesn't curl under them; the arm rigid; the quarters' roots at the arm's edge;
  • the secondary error: the steel spring's share from Dent's error (2-3 s a day over 0-30 °C), and the integral balance's taken as Guillaume's balances' fifth of it at Neuchâtel (1902): the nickel steel's quadratic coefficient isn't to hand.
The dial, motion work and hands (P1e): 12
  • the hour wheel's collar, 3.585 over its wheel's lower face, fitted to end 0.3 under the dial with the wheel on the cannon pinion: its sheet (05 02 03, 1940) gives 5.2 −0.1, which with 05 05 03's tube (1943) would put the Ø 8.0 collar 1.4 into the dial's Ø 6 hole; its overall 8.2 kept;
  • the seconds hand's height, 0.2 over the 1.2 dial on the 1946 post's finished 7.2 (the coloured section, pictorial); the hour and minute hands are at 05 02_1's boxed 2,6 and 3,2 (above), the minute hand's on the tube's shoulder read as the 3,2 (likely), the hour hand's set by how far it is pressed;
  • the pinions' tips and roots (above), the change pinion's leaf thickness 0.55 on the pitch circle, the up-and-down pinion of steel (the coloured section colours it yellow; Wempe 3955's is steel);
  • the cannon tube's square's corners turned to Ø 3.45 (Wempe 3955's, 1_Wempe_3955_Z.jpg at 9.14 px/mm: sides 2.60-2.67, half-diagonals 1.72-1.77; 3_Wempe_3955_uZ.jpg the same: likely, ±0.1; Ø 3.2 until 18 October 2026), its crown so starting 0.6 under the top, and its diameter 3.6 (scaled 3.65); the change wheel's collar Ø 5.5 and the up-and-down pipe's Ø 3.4, 7.7 (its flange: Wempe 3955's, 3_Wempe_3955_uZ.jpg at 8.45 px/mm, 7.71 ±0.15, likely; 7.4 scaled until 18 October 2026) and 2.8, the hour wheel's collar Ø 8.0 (05 02 03 scales 7.95-8.2, Wempe 3955's 8.01, rms 0.03: likely; 7.9 until 18 October 2026) and bore 3.62 (scaled at 5:1); the up-and-down wheel's hub Ø 10.4 (the sheet's 10.5 −0.2; Wempe 3955's measures 10.77-10.92, recorded, not adopted: the drawn figure's tolerance excludes it, and Wempe 3955's version is claude/wempe-photo-inner's); the pipes' bores Ø 1.52, 0.01 clear of the studs (Ø 1.5, train.js; 05 02 02 and 05 02 04 scale them 1.50-1.56);
  • the change and hour wheels' riveting sinks (27 October 2026): 040204-1's "0,4" and "30°" at the hour wheel's hole, on its lower face (05 02 03's section draws the chamfer there, the pipe's foot flared into it), the 30° read to the axis (its leader on a chamfer about 0.2 across at 5:1: likely), Ø 6.46 at the face; the change wheel's, drawn the same on 05 02 02 (enlarged 6×, about 0.4-0.5 deep and 0.2 across) but not figured, made the hour wheel's (likely); the up-and-down wheel's the same since 28 October 2026 (05 02 04's "vernietet", its leader on the bore's lower corner, a small flare drawn there; not figured: estimated; until then nothing under it, the second drawing check's 7-M2). Their bores, and the change pinion's and up-and-down wheel's, are polygons on their pipes' lathe angles (48, 64), the pipes lying on their flats exactly (until then 24-gons the pipes stood 0.02-0.03 into)
  • the hands' outlines between the sheets' stations (traced), the up-and-down hand's length 10.2 (measured on Wempe 3955's dial: its sheet isn't drawn to scale there) and its pipe Ø 3.0 (scaled), read as a collar drawn up from the hand's 0.6 strip (its side view draws the boss a thin-walled tube under the flange, its walls strokes as the blade's: likely), its blade an even 0.3 under the flange's 0.5 (the side view's one stroke, 0.5 tapering to 0.3 until 28 October 2026: 7-M4), the hour hand's flange bevel (Ø 8.2 to the plan's inner circle 6.8), the minute hand's flange's round (r 0.4, scaled);
  • the seconds hand's collet (05 02 01-2, no sheet): Ø 1.5, 4.6 long, scaled on the coloured section, its end Ø 1.2 through the hand;
  • the dial's screws (WN-10 by the coloured section's label, likely: M1, heads Ø 1.7 × 0.8) and the ring's tapped holes for them; the dial's stepped pin (Ø 1.1 in the ring, 3.7 deep as the coloured section draws it, Ø 1.0 in the dial, whose hole scales 1.0; the step estimated); the hole at 185.7°; both small holes under the printed face, which no photographed dial shows pierced (Wempe 3955's 1_Wempe_3955_Z.jpg at 9.14 px/mm, where a Ø 1 would be 9 px and the screws' sinks show; Lange 5177's; Kirov No. 247's: sure; through the printing until 18 October 2026);
  • the dial's printing: traced on one photograph (Wempe 3955's, 9.2 px/mm), so its edges are good to about 0.03 mm and its line widths include the photograph's blur (sigma 0.046 mm) where under it; the 20's 0 and the 8's top, under hands there (above);
  • the Lange dial's printing (Lange 5177's, 8.9 px/mm, blur sigma 0.024): the same, and where its rust lies on the printing (wear, left out by its colour: brown where the ink is grey) the glyphs' edges taken a quarter of the way from the ink to the paper (0.016 mm in from half way, under this blur), so a few edges stay rough by 0.05-0.1 mm (the S of SÖHNE, the R and N of DRESDEN, the seconds 20, 40 and 50); estimated outright: the 16's 1, lost under rust, which is the seconds 10's 1 scaled to the up-and-down figures' height (0.695) and set by the 6 as the 10's 1 stands by its 0; the serial's last 7 (its stem broken by rust), the first 7 fitted on what shows; the 8's top under the minute hand, an ellipse through its upper loop (as Wempe's); its divisions under the hands or moved by rust (the seconds' 6) at the nominal angle, and the outer corners of the up-and-down band's ends sharp (not measured); the ground white, not the photograph's aged cream (above);
  • the glass's underside 17.0 over the plate's dial face (05 02_1's pencil, unsure), until the case was built (P1f: 16.7, above);
  • a fill light from below in the page (app.js), so the dial, which faces down in the page's views, reads white.
The case and box (P1f): 12
  • the bowl's machining: the recess Ø 68.2, the winding hole Ø 8.2, the screw's M3; its inner corner r 7 (scaled); its material: brass as the sheet, though the sheet's "Gewicht 350 gr." is a quarter of the 1.34 kg this bowl weighs in brass (158.4 cm³ with its brackets, weighed on the solid, phase 2d: 350 g is 2.2 g/cm³, a filled moulding's or, machined, aluminium's; a bowl that light wouldn't hang on the drawn pivots: Physics: case, gimbals and the trial);
  • the movement ring's seat 5.3 over the pivots (the bowl's top face until 24 October 2026; 05 01 against its 11,5 reads 4.6 and 4.75, ±0.5, for the 1940 bowl; 5.3 so the cast bowl's foot clears the ring as it turns to be wound; the ring's top level with it until 21 October 2026, the pivots then 11.5 under it);
  • every line of the bezel's section traced on 05 01 01 (no figures, ±0.2); its thread (05 01 01 marks it; its size and pitch: The bowl's thread, rebate and notch, above; until 24 October 2026 its lower bore was opened 0.35 for the ring's pin and plain); the glass's Ø 112 × 2.3 (scaled on the 1941 bezel) and its seat flush with the top; the notch for the ring's pin; the radial pin's 2.0 out and 9.4 long, its hole not cut;
  • the gimbal ring's material (brass; Wempe 3955's is Bakelite; its sizes measured on 05 01 since 25 October 2026, How the case and box were measured); its bracket's length, 57, and its screws, 35.5 apart (Wempe 3955's ring, 16_Wempe_3955_Gehäusetopf_Kardan.jpg, 35.5 ±0.8: likely; 37.2 until 18 October 2026); the ring bracket's screws round holes where 05 01 says they slide in slots ("Kloben verstellen"), M2.5; the pivots' shoulders 0.1 off what they bear against (endshake), the tapped holes' fits (the ring bracket's screws tapped through the ring, as Schnitt A-B draws them, since 27 October 2026; 2.75 into it before); the brackets' five screws' length (8 and 7.5, 3.3 into the bowl's wall; 6 until 27 October 2026), their tapped holes in the bowl not cut (bom.py's embed measures their threads in its metal); the outer pivots' heads 3.8 thick (51 00 10's side view, rough; 05 01's elevation 4.1); the knurls of the lock nuts and the bearing discs; the outer pivot's tip at 9 o'clock ending at the ring's inside (05 01 draws it 4.7 long, ±0.5: no longer, so the 1942 bowl clears it turned over);
  • the case brackets' foot stepped to the 1942 casting's band and body (r 54 over 13.5 under the pivots, 53 under: 05 01's 1940 bowl has no step), the foot's 3.7 thickness carried along its length and its slots not drawn, the boss's sides radial; the WN-78 screws' M4 (their heads Ø 7.0 as DIN 84's M4); the locking bracket's back stepping out under the bowl's band (its face, crank and slot measured on 05 01, 24 October 2026), the transit lock's lever (28 × 8 in plan) and knob (Ø 19.6), brass with its hub and pin on the drawings' brass case as on Wempe No. 6161's (antiquorum_208-4_wempe6161.jpg, one photograph: likely; black knob on steel until 18 October 2026; the 6MX's black on steel, as Wempe 5377's and 4585's): scaled on 05 01 at 2.45 px/mm, their forms after Kirow 422's bowl and the sheet's small views; its heights, which no sheet draws, read on Wempe 5377 (bonhams_26816_327) and Greenwich's wartime Wempe against the knob's Ø 19.6 (likely, ±1.5): the lever 2.5 thick (at the pivots' height, in the locking bracket's slot: above, the locking bracket and the transit lock), a hub Ø 9.8 standing 10.8 under the knob, the knob 4.5 thick, its top 4.6 over the rim, the pin Ø 4 standing 16.7 over the lever (16 to 24 October 2026 the lever lay at the box's rim over the ring's top, the knob's top 18 over the rim; until 16 October at the pivots' height, the knob 10 thick on a Ø 4 stem); its bracket's form estimated;
  • the closing plate's screw and knob (WN-71 head Ø 7.2 × 2.5, M3; WN-72 Ø 5.2 × 2.4); its spring (Verschlußfeder 05 01 00-7: Closed, below): the steel wire Ø 1.0, its two turns from r 12.35 to 6.5, the bends r 0.8, the hole Ø 1.2 × 1.5 for its inner end in the bowl's bottom and its stiffness, all estimated; its turn shut, 25° from open (its Ø 8.1 clears the Ø 8.2 hole from 20.8° on; the 6MX's shutter is a pin in a slot, Uy8A88_j-cA 5:21-5:25, so no measure for this one);
  • winding in the case: the bowl turned a full 180° (the videos show its bottom facing up; the instruction says "onto its side"), the way the dial goes (toward 3 o'clock: DIxob2Rl4dc 7:07, likely), and 0.8 s for each step (the turn, the plate, the key);
  • the pins over the outer pivots (05 01's elevation; How the case and box were measured, The box): their size and place scaled on the sheet's pixel profiles (Ø 2.2-2.9 between the lines' centres and outsides, built 2.5; 2.3 proud of the parting face and 3.3 into the wall, ±0.4; likely), their purpose (a lid-locating pin: inferred), their brass, their foot set in the wood without a bored hole (as the handles' screws), and the closed lid's hole for each, Ø 2.7 × 2.6 (no sheet draws the lid; inferred from the pin standing proud) (25 October 2026)
  • the handles' finger recesses' rounding (radius 10, as deep as the hollow: the section's curve fits it at one more point, 2.5 over the foot) and their place along the wall (centred under the strap); the box's floor 12.5, its boards butted, the lid's top a frame 20 wide round a glass 3 thick, the keeper on the front, the handles' screws: as the photographed boxes show them; the hinge one piano hinge 148 long across the back, its leaves 6.5 × 0.6 let into the parting faces and its knuckle Ø 5 in seven lengths (its length measured on Lange 1947's box from above, drouot_33810142_lange1947_2.jpg, 664 px at 4.48 px/mm by the box's 185; every photographed box has one, Wempe 6161's, 5377's and Greenwich's too: two hinges 40 long until 16 October 2026, which this line then said the photographs show); the lid's stay, a brass bar 91 × 6 × 1.5 with a slot 2.4 wide, pivoted 5 over the parting line and 40 from the back on the lid's left wall and riding on a pin 20 under it and 45 from the back on the box's (Wempe 6161's form, antiquorum_208-4_wempe6161.jpg, on the left as Greenwich's; Lange 1947's on the right), its places fitted to the lid's travel so it clears the back wall at every angle (it leans about 45° open, where 6161's stands nearly upright: its pin and pivot aren't measurable on the photographs); the knurls (the bezel's lowest 2.5 of its Ø 125, 0.4 deep, Wempe 6161 and 5377; the bearing discs' edges, 6161 and Lange 1947; the transit lock's knob, every photographed box; the inner pivots' lock nuts, knurled brass discs Ø 19 × 2.5 inside the ring against its face, Wempe 3955's bowl, where they were hex steel nuts outside it); mahogany, as Wempe No. 6161's round its brass bowl and gimbals (Antiquorum 208-4; its colours sampled on the photograph), where the wartime boxes photographed with Bakelite bowls are oak (Greenwich's, Lange 1947's; the Überkasten's sheet says bleached birch);
  • the outer box (51 00_5): its cushions' cover (felt-grey) and form between their figured middles (a lens (1 − u²)(1 − v²), 1 at the rim), the ends' and floor's 20 (scaled 23, 21), the rim strip's 5 and 4, the strap's leather 3 and its loose ends (not drawn), the eyes' form (a disc Ø 23, 2 thick, its screw's round head the Ø 8.5 centre), the hook and eye (commercial: a ring Ø 9 of 1.6 wire, a hook of the same), the hinges' leaves 12 × 1, which numbers of its parts list are which; the catch (19) inside the front not drawn, its form unread; the wooden box's latch and keeper press 2-3 mm into the front cushion
  • the maker's plate's place on the lid (Greenwich's wartime Wempe), its four pins; the latch (31 October 2026, the owner's choice, built as Lange 5591's: js/case.js's LT): the plates' 1.2 thickness, the latch plate's top 8 deep let into the front wall and its face 7 high (the knob's centre at its lower edge, as 5591's), the slots' 2.4 along the wall, the hook plate's face 7 and the hook's 2 below the parting line, the screws' heads; the tongue as the bolt behind the latch plate, out of view in a mortise in the front wall (its 45 can't lie inside the 43.9 plate: its arrangement unknown), its pivot screw (M6, in its Ø 6.2) put in from inside the box and its knob (Ø 9 measured, in its M3.5) the plate's; until then the tongue hung on the lid's front over an estimated keeper; its outline's arcs scaled (the round end R 6 about the hole, the narrow end R 7 through the tip at 39, both edges tapering 1°20' to 8: 10-M9, 28 October 2026; one edge tapering and a semicircle r 4 until then);
  • the key's tube 8.0 under the body's top (scaled on 51 00 11's section, ±0.15; its free length then 33.0, the sheet drawing the tube 48.1 against its own 50.1), the spring's wire 0.6, pitch 0.95 and about 12 coils (scaled), where its end stands round the axis (free on the sheet), the tube end's relief to Ø 4.6, the joint of its body's four solids 0.12 under the drill point's foot (no figure: where they join is the build's, not the part's), the square's 60° sink 1.1 (its tolerance's middle; built 27 October 2026); the pin's length (15.3, scaled on 51 00 11's section); the paddle's colour (black, as the body: no sheet for it); where it lies in the box.
The case, gimbals and the trial (phase 2d): 5
  • the densities (as wempe_inertia.py's; the dial and the hands' Goldin brass 8.45), and for comparison only Bakelite 1.37, hard Buna rubber 1.18, aluminium 2.7 (published typical values);
  • gimbals.js's: the pivots' damping (0.05 of critical), the stops' wood (1e6 N/m, restitution 0.3), the seas (but for their periods and NORDFORSK's limit), the box's place on the ship; the box turned a quarter as the roll and pitch swapped (small angles, the box on the centreline);
  • the balance's pivots' side friction (mu 0.15 on the Ø 0.21 journal) and the beat error's quasi-static offset; the detent's trip threshold from its preload (0.63 g, the Hamilton's after Op. 78) and weight, static;
  • the masses' setting by the adjuster to ±0.45° of their compensating place (the rim's graduation read as eight punch marks of a 60-division; the spread the 6MX passports' adjusted C gives, sd 0.060; ±0.25° until 18 October 2026); the regulation at 20 °C ±0.5 s a day (the cards' largest period means, 0.81-1.90, come from C and S with it; ±2.5 would fail a fifth of the trials);
  • the scatter and drift (calibrated on the 6MX's eight test cards' E and R: a later Soviet chronometer's, not a Wempe's), the barometer (1 % of the air's density day to day) and the oil (the drive lost 0-10 % a year): the Hamilton's estimates; the air's share of the damping from phase 2a's parts (the middle of its range).
The Soviet 6MX (the Variants' third version, Kirov No. 247; The Soviet 6MX above). Measured on Altmeppen's photographs, each fitted on the sheets' hole table; estimated or chosen where they don't show: 17
  • pillar 39's radius, 37.3: its bearing, 190°, is measured to ±0.5° on both faces, its radius reads 36.6-37.3; on r 37 the drawn pillar would stand 0.11 into the drawn seconds wheel's tips, so the radius is the least that clears them by 0.2 (the 6MX's seconds wheel and its pillar's diameter aren't seen);
  • the chatons: each filling the bush's hole (4.59 in the plate, 4.3 in the bridge; about Ø 4.1 shows), the stone flush with the face the arbor's shoulder bears on (so the drawings' endshake), Ø 2.2 × 0.6, its hole the pivot's and 0.025, the sink a cone Ø 3.4 to 1.2: the photographs show the settings' faces only;
  • the balance staff: the first-drawn one with its brass collar (no 6MX photograph shows the hub), taken because the escapement stands at the first-drawn heights;
  • the barrel: the drawings' 1941 three-piece barrel, its cap underneath (Altmeppen, Knirim); the photographs can't tell it from the later drum turned over;
  • the detent: 1436's revised (bracketed) sections and lengths at the first-drawn heights (the 6MX's detent is not shorter: X5); Kirov 14100's slotted foot not drawn, as the drawings' foot's holes aren't;
  • the detent block: 05 04 01-4's figures, read in first angle (the upright 3.5-6.5 under the plate; Kirov-Poljot 24020's reads 4.3-6.6, Kirov 247's 3.9-6.6); the taper at the upright's root (one photograph); the stud's thread and depth in the block, the near pin's 1.2, the slide's depth, the stop screw's thread and head height; the banking screw centred in the upright, its domed head Ø 3.2 × 1.4;
  • the upright's slit (29 October 2026): 0.12 wide (about 0.1 on both photographs, 2 px at 18.3 px/mm), its floor the screw's hole; the slit end extruded 3.5 long.
  • the Geneva stripes: their profile is the photograph's light; the faces the photographs don't show (the plates' inner faces) take the same;
  • the dial: its screws' and pin's sizes the German dial's; the serial's two strokes under the seconds hand (the 4's upright below its crossbar, the 7's bar 0.4 under the head), drawn straight on from what shows;
  • the bowl: its foot below 8.3 under the cone (unseen, run to the drawn floor), its inside (the bore and a cone 3.4 inside the outer one), the rim's top flush with the flange and its notch for the ring's pin; the brackets' plates' and web's thicknesses, the block's depth taken 4.6 with its outer face at the drawings' r 66.5 for the ring's pivot screw, the locking bracket's crank; the closing plate's spring and screw's head height;
  • the bezel's inside and its glass (the 6MX case review, 8 November 2026, 6mx-audit/case/README.md): the glass 3.0 thick (the blind reading about 3, doubtful), flush with the lip as Kirow 422's, Kirova 14100's and Poljot 09777's are (likely), its underside where it was, 0.1 over the cannon tube's crown, so the bezel 14.5 high, a plain band 1.5 between the traced bell and the knurl added to reach it (DEC 422 shows a longer band, on a scale set by its collar); the seat and shoulder not seen; the sight cone (10 November 2026) measured, likely ±1: a white cone from r 52.3 under the glass to r 46.0 at the dial, 0.1 over it, on Kirov 14917 (rAw2zmulpk8 0:22) and Poljot 09777 (Uy8A88_j-cA 2:08) with their bezels on, each frame's camera an ellipse on the minute track (its r 45.13 Kirov 247's, so the scale is borrowed); its paint's colour by eye; Kirov DEC 422's bezel has none (its dial screws at r 49.9 show), so the model follows the two; the knurl a diamond (±28° to the axis, 270 round, 0.4 deep: its pitch doubtful); the thread's pitch 1.0 on Kirov 247's rim (likely), its hand not seen (right-hand, estimated);
  • the brackets (8 November 2026, the same review): their heights read blind on 247's and 2973's bowls (±1-2, 247's taken where the two part by 2-3), their slots' sizes (18 × 5 on the case brackets, 4.5 wide on the locking bracket, square-ended as drawn), the screws' threads (M4) and lengths, the locking bracket's crank (3 thick) and its block's stand-off (8 ±2); the pivots' height kept at 3.0 under the rim (247 reads 0 ±1.7, 2973 2.5 ±1.5);
  • the key: its form the key of 1943's, which no 6MX photograph measures; only its colour, brass, is read (two videos);
  • the boxes: the walls, the glazed lid's glass and frame, the hinges, the lids opening together; the outer box's height, parting line, walls, felt and straps (its width, notch and catch measured: X5); the lid's tab drawn filling the notch;
  • the physics: the 6MX's own escapement, from its bodies weighed on the page's solids (X5); its mainspring the drawings'; the gimbals' periods the drawings' (its bowl, hanging 7.0 lower, weighed apart and run on 7 November 2026: 0.578 and 0.661 s, every check passing; Unresolved);
  • the electrical contact: the lever, staff, stone, cock and spiral's places from Romanov's Fig. 100 mapped onto the photographed parts (±1.5), the base's inner edge (±2), the blocks' outlines, the contact wheel's tooth depth and form, the springs' sizes (±15 %), the jewels' sizes and the staff's pivots (each 0.02 off its cap jewel, 0.04 endshake: no photograph shows them inside their settings; the upper ended 0.1 into its stone until 24 October 2026), every screw's thread; which half-second the contact is open; the stud's outer face kept at r 44.0 and the stone's bevel to 0.3, so the lever swings clear.
  • the contact's drag on the train (7 November 2026, 6mx-audit/delta.md §4, tools/wempe_6mx_contact.js; the owner's choice: recorded as a range, the page's physics left without it): on every other beat the contact wheel's slope lifts the lever 4.6° up a 58° flank against the spiral 59, its work out of that beat's impulse (36 µJ at the escape wheel). Through the 6MX's makeEsc it costs about −4° of swing at a middle estimate of the contact's force (0.14 gf at the screw: a spiral 0.05 × 0.3, one turn), −1.4° to −30° over hairspring-like spirals of the measured size (0.01-1.4 gf), the rate under 0.25 s a day to 0.2 gf; from about 2 gf a lift beat falls under the least drive. The TU holds the contact and contactless 6MX to the same requirements and trial, which rules out the stiff end. Unknown: spiral 59's section and preload (Romanov 1975, p. 208, calls it a hairspring, волосок; the TU gives only the current, 5 mA), and the screw's travel off the disc (about 1.15 mm from the contact wheel's estimated root, the stone's 0.4 rest and Fig. 100's arms), which the drag scales with; each figure is an upper bound for its force (its arm taken whole).
Left out, small: 1
  • Left out, small: the stud's Ø 0.8 round at its point; the balance rim's graduation (05 03 01-1: "8 Körner in 60 Teilung"). The barrel bridge's pin holes' 0.3 × 90° countersinks, the upper cock's pins' 0.2 × 90° sinks (16 October 2026) and the dust bush's bottom, a flat ring 0.5 wide with the face dished 5° to the bore (26 October 2026; a 5° × 0.5 chamfer until then) are in; the dust plate's 0.5 × 45° (05 06 00-1, on the face toward the plates) and the cover plate's outer 45° (05 06 00-4, on its outer face with the sinks; not sized, scaled on the section 0.6-0.7 and drawn 0.65, leaving a 0.15 land: estimated) since 25 October 2026 (fusee.js, chamf).

3. The escapement

The Wempe's spring detent escapement is laid out on one sheet, 05 04_1 (Hemmung für Feder-Chronometer, 10:1): the escape wheel 10.00 mm from the balance, the locking stone 36.5° on from the line of centres, the impulse 21° either side of it. Its parts come from their own sheets: the escape wheel's tooth (05 04_23), the detent and its gold passing spring (1436, 05 04 01-2), the rollers and the balance.

3.1 Solved from the layout (makeEsc)

Method. The shared escapement solver, makeEsc, takes the layout's points put in its frame, the parts' shapes and bodies (the balance weighed on the model's solids, the train behind the escape wheel, the detent about its flexure), and solves the escapement as the Hamilton's is solved: the wheel released from rest chasing the impulse stone, the strike's loss, the unlocking's work, the detent's fall and ride on the tooth it has just let go; then the swing at which the escapement's work each oscillation balances the balance's damping, at the drive 05 06_20's measured force gives through the train. The rate is Airy's sum of the escapement's disturbances with the drawn hairspring's own isochronism (section 8).

Since 11 November 2026 the tooth is let go where the layout has it. 05 04_1's "5°" stands on the balance's arc beside its 6° and 22°10': the impulse stone is 5° inside the teeth's path when the tooth is let go. The model had the discharge stone 9.2° earlier on the staff, an estimate. Put where the sheet has it (the owner's choice), the wheel, let go from rest, reaches the impulse stone only about 5° before the line of centres and drives it through little more than half the drawn impulse, and the swing falls from 194° to 161° at 12 h. On the measured drive no damping the balance's parts allow reaches the sources' 225° (section 5), which two checks report and no longer fail. The drive (05 06_20) and the damping (TF 111 s) are kept, the owner's choice; the model's README has the reading (Readings corrected) and what is open (Unresolved).

Figure 3. The swing over the 56-hour wind, at the drive 05 06_20 measured at each fusee turn (the escapement's equation, ampAt). The green band is the sources' 225-270° (the 1954 instruction for chronometers in general); the model's swing lies under it (section 15). wempe_escapement.js.
Figure 4. The rate over the wind against 12 h from full wind: the escapement alone, and with the drawn hairspring's own isochronism, which takes back only a little of it (the escapement gains about 2 s a day for each 10° of swing, the tooth reaching the stone late after the dead point). Wound daily, the chronometer runs in the first day's span. wempe_escapement.js (rateAt).

3.2 In time (dynamics.js)

Method. The balance, the escape wheel with the whole train behind it, and the detent move as bodies through every contact, beat by beat, on contact geometry worked out apart from makeEsc's: the bevelled discharge stone on the gold spring's square end by separating axes, the stone on the tooth it let go from the tooth's drawn arcs. The result is held against makeEsc's: the swing, where the tooth lands, the detent's fall and its home, the rate and its isochronism. Then the detent's estimates are varied, the chronometer is started by a twist, starved of drive, knocked, and taken to sea.

3.3 The detent as a frame (tripping.js)

Method. The detent becomes a plane frame: beam elements with stretching, bending and their consistent mass for the spring (0.07 × 1.9 mm, clamped at its foot), the blade's two sections, the horn and the gold spring along its drawn path; the square with the gold spring's head and the pipe with the stone rigid. Its modes carry it in time with the balance and the wheel through elastic contacts: the discharge stone on the gold spring, the gold spring on the horn, the pipe on its banking (the block's cross screw), the tooth on the stone and on the impulse stone. Validated rigid against dynamics.js, it is then run as drawn, with its estimates varied, pushed to larger swings, and knocked.

Figure 5. The swing pushed by raising the drive: the stone's largest lift while a tooth is held, as a share of the depth of lock; a dot on the top line is a run where a tooth got past. The dashed line is makeEsc's TRIP. tripping.js --model wempe, section 3.

4. The mainspring and the fusee

The drawing set has a sheet no other chronometer's drawings are known to have: 05 06_20 (Herb. Müller, 24.7.41), the fusee's curve and its measured force, the pull in grams on a lever 262 mm long over the whole wind. It gives the strip (15 × 0.37) but not its length or free shape.

Method. The Hamilton's mainspring.py solves the strip in the drawn barrel (Ø 38.5 on the arbor's Ø 13.3 core) for its shape of least bending energy at every quarter turn, coils pressing on each other, its torque the energy's slope. Its free curve is the Hamilton's measured one scaled to the length, its turns scaled by a set. Set up ¾ turn (the sheet's "3/4 Umgang gespannt") and wound through the drawn fusee and chain, its torque on the fusee arbor at each of the sheet's readings is set against the measured force, each scaled to its mean. Twelve lengths and sets were solved; the best fit is the model's spring.

Figure 6. 05 06_20's measured force on its 262 mm lever (dots) against the solved spring's torque through the drawn fusee, scaled to the same mean (line), over the fusee's turns from run down. wempe_mainspring.py, wempe_mainspring.json.
Figure 7. Each length and set tried: the shape's misfit (rms) against the set (the free curve's turns scaled), one line a length. The sheet's own reading is good to about 0.5 %. wempe_mainspring.py.

The level is the drawing's open figure.

5. The energy budget

Method. Where the drive goes, from the mainspring through the fusee and chain, the train and the escapement to the balance's damping, each loss from the Wempe's own parts: the coils' friction (section 4), the barrel on its journals, the chain's links on their rivets, the fusee arbor's pivots; each mesh tooth by tooth (section 6) and each pivot on its tooth loads; the escapement's strike and unlocking (makeEsc). The balance's damping, which no Wempe source gives, is built from its parts: the air's drag on its masses, timing weights, screws and arm (a pressure drag at the masses' Reynolds number), the endstone's friction under the balance's weight, the spring's own. Then: does some balance within its parts' damping swing the sources' band on the measured drive?

Figure 8. The balance's quality factor Q: what its parts give against what the measured drive, through the train as built, needs to swing the sources' 225-270°; the model's middle estimate marked. wempe_physics.js.

6. The going train, tooth by tooth

Method. The train is four meshes: the fusee wheel (90) on the centre pinion (14), centre wheel 90 on the intermediate pinion 12, intermediate 80 on the seconds pinion 10, seconds 80 on the escape pinion 10; every count, module, tip and root from the wheels' tables, the two pinions no sheet gives fixed by the drawn centres. gearing.js takes the page's own outlines, places each pair at its drawn holes, and for 120 steps through a pitch turns the pinion back against its load until the outlines touch, giving the torque it receives with the teeth's friction (µ 0.10-0.20, steel leaves on brass). The four ratios as their wheels turn drive the balance through the escapement's equation over six hours. Then each mesh's centres are moved until it jams or its friction takes the swing out of the model's band: the window the holes must hold.

Figure 9. The torque each pinion receives over one pitch of its wheel against the nominal ratio, µ = 0.20. gearing.js --model wempe.
Figure 10. The balance's swing over six hours of the train, the teeth's ripple included, at the two friction coefficients. gearing.js, day().
Figure 11. Each mesh's depthing window about its drawn centres (light bar), and within it the most the centres may close with each hole off by the tolerance it sets in any direction and the pivots shaking (dark bar: twice the hole tolerance and the shake, either way); the tightest mesh sets the holes' tolerance for the whole train. gearing_wempe.json.

7. The parts' strength

Method. Fully wound, at the measured level of the drive, every stage lossless (the largest forces): the mainspring's bending stress; the chain's plates, rivets and hooks under its pull; the band; each mesh's teeth by Lewis' formula (Y 0.2, clock teeth); the pivots as cantilevers carrying their arbor's whole load; the maintaining spring, ratchet and click wheel; a strong hand on the key against the stop. Each against its material's published range: the margin at its weak and its strong end (~~ where only the strong end holds).

Figure 12. Each part's margin, its stress against the weak and the strong end of its material's range (logarithmic). Under 1 at the weak end (left of the dashed line) the part holds only if made of the stronger material. wempe_strength.js.

The maintaining spring. Its eye's pin goes up into the ratchet and its other pin down into the fusee wheel (05 06_4's Schnitt A-B: the eye's pin drawn whole, "Stift nach oben", the other hidden), so the drive closes the open ring until its ends nearly meet, its outer edge leaving the wheel's recess; the end at the wheel's pin, its short wide tail against the recess's wall, is held as clamped. Its stiffness from its drawn section (the width traced round it on 05 06 01-1), by Castigliano at small turns and as an elastica as it closes; how far the fusee's drive (05 06_20) closes it and where its ends would meet; its stress; and, the key held, the train run on it, the swing following, with the maintaining detent catching the ratchet up to a tooth late. wempe_sustaining.js.

8. The hairspring and Phillips' curves

The hairspring is a helix (05 03_6: 0.285 × 0.52, Ø 12.8, left-hand), its ends bent to a curve drawn on 05 03_7 as a table of radii every 10°, and cut, before the curves are bent, "auf 13 Umgänge −3,5 mm": to 13 turns less 3.5 mm.

Method. The length is read four ways and each gives the balance a period. The curve's table is fitted as two compass arcs and tested against Phillips' conditions (1861): its centroid on the perpendicular to the junction's radius, at R²/l from the staff, so that the spring pulls the balance with a pure couple. The force it leaves on the pivots under a couple follows (HSPR.lateral), with the helix's turns varied. At large swings the drawn spring is solved as an inextensible elastica by shooting and again as co-rotational beam elements; the two must agree on the couple's shape, the force and the rate.

Figure 13. The rate the drawn spring alone gives against small swings, by the elastica: it stiffens a little as it winds, and loses as the swing grows about the model's swing. HSPR.large on the drawn spring (wempe_hairspring.js, section 4).
Figure 14. The force the drawn curves leave on the pivots under a couple (× R / couple) with the helix's turns between them varied; the drawn spring marked. The two curves' errors cancel at a whole number of turns. wempe_hairspring.js, section 3.

9. Temperature: the cut bimetallic balance

A steel hairspring weakens as it warms, so the balance must open to compensate: the rim, steel inside and brass outside, cut beside the arm, curls in as it warms and carries its four masses toward the staff. How much depends on where the masses sit on the free arc.

Method. Each quarter of the rim is a bimetal cantilever from the arm's edge to the cut, curling by Timoshenko's bimetal (1925), each element's radius moving by the layers' mean growth, the masses clamping the rim over their slot. Every part is weighed on the model's solids. The spring's own loss is its thermoelastic coefficient's (−240 to −270 × 10⁻⁶) with its expansion. The rim's "Nickelstahl Sonderlegierung" is read as Guillaume's 42-44 % nickel steel (estimated); a plain nickel steel is tried beside it. The secondary error (Dent's, 2 s) against Guillaume's nickel steel as NBS Circular 58 tabulates it, checked against the 6MX passports' S.

Figure 15. The rate gained a day a °C against the masses' place on the free arc (degrees from the cuts' line), with Guillaume's nickel steel at the constants' middle (line) and over their 32 corners (band), and with a plain nickel steel. The green band is the TU's ±0.10 s a day a °C; the drawn place (Wempe 3955's and the 6MX's) and the place that compensates are marked. wempe_thermal.js.

The frame, train and escapement in heat and cold

The plates, pillars, bridges, wheels and, unlike the Hamilton's, the escape wheel (05 04 05-1, Ms 63) are brass; the pinions, arbors and detent steel, the gold spring gold, the impulse roller beryllium copper. Over the TU's working range, 4 to 36 °C, assembled at 20 °C, what isn't brass moves against what is.

Method. Each mesh's depth changes by the pinion's pitch radius times the brass's and steel's difference, against gearing_wempe.json's windows; each arbor's endshake by its brass gap and steel length (the bench's stacks, wempe_endshake.js), against the 1958 MX repair book's bands. The escapement runs through makeEsc in the brass wheel's own frame: the steel detent from its brass-held foot (the depth of lock, the gold spring's tip), the rollers on the steel staff, the springs' moduli and the mainspring's torque by their thermoelastic coefficients; at the published constants' middle and at the corner that moves each part most (brass 20.9, steel 10.8 × 10⁻⁶ a °C), with the drawn lock and with the bench's, the window's middle. Where makeEsc loses the impulse, dynamics.js --model wempe in time was run on the same settings and agreed when this was found (27 October 2026: the swing 91-97° then; not run again since the discharge stone was put at 05 04_1's 5°, where makeEsc gives about 80°).

Found. The depths, endshakes, lock and chain hardly notice (under 4 µm). The gold spring's tip does: 1.8-2.3 µm farther from the staff at 36 °C, taking 0.3-0.45° of the 1.50° let-off, and the escapement loses its impulse at about 1.0-1.1° (the same edge that tops wempe_tolerances.js's lock window). The lock window that holds over the range is about half the one at 20 °C, and since 3.45.00 it is the window wempe_tolerances.js gives the bench (the owner's choice): 231.7-237.9 µm, the bench aiming just under its top (237.8) so the teeth's wear, which only makes the lock shallower, has three years' room. At the constants' corner no lock holds at 36 °C (open).

10. The case and gimbals

Method. The bowl with the movement in it and the gimbal ring are weighed on the model, every closed solid at its material's density (massprops.py --model wempe): mass, centre of gravity and inertia about the gimbal centre. 05 01 says to move the case brackets "bis Gleichgewicht erreicht ist", until it hangs level, and the tool does so. The two bodies then run on their pivots under the Hamilton's seaway (gimbals.js: narrow-band random roll, pitch, yaw and heave; NORDFORSK's 0.12 g limit), and what reaches the balance's staff is run through the Wempe's escapement.

Figure 16. The case's lean for each degree of a regular roll about the box itself, from the apparent vertical (gravity less the box's acceleration, which is what the balance's pivots feel) and from the true, against the roll's period; the gimbals' own periods and the balance's beat marked, merchant ships' roll periods shaded. gimbals.js --model wempe, section 2.

11. The 6MX's trial and a voyage

The Soviet copy of the Einheits-Chronometer, the 6MX, was made to a published specification (TU 25-07.1533-84) with a trial: 35 days at 36, 20, 4, 20 and 36 °C, the first two days of each period not counted, wound daily, its figures worked as E (the mean variation of the daily rate), R, D, C (the temperature coefficient), S (the secondary error), each period's mean, and N = 42E + 4.5R + 60C + 2.4S. No Wempe test card is to hand, so the trial is the 6MX's.

Method. Each day's rate is summed from its parts: the temperature curve (section 9, with the masses set by the adjuster at their compensating place to ±0.25°), the air's density, the swing over the daily wind, the day-to-day scatter and drift (calibrated on eight 6MX test cards' E and R), the barometer and the oil; the trial is drawn 2,000 times. Then a 60-day voyage from a winter port into the tropics, rated over ten days in port, gives the error a navigator would carry.

Figure 17. One trial drawn: each day's rate (the balance's linear temperature term taken out, as the essay draws it), the periods' temperatures shaded. wempe_errorbudget.js.

12. Building it: tolerances, the bench, the Monte Carlo

The checks above take the Wempe as drawn. A chronometer made to the drawings is made to tolerances, some the sheets give and some they don't, and is then tested and adjusted on the bench as its makers did. Three tools follow it there, as the Hamilton's tolerances.js, bench.js and montecarlo.js do. They are guidance for anyone who would make one, proved by their own runs: the model was never made.

12.1 How closely the unsized parts must be made

Method. From the escapement as the page runs it (makeEsc with the Wempe's settings, driven by 05 06_20's force through the train), the swing and rate are solved with the escape wheel's torque and the detent's and gold spring's stiffness each scaled (a strip's stiffness as its thickness cubed). Each part's range is the one that keeps the swing in the model's guard band, 147-270° (its floor 14° under the model's own swing at 12 h; the two builds' 150-270°), over the whole 56-hour wind; the sources' 225-270° is reported beside it. The parts: the mainspring's strip (05 06_20's "15 × 0,37", no tolerance), the detent's spring (1436's faint "0,07") and its preload, the gold spring, the depth of lock (05 04_1's 0.235, set by the banking), and the hairspring against the period, which the timing weights must take up.

The drawn detent's ring, and how the drawn version is judged. The lock's floor and the preload's rest on how far a held tooth's stone rings in the frame (section 3.3). With the discharge stone at the sheet's 5° it rings 126.9 µm, 53.9 % of the depth of lock, under half only from a preload of 0.93 g (116 µm, 49 %, with the stone as it was). By the rule of half the depth the drawn version would have no window: twice the ring, 0.254 mm, lies past the deepest lock that unlocks (0.2436 mm), and the preload that halves the ring past its cap (0.80 g). The owner's choice (12 November 2026): the drawn versions are judged on no tooth getting past, the ring reported; the builds, their gold spring thinned at its bend, keep half the depth. The frame run at shallower locks holds every tooth down to 0.130 mm and lets one past at 0.120: the ring is the gold spring's blow on the return swing, 127-133 µm whatever the depth. The tables below, the bench and the Monte Carlo are worked on that floor (tools/tripping_wempe.json, which tripping.js --model wempe --write writes). Still open: the ring turns on one estimated stiffness, the discharge stone's on the gold spring's tip on the return.

12.2 The tests on the bench

Method. Each test turns the model's figures into what a made part or movement must read, and how to measure it with a maker's means: the mainspring's pull against 05 06_20, the fusee's evenness, the depths, the endshake from the sheets' fits, the balance vibrated, its free run-down, the detent's ring, the escapement's figures and the banking's preload, the swing, the maintaining spring, and the 6MX TU's 35-day trial as the acceptance test (the Wempe's sheets give no trial of their own).

12.3 The whole movement made to tolerances

Method. 2,000 movements, each drawing every tolerance at once (uniform within the sheets' limits where they give them, normal where estimated): the arbors' holes, every term of each arbor's endshake stack from its sheet, the mainspring's strip, the teeth's friction, the detent's and gold spring's strips, the wheel's tips and the banking (the depth of lock), the preload, the hairspring's strip, coil, cut length and modulus. Each is judged as made (the train turns, every arbor free, the escapement unlocks and locks in its window, the swing in the guard band over the wind, it times within the weights' reach), then after the bench's tests as a maker applies them (bench_wempe.json's figures): a close mesh re-bushed, endshake set again, a detent spring out of its band replaced, the banking and preload set, another strip, the set-up turned, the hairspring matched to the balance.

Figure 18. The lowest swing over the wind of the movements made to tolerances whose escapement unlocks, as made: how many in each 2.5° band. The green band is the guard band's floor and above. wempe_montecarlo.js, montecarlo_wempe.json.

13. The whole transmission in time

Sections 3 to 7 take the movement a mechanism at a time, each with the others held still or rigid. Here every part that moves is one system: the barrel on its solved mainspring, the chain, the fusee with its two clicks, the maintaining ratchet, spring and detent, the fusee wheel and the going train to the escape wheel, the motion work and hands, the up-and-down work, the key and the stop: each a body with its inertia weighed on the model's solids (or estimated, said where), joined by the elements a real movement has, none of them linearised.

Method. wempe_transmission.js. The elements: the mainspring's torque as solved (section 4) with its coils' friction as a stick-slip element, so winding takes more than running gives; the chain in tension only, so it can go slack, its lever on the fusee where it lies and rising and falling as each 2.05 mm link comes onto a drum; each pawl (the clicks, the maintaining detent) a body on its pivot under its spring, riding the teeth's backs and catching a face only if it is down; the maintaining spring's drawn ring, stiffening as it closes, with its bank; every mesh tooth by tooth on the page's own outlines (section 6's), the driven gear's place under load through each pitch, the backlash to its back flank and the teeth's friction, contact on either flank or neither; the cannon pinion's friction fit, which slips past its torque; the key's one-way coil spring (it grips whenever the hand holds the key against the fusee, and slips only if the key is turned the wrong way); the stop bar on its pin. The escapement is dynamics.js's own (section 3.2), its escape wheel driven each step by the train through the escape pinion's mesh, at its own place and speed. The beat runs from 8 starts of the train's meshes, and each mesh's peak is judged on the worst of them, a figure that holds as the step is refined (wempe_transmission_converge.js: the seconds wheel's, the escape pinion's take-up after the escape wheel lands on its stone, within 2 % from a quarter of the step to an eighth). It is stepped in time at a microsecond. The stiffnesses no sheet gives (the teeth's give, the chain's, the clicks' faces, the bank) are estimated from the parts, and those the results rest on are varied.

Watch the run: the movement in plan, every part coloured by the load it carries against its running load and its deviations from rigid motion magnified, as a finite-element post-processor shows a deformed shape; a beat, the key let go outright and eased back, and the stop, played in slow motion with their time histories (wempe_transmission_trace.js).

Figure 19. The key let go at the end of a stroke: the force with which the click catches the click wheel's tooth against how far short of its face the nearer click was. Let go outright, the fusee and barrel run free to the click; eased back, the hand lets them forward under its hold. The dashed lines are the force at which the tooth's brass reaches its yield (300-450 MPa) and the running load. wempe_transmission.js.

Every moving part in 3D

The runs above give the forces between the parts; here each moving part's own stress under them. motion3d.py takes the page's solids and moves each part as the page's own update() moves it; motion3d_fe.py solves each part as finite elements (voxel bricks on its solid, a modulus by its material, finer windows at its contacts, the root's bending kept apart from the contact's pressure) at its worst load in the four runs, held where it is held; the springs (the mainspring by its solved curvature, the hairspring, the maintaining spring, the gold spring) from their large-deflection solutions. Each part is read against its drawn material: the 1950s build beside the drawing's.

Open the 3D viewer: every moving part in 3D through the four runs, coloured by the load it carries against its load in running (the plan above in 3D, with its moments and its probe of every joint's force), by its stress, its stress against its material's limit or its speed, the deviations magnified, with each part's load, margin and motion, the kinematics against the tooth counts, and the assumptions.

Every part past its limit but the mainspring (its inner coils, fully wound: section 17's builds and wempe_build.js take it up) is past it only with the key let go outright. Those peaks lie at the contacts, and the methodology's contact study finds them the contact's own pressure: it halves as the load is spread twice as wide, so a real blow dents the click's nose and the tooth's face where they meet. By the beam formulas only the maintaining ratchet's teeth bend past brass (about 500 MPa at the blow); the click wheel's tooth carries about 200 MPa in the contact study (240-280 until 26 October 2026), inside it. The let-go study (wempe_letgo_fix.js) runs that let-go (the worst place, the nearer click 97 % of the way to the next; and eased back beside it) with the fixes a maker could make: more clicks, evenly spaced over a tooth, so the fusee runs a shorter way free; a brake, friction between the click wheel and the maintaining ratchet, which the clicks hold still running (so it costs the train nothing) but which the key works against winding; the click's face cushioned. The stresses are the 3D results' scaled to each case's force; the click wheel's tooth by wempe_strength.js's beam, the highest of three readings (the FE on the tooth gives 0.49 MPa a newton (0.60 until 26 October 2026), wempe_ratchets.js 1.16, the beam 2.05: the methodology's cross-checks).

14. Against a period text

The drawings decide the Wempe; a period text says how a detent escapement was usually proportioned, and so where the Wempe's makers chose otherwise. keystone.js --model wempe measures the escapement against Watch and Clock Escapements (The Keystone, 1904), Ch. III: the wheel and roller, the balance planted where the rims cross over a pitch, the impulse and unlocking arcs, the locking stone, the tooth's face and back, the detent and its spring, the balance. The text is generic, so the tool reports and never judges; where the two differ, the drawing and its sheet are named.

15. The swing measured on video

No Wempe source gives the balance's damping, the figure the swing rests on most. A running 6MX, whose balance is of the same kind (a cut bimetallic rim, four masses, two nuts on the arm), is filmed from above at 30 frames a second in moMvyx8eXfo, 17:53-19:50 (another person's video: linked at the moment used, no frame copied here).

Method. wempe_swing.py reads the balance's arm in every frame as a drop in saturation along the rim's ellipse, folded every 180°, and fits each 4-second window with a sine: the swing, its alias settled by the frames between the turning points. The movement is set down at 17:52 and the swing rises; it is turned in the hands at 18:46 (the 6MX's start) and the swing falls back. wempe_startup.js fits all the stretches together with the Wempe's averaged balance equation, d(A²)/dt = 2 Wn(s, A)/T − 2A²·damping/TF, the escapement's work Wn from makeEsc with the Wempe's settings, the drive and each stretch's start free: rising and falling toward one swing, they fix TF, the free run-down.

Figure 20. The 6MX's swing read off the video (dots, each a 4-second window), with the fit through all the stretches: the Wempe's escapement work and the damping split as its parts give it (line), and the plain A² law (dashed). The time is the video's, in seconds. wempe_swing.json, wempe_startup.json.

The clicks of two more running chronometers were read with this escapement in time: a Lange VEB (gnIP3NqNbEU, 0:03-1:03) and a Kirova under a dome (KxInG08LWXg, 0:00-0:34). Their pairings leave the swing between readings 80-120° apart, since which click is which isn't settled, so no swing is read from them (References/wempe/VIDEOS.md, "Measured").

16. The Soviet 6MX

The page's third version (Variants, Version) is the 6MX the 1st Moscow Watch Factory made from these drawings, built as one photographed chronometer is made: Kirov No. 247 of February 1952, from Johannes Altmeppen's photographs, with videos of other 6MXs where his don't show a part. Each difference from the drawings is measured on a photograph, fitted on the sheets' hole table, or marked estimated; the drawings' two versions are left as they were (every view the same to the pixel). The section below is the model's README as it stands: how each difference was read, and how sure. The page keeps the drawings' swing and rate in every version; what the 6MX's differences would change is worked out apart (wempe_6mx.js, below).

How each difference was read (the README's "The Soviet 6MX")

Status (18 October 2026): complete. X0-X4 (16 October 2026) built the switch, the movement, the dial and finish, the case and boxes, the cards, the essay's paragraphs, the report's section and the physics worked apart; X5 (18 October 2026, below) settled what they left: every one of the 6MX gallery's photographs read for the model, the parts measured on them built (the escapement's collet and detent block, the lower cock, the ring's screws, the bowl, brackets, closing plate and outer box), its electrical contact built, and the page running the 6MX's own escapement. What no source shows stays in "Estimated" (The Soviet 6MX) and "Unresolved"; the asks for the photographs that would settle it in REQUESTS.md.

The Variants' third version, Soviet 6MX (Kirov No. 247, 1952; rev=6mx in the link): the 6MX (6МХ) the 1st Moscow Watch Factory made from these drawings after 1946-47, built as one photographed chronometer is made, as Wempe 3955 stands for the first-drawn sheets. Kirov No. 247 (February 1952, first generation) has Altmeppen's full set (photos/dg-chrono-6mx/: its dial, top plate on millimetre paper, dial side, four obliques, detent and box) and his arrowed comparisons with a Wempe. Its electrical contact (E-Kontakt) is left out. The plan is PLAN-wempe-6mx.md at the root.

X0, the switch (16 October 2026). The version control (heading "Version") has three buttons. userData.rev now lists the versions a group shows in ('last 6mx': the revised sheets' pieces the 6MX kept; 'first'; '6mx' for its own), setRev(v) shows a group if its list holds v, and each onRev callback gets the version's name. Until its own parts are built, the 6MX shows every revised piece: its views are pixel for pixel the last-revised ones (views.py --model wempe --eval "__mv.userData.setRev('6mx')"). Both earlier versions rendered unchanged, every view 0 px. The tools: audit.py --rev 6mx (its keys movement.6mx, box.6mx), ci.py's wempe-audit-6mx and wempe-exploded-6mx; bom.py's dev entries take a list of versions in rev; app.js's cards per version in REVC.

How Kirov No. 247 is measured. Its top plate (2_Kirow_247_A.jpg, square-on) is put on the sheets' hole table by a similarity on four arbors seen on it (the fusee's square in the cork ring, the barrel's square, the balance's cap jewel, the escape wheel's endstone) and pillars 11 and 49: 8.98 px/mm, every residual under 0.55 mm; holemap.py --model wempe lays it over the model's plan hole by hole (it now takes --model). Its dial side (3_Kirow_247_uZ.jpg) the same way on seven holes (the centre, seconds, change wheel, intermediate, escape and the intermediate bridge's two screws): 8.63 px/mm, under 0.42. The millimetre paper under the top plate gives the paper's scale (8.8 px/mm), not the plate's, which stands well above it. Screws' centres are circles fitted to their edges.

X1a, the frame (16 October 2026). Each piece and how sure:

X1b, the escapement (16 October 2026). The 6MX has the escapement at the heights the sheets first drew, with the revised block's form:

X1c, the barrel (16 October 2026). The barrel cap underneath (Altmeppen's list; Knirim's, against Lange's Kal. 100): the drawings' own barrel, the 1941 three-piece barrel the first-drawn version builds (0207, 020701; its arbor's core and the mainspring 0.44 lower with it): likely. Kirov 247's side views (4_, 6_Kirow_247_N, S) show a drum about Ø 41.5 × 18 under the chain, too fine to tell that ring from the later drum turned over. The maintaining spring stays the revised one. Not built: the lower balance cock's other shape (Unresolved). Since 28 October 2026 the drawings' 1941 barrel is read with its lid on the bridge's side and its bottom on the pillar plate's (drawing check 2, 4-M1): the 6MX is built with that barrel, and which side the 6MX's own cap is on (Altmeppen's "underneath") isn't modelled apart: its discs are the same sizes either way.

X2, the dial (16 October 2026). Kirov No. 247's dial (dg-chrono-6mx/1_Kirow_247_Z.jpg), traced as the two German dials are (dialtrace.py --dial kirov, into js/dialface_kirov.js, WDIALFACE_KIROV; the Variants' dial switch, 1st Moscow Watch Factory, which choosing the 6MX selects):

X2, the finish (16 October 2026). Geneva stripes on the plates, cocks and bridges (Kirov 247's, 1952; the Wempe's are hammered, drawn plain gilt): straight, 3.95 mm apart (an FFT on plain patches of its top plate, 3.73-4.28 at 8.98 px/mm, and its dial side, 3.71-4.35 at 8.63: the median), running at 50 deg in each face's own photograph (49.8-50.3 and 48.5-52.1): in plan along (-0.643, 0.766) over the top plate, cocks and barrel bridge as one sweep (its stripes run on across them), and along (0.643, 0.766) over the pillar plate's dial face and the intermediate bridge (each plate striped from its own face). Across a stripe the photograph's brightness, folded on the pitch over two patches and averaged (each to its brightest), falls slowly from 1 to 0.61 and rises quickly back: the gilt material's map (frame.js's G6), on solidGeo's plan UVs. Likely: the profile is the photograph's light; the faces it doesn't show (the plates' inner faces) take the same. Altmeppen's "the finish changed in the late 1950s" is later than Kirov 247, which is already striped. The barrel and intermediate bridges are built per version for it, their pins with them.

X3a, the bowl and bezel (16 October 2026). The 6MX's bowl isn't R-05 01 00-1 outside: no Ø 118 flange, but a threaded rim Ø 110.5 for a knurled bezel screwed onto it (Altmeppen's "the bowl and bezel brass", and Kirow 422's knurled bezel). Inside it is the drawings' (the bore, its round, the bottom, the recess and closing plate, the winding hole's seat): the movement is the drawings' and stands in it as drawn. Each reading and how sure:

X3b, the boxes (16 October 2026). Kirov No. 247's box and outer box (dg-chrono-6mx-gallery/leben-des-6mx/13_, 14_Kirow_247_Kasten_*.jpg), built in the drawings' box's place (the gimbals, pivots and transit lock as drawn):

X4, the 6MX's physics (16 October 2026). tools/wempe_6mx.js (with wempe_inertia.py --rev 6mx, which weighs its escapement's bodies on the page's solids in that version) works out what its differences change; the page keeps the drawings' swing and rate (the user's choice):

X5, completing it (18 October 2026). Every photograph in the 6MX gallery (dg-chrono-6mx-gallery/, thirteen more 6MXs), Kirov 14100's parts laid out (chasy-ru-9985-kirova-14100/) and Romanov's Proektirovanie priborov vremeni (1975, Figs. 95-101 and their parts lists) read for the model, each part then measured as below (the 1950s Kirovs first, Kirov 247 before the rest; the user's choice, 17 October 2026). Each reading and how sure:

Checked again against the photographs (29 October 2026). The 6MX's escapement, its block and the parts at their heights, each read on the photographs before the model's figures were looked at (Kirov 247's 8_Kirow_247_F.jpg at 18.3 px/mm, scaled by the escape wheel's tips, 257.5 px for Ø 14, and its pinion's Ø 2.97, 54 px; Kirov-Poljot 24020's 9_KirPol_24020_Feder2.jpg at 26.6): WEMPE-VARIANT-ESCAPEMENT.md at the root has every reading. Changed:

Matched: the base's 6.5 at its end, the stop screw's head (Ø 4.4 ±0.15, the model's 4.2) and place, the collet's length (5.19 ±0.1) and its diameter at the wheel, the wheel's rim 1.37. Not settled by any photograph: the adjusting slide and stud, the lower cock's plan and step (Unresolved), the upright's height (3.45 ±0.3 read against 3.0: possible) and the screw's axis 0.4-0.6 under its middle (one view: possible).

The sheet, Soviet 6MX against the drawings (18 October 2026). The panel's section after Variants lists every difference the 6MX version builds (app.js's X6: 22, grouped as frame, escapement, barrel, electrical contact, ring and dial, case and boxes), each with what the drawings have, what the 6MX has and how sure the reading is, as this section gives them. Choosing one turns the model to the 6MX and closes in on it, the rest see-through (VIEWER.focusOn, shared/viewer.js: the parts' box, or a place where a part's own box would lose the change, as for pillar 39, the hand-setting holes, hole 53 and the moved ring and dial screws); Drawings and Soviet 6MX flip the version where the camera stands; a view chosen, Reset, Back or Whole model ends it.

The geometry checks in the 6MX

The browser checks of section 1, run with the page set to the 6MX: its own parts, its case and boxes, against the same expected contacts and leftovers.

Its physics against the drawings

Its jewels in chatons change the holes' tolerance, not the train's efficiency (oiled steel in ruby and in brass has the same friction in the model's sources); its escapement at the first-drawn heights, with the first-drawn balance staff and collar, changes the bodies only, weighed on the page's solids in that version.

The escapement of 1943 against the drawings

The page's fourth version stands its escapement as the German movements photographed with the 1943 detent block are made: the first-drawn balance, rollers and escape wheel at their heights, the 6MX's detent and the block with its banking screw (the model's README, "The escapement of 1943 as photographed"). Its plan is the drawings' (the banking screw sets the drawn depth of lock), so, as for the 6MX, only its bodies change: weighed on the page's solids in that version and run through makeEsc (wempe_1943.js: the swing at the reference 160.7°, as the drawings'), its bodies within 0.3 % of the drawings', whose run in time stands for it (dynamics.js --model wempe: 300 beats, no tooth lost; the swing 160.5° against makeEsc's 160.7°; run on the 1943 bodies themselves before the discharge stone's move, 198.7° against 197.6°). The page itself runs the drawings' escapement in this version.

17. The builds: a Wempe to navigate by

What a maker would build to navigate by: about two miles of longitude a month, 8 s after 30 days at sea. As drawn, the balance swings 155-166° over the wind, under the sources' 225-270°; the depth of lock has a window of a few µm at best (section 12), too fine to set and keep as the stone and the teeth wear; and set as the 6MX passports show adjusters set the masses, it carries about 17 s after a month (section 21). Two builds answer it, and the maker chooses (the model's README, "The build"; each choice the user's, 19 October 2026):

The drawn versions' cross-checks between makeEsc, dynamics.js and the tripping frame are calibrated on the drawn detent; on the thinned gold spring three of them part on detail of the detent's motion (none bears on its safety), so each build is judged on what decides whether a made one works (wempe_build.js).

Since 11 November 2026 (the discharge stone at 05 04_1's 5°) both builds have their hairspring set against the escapement, a bench adjustment (the escapement gains with the swing; the spring is set to lose as much: the owner's choice), and their swing's floor is 150°. Both builds pass every line below: the 1950s build swings 197-209° over the wind, the 1940s build 174-182°, both under the sources' 225°. On 11 November the 1940s build was reported here as failing (a held tooth's stone lifted 84 % of the depth of lock, a tooth past at 1.02 times its drive). That was a fault in the tripping tool, found and fixed on 12 November: it pressed a held tooth on the stone through the one step in which the tooth comes free, throwing the wheel back at up to 11 rad/s, and in about 1 beat of 100 the tooth met the returning stone (5 wrong beats of 560 before the fix, none of 560 after).

Made to the tolerances, and after the bench

Each build's tolerances, bench and Monte Carlo, run as section 12's with the build loaded (MC_WREV=build40, build): 2,000 movements each.

18. The mechanisms in turn: wear, the key, the motion work, the estimates

The Hamilton's model was checked mechanism by mechanism, as a maker would take its parts, by five tools beside the escapement's and the train's (its build report's section 11): wear, the key forced, the motion work, the train on sprung meshes and the estimates swept. The Wempe's counterparts follow, each reading the Wempe's own data (wempe_strength.js's loads, gearing_wempe.json's windows, tolerances_wempe.json's lock window, mechanisms_wempe.json, wempe_inertia.json) and its sheets. The train on sprung meshes (the Hamilton's traindyn.js) has no counterpart here: section 13's wempe_transmission.js already runs every mesh tooth by tooth through the beat (the meshes' forces, which rattle, the seconds hand's bounce) and weighs every body on the solids (its section 1; the escape wheel's inertia against WESC.SET's is invariants.py --model wempe's), so it is cited, not repeated.

Wear: how long each contact keeps its figure

Archard's law at every sliding contact (V = k F s / H; k from Rabinowicz, 1e-7 well oiled, 1e-6 poorly, 1e-5 once the oil has gone; the escapement runs dry, as a chronometer's does, at Rabinowicz's unlubricated non-metal on metal, 1.5e-6), the loads wempe_strength.js's fully wound, each depth against the tolerance it eats. An escape tooth's tip (gilt brass) wears as a wedge whose flat widens. The line is a service interval: the 1954 Soviet instruction's cleaning and fresh oil every three years (6MX-figures.md, п. 19); the TU's ten years' service life is reported against.

Figure 21. The share of its tolerance each contact wears away as it runs (well oiled; the escapement dry at 1.5e-6), year by year, as last revised (solid) and the escape tooth's tip in the 1950s build (dashed); the shaded band the three years between overhauls. Past 1 the figure is gone: the balance's endstone wears its endshake band (a repairer's setting the overhaul restores) and the drawn escapement's 15 µm lock window is spent by its brass teeth's wear in about four years.

The key forced

A strong hand on the key's Ø 34.5 paddle (Imrhan and Munoz-Vasa's handles carried on past their 19 mm, an extrapolation; Swain's 1.06 N·m beside it), wound against the stop, and turned the wrong way. The 1943 key (51 00 11) is a one-way clutch, its coil spring gripping the tube one way ("Linksaufzug b. Rechtsdrehung entspannt sich die Feder"); the 1940 key it replaced is one-way too, its click spring drawn round its pipe (05-01_6; read as a plain key until 28 October 2026). The table's wrong-way lines are what a key that grips both ways would do.

The motion work and the up-and-down work, made on a CNC

Each mesh tooth by tooth on dial.js's tables, its depthing window (closer until its tips reach the roots, wider until it loses contact or its hand has a quarter of an hour of play) and the hole its stud must be bored to (gearing.js's rule, the window's close side less the stud's shake, halved). The cannon pinion's friction window is wempe_transmission.js's (section 13), put as a weight hung on the minute hand.

The estimates swept, with the bench's correction

What wempe_tolerances.js (section 12) doesn't sweep: the balance's free run-down (TF) over 80-200 s with the strip the bench's run-down test chooses, the pivots' shake against each mesh's window, the hairspring's length read four ways off 05 03_7, and its outer terminal curve formed by hand; for the drawings as last revised and both builds.

Figure 22. The swing over the whole wind (band: its lowest to its highest) against the balance's free run-down TF, with the strip the run-down test chooses: as last revised and the two builds, against the guard band (shaded). The model's TF, 111 s, is a running 6MX's on video (section 15).

19. Making it: the frame under load, the order it goes together, each part on the machine

Sections 7, 12 and 17 ask whether a Wempe made to the drawings would hold together, keep time and pass its bench. Three more checks ask whether it can be made and put together at all, each on the page's own solids: the parts that hold the pivots solved as elastic bodies under the mainspring's load, the movement taken apart and put back in the order a repairer's book gives, and every machined piece read against what a lathe, a mill and a drill can cut.

The frame under load

fea.py --model wempe (the Hamilton's fea.py, its parts, arbors and fixings in wempe/checks.py's FEA): each part that holds a pivot of the going train (the pillar plate, the top plate, the barrel bridge, the fusee bush and the intermediate bridge) cut into brass bricks (E 100 GPa, ν 0.34) from its solid, at two sizes, the coarser a check on the finer; loaded fully wound and run down with the chain's pull on the drums' common tangent (the spring's moment at 05 06_20's level, wempe_physics.js, on the barrel's pitch radius; the fusee's radius and the chain's height at that turn) and each mesh's tooth force with its pressure angle (gearing_wempe.json), shared between each arbor's two bearings by the lever rule; each part held at its screws, the top plate on the three pillars (Ø 8.1, 19 long, bending as beams), the pillar plate at its rim in the movement ring, the parts put together by superposition. What it leaves out: the arbors' own bending, the jewels' settings (taken as the brass round them), friction at the fixings, the cocks (they carry the balance, none of the train's load).

The order it goes together

The Einheits-Chronometer has no overhaul manual of its own in the set; the Soviet Navy's repair book of 1958 (Эксплуатация и ремонт измерителей времени, pp. 59-60, "Порядок разборки и сборки") gives the order for the MX, its copy on the same frame, step by step: the hands, the dial, the motion work and the up-and-down pinion from the dial side; then, dial down, the balance's cock and the balance, the detent's block, the barrel arbor's ratchet, the barrel bridge, the fusee arbor's cork washer, flange and washer, the top plate on its three screws, and the rest; and together the same backwards. assembly.py --model wempe rasters every part as assembled into columns 0.2 mm apart (exploded.py's and overhaul.py's method), freezes the movement in one state, and draws each step's parts straight off along the arbors, or along a path of straight legs where a part leaves sideways: a pair fails if one passes through the other, or is dragged the long way out of a fit it sits in. Where the model's parts can't follow the book, or a real movement takes a part off another way (unhooked, unwound, turned to a gap, slid sideways), the step says so.

Each part on the machine

dfm.py --model wempe reads every machined piece off its solid (screws, jewels, springs, the chain, the dial's enamel and the hands left out: bought, stamped, rolled or wound): its sections sampled between every two heights at which it has a flat face, on a grid 0.01-0.06 mm square as the piece's size allows. It reports, against stated limits: walls in plan under mm (fragile in brass; teeth left to the gear cutter), slots narrower than an end mill of radius mm (a smaller mill, a slitting saw or EDM), the inside corners that mill leaves rounded, round holes and their depth over diameter (over 10 peck drilling, over 20 a gun drill or EDM; under Ø 0.2 a micro drill), the thinnest floor along the arbors, what is cut from each face and whether the piece is round about one centre (turned) or nearly (turned, then cut), undercuts (material above and below a gap: a lathe's grooving tool, or a side cutter on a milled piece), the setups that follow, and any void sealed inside one piece (it fails: no tool reaches it).

Every piece

20. The build book: what each part is made of, and how

The page's Making it (the Hamilton's maker's sheets, in the Wempe's words): each part card has its maker's sheet, its lines of the parts list with the fit bom.py measured on the model, the material and treatment its own drawing gives (the title block's Werkstoff, the notes' hardening and tempering, the surface marks, read off References/wempe/transcriptions/), and how it is made; its drawing to scale, drawn from the model's solids; and the whole build book to print, ending with the bench's tests for the 1950s build (section 17). Nothing is filled from practice: a part whose sheet leaves the Werkstoff blank, or which has no sheet in the set (the Wempe-Norm screws and pins, the parts drawn only on an assembly), says so. tools/wempe_makers.py writes the data (wempe/js/makers.js; --check in ci.py), book.py --model wempe checks the book as paper (every shown line on a sheet, each with a material or the reason it has none and a fit, every drawing at its stated scale and legible on A4).

235 lines as last revised (14 assemblies): 60 parts with their material on their own drawing, 161 without (56 Wempe-Norm parts whose standard sheets aren't in the set; the rest drawn only on an assembly, or with the Werkstoff left blank). 166 carry a fit measured on the model.

How it is madeLines
CNC milled or turned, the material the maker's choice64
bought in or turned56
CNC milled or wire-cut27
bought in or made by a case maker22
CNC milled or turned18
bought in17
CNC turned8
by hand and special supply2
special supply2
by hand1
special supply and hand work1
special supply or by hand1
cast and machined1
joinery1

What needs a hand, or a special supply

DrawingPartMaterial (its drawing)How
05 05 04-2Fassung (the seconds jewel's setting)not on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 05 04-1Lagerstein f. Sek. (the seconds arbor's lower jewel)not on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 03 01-1Unruhreifen (balance rim with arm, nickel steel and brass)„Nickelstahl Sonderlegierung u. Messing“ (nickel steel, special alloy, and brass)by hand and special supply: the cut bimetallic rim: 05 03 01-1's „Nickelstahl Sonderlegierung u. Messing“, a nickel-steel special alloy with brass fused or brazed on, turned, then cut through beside the arm and its masses' holes drilled; no stock supplier makes it: a maker fuses brass onto the nickel steel or buys a balance blank, then adjusts it for temperature on the trial (wempe_thermal.js, the Wempe README's "Physics: hairspring and temperature")
05 03 02-1Spiralfeder (helical hairspring)„Bandstahl 0,52×0,285 poliert“ (strip steel, polished)by hand and special supply: the helical hairspring: 05 03_6's „Bandstahl 0,52×0,285 poliert“ wound 15 turns left-hand on a form, hardened and tempered dark blue on it, its terminal curves bent to 05 03_7 by hand with curve pliers on a drawing of them; vibrated against the balance (its tolerance)
05 04 03-2Hebestein (impulse stone, sapphire)not on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 04-2Auslösestein (discharge stone)not on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 01-1Gangfeder (spring detent, drawing 1436)„Stahl“by hand: the spring detent (1436, „Stahl“): one piece of steel, its spring (Federung) 0.07 thick (0.09 at its ends) and 5.0 long filed and stoned to its ring (the bench's test), the locking stone set in its pipe (Röhrchen); 1436 gives no hardening
05 04 01-3Ruhestein (locking stone)not on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 01-2Goldfeder (gold passing spring)„Goldblech 333 federhart“ (gold sheet, 333 fine (8 ct), spring-hard)special supply and hand work: 05 04 01-2's „Goldblech 333 federhart“: 8 ct gold sheet rolled spring-hard, from a bullion or jewellery supplier; cut, thinned at its bend (the build) and stoned
05 04 08-1Deckstein (endstone), balance, uppernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 06-2Fassung (jewel setting), balance, uppernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 07-1Lagerstein (hole jewel), balance, uppernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 06-2Fassung (jewel setting), balance, lowernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 07-1Lagerstein (hole jewel), balance, lowernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 08-1Deckstein (endstone), balance, lowernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 08-1Deckstein (endstone), escape wheel, uppernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 06-2Fassung (jewel setting), escape wheel, uppernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 06-1Lagerstein (hole jewel), escape wheel, uppernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 06-2Fassung (jewel bush), escape wheel, lowernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 06-1Lagerstein (hole jewel), escape wheel, lowernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 04 08-1Deckstein (endstone), escape wheel, lowernot on the drawingbought in: a synthetic jewel: choose the standard hole and outside diameter nearest the drawn (the hole to the pivot plus 0.01-0.02), or have it made
05 07 00-6Zugfeder (mainspring)not on the drawingspecial supply: the mainspring: 05 07 00-6's strip, or the build's Nivaflex 45/5 (Straumann's, from a mainspring supplier) cut to 1240 mm, its ends formed, set
05 07 04Kette (fusee chain)„Bandstahl 5×0,3; 10×0,9; Präz. Rundstahl 0,6“ (strip steel 5 × 0,3 [the link plates, inferred]; 10 × 0,9 [the end hooks, inferred]; precision round steel 0,6 [the rivets, inferred])special supply or by hand: the fusee chain (05 07 04): links blanked from thin steel strip and riveted on pins; a chain from a watch-parts supplier, or blanked with a press tool
05 07 03-1Band (fusee band)not on the drawingspecial supply: the fusee band (05 07 03, 0.22 +0.03 thick, 730 ±3 between its hooks; no width or material on its sheet): a rolled steel strip from a strip supplier, its hooks brazed on (1.3.43); a post-war GUB Glashütte's band was Sandvik's, 730 mm between its hooks and about 0.2 mm² across (chronometerbook.com, post 35, 10 January 2019). Fitting it (the same post): too springy to lay on the fusee beforehand, it is hooked to the fusee in place and wound on under tension, then hooked to the barrel

The materials the drawings give

DrawingPartMaterialTreatment
05 08 02-1Unterplatte (pillar plate)„Messingblech 4 / SonderMs α-βGef.“none on the drawing
05 09 01Oberplatte (top plate)„Messingblech 4 / Sonder Ms α-βGef.“ (special brass, α-β structure)none on the drawing
05 08 00-1Pfeiler (pillar)„Rundmessing 8,5 Din 1758 Ms 58“ ("hart" struck through)none on the drawing
05 08 05-1Federhausbrücke (barrel bridge)„Messing blech 3,5 Din 1751 Ms 58“untoleranced sizes ±0,1
05 08 05-2Sperrkegel (click)„Messingblech 3 Din 1751 Ms 58 federhart“untoleranced sizes ±0,1
05 08 01-2Schneckenfutter (fusee bush)„Rundmessing 13,5 Din 1758 Ms 58“none on the drawing
05 08 01-3Staubfutter (dust bush)„Rundm s. 15 Din 1758 Ms58“none on the drawing
05 08 03-1Unruhkloben oben (upper balance cock)„Profilmessing 50×15×3,3×5,5 Ms58“ (the "5,5" likely)blank (Rohmaß) Profil ×18; untoleranced sizes ±0,1
05 08 00-2Unruhkloben unten (lower balance cock)„Profilmessing 20×10×8×3,3 Ms 58“blank (Rohmaß) Profil ×18
05 05 03-2Minutentrieb (centre pinion and arbor)„Präz. Rundstahl 7,2 Din 175“„härten, 295° anlassen“ (harden, temper at 295°); surface ▽▽▽ ▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
05 05 03-1Minutenrad (centre wheel)„Messingblech 1,5 Din 1751 Ms 58 federhart“surface ▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,01
05 05 02-2Zwischentrieb (intermediate pinion and arbor)„Präz. Rundstahl 5,2 Din 175“surface ▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
05 05 02-1Zwischenrad (intermediate wheel)„Messingblech 1,2 Din 1751 Ms 58 federhart“surface ▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,01
05 05 01-2Sekundentrieb (seconds pinion and arbor)„Präz. Rundstahl 3,5 Din 175“ (precision round steel bar 3.5 mm, DIN 175)surface ▽▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
05 05 01-1Sekundenrad (seconds wheel)„Messingblech 1,2 Din 1751 Ms 58 federhart“ (brass sheet 1.2 mm DIN 1751, Ms 58 spring-hard)surface ▽▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,01
05 03 01-1Unruhreifen (balance rim with arm, nickel steel and brass)„Nickelstahl Sonderlegierung u. Messing“ (nickel steel, special alloy, and brass)„∇∇∇ poliert“; surface ∇∇∇ ∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,03
05 04 02-2Unruhbutzen (balance hub, brass)„Ms.“ (brass)
05 03 02-1Spiralfeder (helical hairspring)„Bandstahl 0,52×0,285 poliert“ (strip steel, polished)„härten, dunkelblau anlassen“ (harden, temper to dark blue); blank (Rohmaß) 0,52 × 0,285 × 603
05 03 02-2Spiralrolle (hairspring collet)„Messingblech 3,5 (overwritten, likely 3,5) Din 1751 Ms 58“surface ∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished)
05 03 02-3Klötzchen (hairspring stud)„Flachmessing 15×7 Din 1759 Ms 58“surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished)
02 04 02Heberolle (impulse roller, beryllium copper; 05 04 03-1 in the coloured section)„Kupfer Beryllium, Saphir“ (beryllium copper; sapphire)none on the drawing
05 04 05-1Gangrad (escape wheel, 15 teeth)„Messingblech 1,6 Din 1751 Ms 63 federhart“ (brass sheet 1,6, spring-hard)none on the drawing
05 04 05-2Gangtrieb (escape pinion, z 10, m 0.263, and arbor)„Präz. Rundstahl 3,4 Din 175“ (precision round steel ∅3,4)none on the drawing
05 04 01-1Gangfeder (spring detent, drawing 1436)„Stahl“none on the drawing
05 04 01-2Goldfeder (gold passing spring)„Goldblech 333 federhart“ (gold sheet, 333 fine (8 ct), spring-hard)surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished)
05 06 04-2Schneckenkörper (fusee body)„Rundmessing 36 Din 1756(?) Ms 58“ ("Rundmessing 36", "Ms 58"; the DIN number reads 1756 or 1758)surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,05
05 06 04-1Schneckenwelle (fusee arbor)„Präz. Rundstahl 7,5 Din 175“ (precision round steel 7,5 dia, DIN 175)untoleranced sizes ±0,1
05 06 04-3Schneckensperrrad (click wheel)„Ms Blech 2mm DIN 17.51 Ms 58 federhart“ (brass sheet 2 mm, DIN 1751, Ms 58 spring-hard)surface ∇∇ ( (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished)
05 06 00-4Deckplatte (cover plate)„Bandstahl 24×1 Din 1016“ (strip steel 24 × 1, DIN 1016)surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished)
05 06 00-2Vorreiber„Präz. Rundstahl 14 Din 175“ (precision round steel bar 14 dia, DIN 175; "Rundstahl" cut off at the box edge, likely)surface ∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished)
05 06 00-1Staubplatte (dust plate)„Rundmessing 8 Din 1758 Ms58“ (round brass bar 8 dia, DIN 1758, Ms 58)none on the drawing
05 06 00-3Schneckenrad (fusee wheel)„Messingblech 5,5 Din 1751 Ms 58 federhart“ (brass sheet 5,5 thick, DIN 1751, Ms 58, spring-hard)surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished)
05 06 02-1Gegensperrrad (maintaining ratchet)„Messingblech 3,5 Din 1751 Ms 58 federhart“ (brass sheet 3,5, DIN 1751, Ms 58 spring-hard)surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished)
05 06 02-3Sperrkegel (click)„Bandstahl 11×2 Din 1016“ (strip steel 11 × 2, DIN 1016)surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,05
05 06 02-2Sperrfeder (click spring)„Bronzeband 1,5×0,4 federhart“ (bronze strip 1,5 × 0,4, spring-hard)surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
05 06 01-1Gegenfeder (maintaining spring)„Bandstahl 35×1,75 Din 1016“ (strip steel 35 × 1,75, DIN 1016)„400° anlassen“; „härten in Öl“; „310° blau anlassen“; surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,05
05 05 05-1Gegensperrkegel (maintaining detent)„Bandstahl 2,25 × 18 DIN 1016“ (steel strip 2.25 × 18)„härten in Öl, 310° hell-blau anlassen“ (harden in oil, temper light blue at 310°); „um die Bohrung 400° anlassen“ (temper round the hole at 400°); surface ▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
05 07 01-1Federhaustrommel (barrel drum)„Ms. 63 3“ (brass Ms 63, 3 mm, i.e. 3 mm sheet)surface ∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
05 07 01-2Federhausdeckel (barrel lid)„Ms 58 DIN 1751“surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished)
05 07 02-1Federkern (barrel arbor)„Präz. Rundstahl 14 Din 175“ (precision round steel 14 dia, DIN 175)„härten bei 800(?)“ (harden at 800?); „anlassen 270“ (temper at 270); „Kegel 1:70 poliert“; „härten bei 800 / anlassen 270“; surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); blank (Rohmaß) 14 × 45; untoleranced sizes ±0,05
05 07 00-3Sperrad (set-up ratchet)„Messingblech 3 Din 1751 Ms58“ (brass sheet 3, DIN 1751, Ms 58)surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished)
05 07 04Kette (fusee chain)„Bandstahl 5×0,3; 10×0,9; Präz. Rundstahl 0,6“ (strip steel 5 × 0,3 [the link plates, inferred]; 10 × 0,9 [the end hooks, inferred]; precision round steel 0,6 [the rivets, inferred])„leicht polieren“ (polish lightly); „∇∇∇ härten, 310° anlassen“; untoleranced sizes ±0,05
05 07 03-2Bandhaken rund (band's round hook)„Bandstahl 10×0,8 Din 1016“ (strip steel 10 × 0,8, DIN 1016)surface ∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,05
05 07 03-3Bandhaken f. Trommel (band's drum hook)„Bandstahl 10×0,8 Din 1016“ (strip steel 10 × 0,8, DIN 1016)surface ∇∇∇ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,05
05 02 02-1Wechselrad (change wheel)„Messingblech 1,2 Din 1751 Ms 58 federhart“surface ▽▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); blank (Rohmaß) 1,2 × 22,5 × 22,5; untoleranced sizes ±0,1
05 02 03-1Stundenrad (hour wheel)„Messingblech 1,2 Din 1751 Ms 58 federhart“surface ▽▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
05 02 04-1Auf u. Abrad (up-and-down wheel)„Messingblech 1,2 Din 1751 Ms 58 federhart“surface ▽▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,03
05 02 00-1Minutenzeiger (minute hand)„Goldin“ (a gold-coloured brass alloy)„Oberfläche poliert“ (surface polished); surface ▽▽▽▽ (▽▽▽) (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
05 02 00-2Stundenzeiger (hour hand)„Goldin“„Oberfläche: hochglanz poliert“ (surface high-gloss polished); surface ▽▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
05 02 00-3Auf u. Abzeiger (up-and-down hand)„Bandstahl 0,6 × 15“surface ▽▽▽ (▽▽▽) (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
05 02 01-1Sekundenzeiger (seconds hand)„Bandstahl 30 × 0,6 Din 1016“surface ▽▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
05 02 00-4Zifferblatt (dial)„Messingblech 1,3 Din 1751 Ms 58“ (1,3 struck, pencil "1,2")„Eisenblech 1,3 gebondert“ (iron sheet 1.3, bonderized); „weiß lackiert“ (struck through: white lacquered); blank (Rohmaß) 1,3 × 110 × 110
R-05 01 00-1Gehäusetopf (bowl)„Ms-Sandguß (Kokillenguß)“none on the drawing
05 01 00-2Verschlußplatte (closing plate)„Ms.“ (brass; sure)surface ▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished); untoleranced sizes ±0,1
R-05 01 01-2Glasrand (bezel)„Ms.“none on the drawing
00 01 01-9Firmenschild (maker’s plate), Type 3371„Alu 10“none on the drawing
P. 51 00 02-2Zunge (latch tongue)„Eisen 10 × 4 blank gezogen“„gebondert, schwarz lackiert“
127-556.01 (52 00 16)Überkasten (outer box): body and base plate, bleached birch„gebleichte Birke“none on the drawing
05 01 02-2Schlüsselmittelteil (key body)„St. Az / Din 668“ ("St. Az" = bright-drawn steel ("Automatenstahl"/"Stahl, gezogen", inferred) to DIN 668)„** (sure) = sand-blast, phosphate, spray black. - **Number note:** on the key assembly 51 00 11 this part is “; „sandeln, phosphatieren, schwarz spritzen“ (sand-blast, phosphate, spray black)
05 01 02-3Schlüsselrohr (key tube)„Ms. 58“ (brass, CuZn40Pb, inferred)surface ▽▽▽ (DIN 140: ▽ roughed, ▽▽ smoothed, ▽▽▽ fine, ▽▽▽▽ polished)

The 1950s build's deviations on the cards

The book's last chapter: the bench's 15 tests (tools/wempe_bench.js, the 1950s build), each on its part's sheet too. Not in the book yet: an order of assembly (the Hamilton's is its manual's Sec. VIII reassembly; no Wempe or 6MX assembly instructions are in the sources).

Sections 11, 12 and 17 ask whether a Wempe passes its trial, can be made to its tolerances, and keeps about two miles a month in the builds. This section follows the whole chain a navigator depends on, from a chronometer as made to the ship's position. tools/wempe_navigation.js takes 2,000 made chronometers for each version, as wempe_montecarlo.js makes them after the bench. Their swing sets how much the oil's thickening moves the rate, and they are adjusted as each version is: as last revised, the masses within ±0.45° of their compensating place (the TU's trial, the 6MX passports' spread of C); the two builds within ±0.1° on further rounds of the trial. Each one is rated over ten days in port and carried sixty days to sea on wempe_errorbudget.js's voyage: a winter port (55 °F) into the tropics (85 °F) over the first twenty days, the scatter and drift calibrated on the 6MX's test cards, the barometer, the oil, and heavy weather one day in five. The navigator carries the rated rate, as rating.py's backtest does, so the error at sea is what the rate has wandered from it, added up.

Four seconds of time are a minute of longitude, which is one nautical mile at the equator and cos(latitude) miles elsewhere. The route is estimated, since no source gives one: from 45° N in port to 15° N by day 20, where the box reaches 85 °F. The fix itself has an error too. Bowditch gives a celestial fix by an experienced observer in favourable conditions a probable error of about ±2 miles (The American Practical Navigator, Pub. No. 9, 2019, Vol. 2, Ch. 3, "Navigational Errors"). It is taken here as a circular normal whose median radius is 2 miles (σ 1.70 a side, 95 % within 4.16), and the clock's error is added along the parallel. The criterion was stated before the results were known: after 30 days without a time signal, the chronometer should add no more than the sight itself. Its longitude error at the equator must be under 2 miles in the median (Bowditch's probable error, and the user's aim of about two miles a month) and under 4.16 miles in 19 voyages of 20 (the fix's own 95 % radius). Each figure is itself a sample of 2,000 voyages, so each is judged with its sampling error: 1,000 bootstrap resamples of the voyages (seeded, the same each run) give each figure's 95 % interval, and a half fails only if that whole interval lies over its limit. A figure over its limit whose interval still reaches it passes within sampling: 2,000 voyages can't tell it from one on the limit. One whose interval misses the limit by under 1 % of it is on the line, reported and not failed (the owner's choice, 26 October 2026, when the 1950s build's 95 % figure came to 4.34 miles with its interval from 4.160, against 4.158, once the error budget's calibration was made to stop wandering with its random draws: before, a run could land it either side). The limits were not moved. The 1950s build must not fail, and the tool fails if it does.

Figure 23. The longitude error the navigator carries, at the equator, day by day at sea after ten days' rating in port: the median (solid) and the error 19 voyages in 20 stay within (dashed, the same colour), for each version; the two builds' lines lie on each other (they differ in the drive and the escapement's swing, which the rating takes out, not in what moves the rate at sea); the grey dotted lines are the fix's own probable error (2 miles) and its 95 % radius.
Figure 24. How the 2,000 made chronometers of each version spread after 30 days at sea: the share of voyages in each quarter mile of longitude error at the equator, beside the fix's 2-mile probable error.

22. An independent pass: can it be built, and would it navigate?

The per-part audit of 25-26 October 2026 (six batches, WEMPE-PROVENANCE.md) changed parts the analyses above rest on: the escape pinion (4.0 long), the detent (its spring tapered, its heights), the endshake (0.06, the 1958 book's 0.05-0.07), the fusee body's place, the barrel's internals, the maintaining clicks and springs, the case and gimbals, and the mainspring's set (×0.75). This section is not a rerun. It reads each analysis again, adversarially, and asks three questions. Does it read the changed parts from their one source, or from a copy of their old numbers? Do its criteria match their sources? What does it assume that a maker couldn't meet? Two read-throughs were made, one hunting stale numbers and one weighing criteria and assumptions, and every finding was checked against the code before anything was changed.

Stale copies, found and fixed

Everything else reads its source: the energy budget's level (0.913) is computed live from the re-solved spring; the gimbals' bodies come from gimbals_wempe.json, reweighed after the threads (periods 0.67 and 0.77 s); the bench's endshake stacks agree with the escape arbor's and the fusee's drawn lengths; the chain's lever on the stop and in the strength checks is the fusee's own table's.

The criteria against their sources

CriterionAs the tools encode itThe sourceAgreement
The TU's trial36/20/4/20/36 °C, the first 2 days of each period not counted; E 0.35, R 2.00, D 2.30, C 0.10, S 1.20, ω 3.50, N 30 (wempe_errorbudget.js)TU 25-07-1533-84, Table 5 (6MX-figures.md): GOST 8916-58's class II trial, GOST 8916-70's class 2 limitsevery limit and formula matches
Endshakefusee 0.08-0.12, escape 0.05-0.07, balance 0.04-0.06 (wempe_bench.js)the 1958 repair book, Appendix 4matches
The swinga guard band over the whole wind, 147-270° for the drawn versions and 150-270° for the builds (wempe_build.js, wempe_tolerances.js, wempe_montecarlo.js; 180-270° until 11 November 2026)225-270° (the 1954 instruction)relaxed, and the tools say so: the drawn version (155-166°) and both builds (the 1950s build 197-209°) swing under the sources' 225°
The fix's errora 2-mile median radius (σ 1.70 a side), its 95 % radius 4.16; the clock's error along the parallel against that radius (wempe_navigation.js)Bowditch: a fix "seldom in error by more than 2 miles"the arithmetic is right, the reading the lenient one: "seldom" reads more like a 95 % bound, and the clock's error is along one axis. Both stricter readings are reported below; the verdict stays on this one (the owner's choice, 26 October 2026)

Can it be made and put together?

Yes, as drawn, with the work a 1950s shop would expect, and on nothing a maker can't do. Section 19's checks run on the page's own solids, which since 26 October carry their threads (ISO metric, the tapped holes threaded and each screw on its hole's helix). dfm.py finds no void sealed inside a piece. What it flags is ordinary watchmaking: walls under 0.15 mm (the escape wheel's and fusee's teeth and groove lands, the cannon pinion), two slots narrower than a 0.4 mm end mill (the upper cock's and the fusee's: a slitting saw), undercuts on eleven milled pieces (a side cutter) and the sharp inside corners an end mill leaves rounded. It also reads eight holes of Ø 0.077 in the fusee body, which no sheet has: a reading of the groove's surface, to be confirmed. assembly.py takes the movement apart and back in the 1958 book's order with nothing passing through anything. The fits the sheets left impossible were found and corrected by the audit itself (README, Readings corrected; the collet's bore against the 1946 staff, the hour wheel's collar against the dial). One fit a maker must make on the bench: the endshake. Made to the sheets' fits, every one of the Monte Carlo's 2,000 movements has some arbor outside the 1958 book's band (about five arbors each out of band, some of them binding), and the bench sets each.

Would it keep time to navigate by? The chain and its weakest link

The chain is the made chronometer's swing (wempe_montecarlo.js), its trial and voyage (wempe_errorbudget.js) and the position (wempe_navigation.js). What each source adds alone after 30 days at sea, median |error|:

As adjusted on the TU's trial, the masses set to the 6MX passports' ±0.45°, the temperature's residual dominates: the rate moves by up to 0.1 s a day for each °C, and the voyage climbs 16.7 °C in its first twenty days. With the masses set to ±0.1° (two or three more rounds of the 4 °C and 36 °C periods, about two weeks on the bench), drift becomes the largest term. That is a slow change of rate, calibrated on eight test cards as a constant slope, which a ten-day rating in port cannot see.

Why the drawn version fails. It is mostly not the low swing. It is the masses' setting: run the drawn escapement with the builds' ±0.1°, and it carries . The swing (161°, from the drive and damping the owner chose) is left only the difference between that and the builds' figures. Where the swing does matter is over a year rather than a month: with the drawn hairspring the drawn version's rate moves about 2 s a day for each 10° of swing, where the builds' hairspring is set against the escapement's gain, so as the oil thickens between services it wanders more, and each voyage's fresh rating has more to take out.

What a maker must do on the bench

Pass the TU's trial; then keep repeating its 4 °C and 36 °C periods, five counted days each, moving all four masses by C/0.228 degrees each round (a pair alike, to keep the poise), until |C| is under 0.023 s a day per °C. That step decides whether a made one navigates. Then set every arbor's endshake to the 1958 book's band, the banking to the middle of the lock's window and the preload to its aim with wire riders, replace a detent spring that rings outside its band, change the strip if the mainspring's pull is out of its band, and turn the set-up until the swing stays in band over the whole wind (section 17's tests).

What the model assumes the maker achieves, and doesn't model. The ±0.1° assumes that a compensation, once set, holds: that moving the masses and tightening them again doesn't shift the poise, and that the nickel-steel rim doesn't age between the trial and the voyage. The scatter is white noise, and the drift a constant slope. The oil loses 0-10 % of the drive a year, linearly, its viscosity's change with temperature left out (it matters most to the drawn version). Missed or irregular winding is not modelled.

The verdict, version by version

23. What stays open

What the sources haven't settled, as the model's README lists it ( questions), each with the reading the model takes and what would settle it:

What the tools flag

Every ~~ the checks above print, together: what they report without failing on.

Reproduce

From marine-chronometer-source/wempe/verification/:

node reportdata.js        # every Wempe check run (the full runs, at once), and the figures: assets/data.json (about 9 minutes)
python geometry.py        # the browser checks on the Wempe's page, one after another: assets/geometry.json (about 11 minutes; section 19's fea, assembly and dfm about 16 more, each also writing its own JSON in assets/)
python build_report.py    # this page, index.html (a second)

Each tool can be run on its own from marine-chronometer-source/chronometer-working-model/tools/ (the commands are beside each tool's checks above). The model's ci.py runs the quick forms of all of them on every change, and the full runs with --full, but three: the readings off the videos (wempe_swing.py, acoustic.py --model wempe), which need the videos, and the start-up fit made on them (wempe_startup.js), whose result wempe_escapement.js checks the model's damping against. This page is rebuilt when their results change.

Scripts

ScriptWhat it does
solids.py, fine.py, audit.py, exploded.py, bom.pyThe geometry checks, the Wempe's tables in wempe/checks.py, bom.json and audit-expected.json (section 1)
wempe_escapement.jsThe escapement as makeEsc solves it, the drive, the damping, the swing and rate over the wind (3.1)
dynamics.jsThe escapement in time (3.2)
tripping.js, wempe_tripping.jsThe detent as a frame of beams (3.3)
wempe_mainspring.pyThe mainspring solved and fitted to 05 06_20; its coil friction (4)
wempe_physics.jsThe energy budget (5)
gearing.js, wempe_train.jsThe meshes tooth by tooth (6)
wempe_strength.jsThe parts' strength (7)
wempe_hairspring.jsThe hairspring (8)
wempe_thermal.jsTemperature (9)
wempe_thermal_frame.jsTemperature (9)
massprops.py, gimbals.js, wempe_gimbals.jsThe case and gimbals (10)
wempe_errorbudget.jsThe TU's trial and the voyage (11)
wempe_navigation.jsMade chronometers carried to sea, their longitude error and the fix's (21)
wempe_tolerances.js, wempe_bench.js, wempe_montecarlo.jsThe tolerances, the bench's tests, the movement made to tolerances (12)
wempe_transmission.jsThe whole transmission in time, every part that moves (13)
keystone.jsThe escapement against the 1904 text (14)
wempe_swing.py, wempe_startup.jsThe swing read off a video and fitted (15)
wempe_wear.js, wempe_overwind.js, wempe_motionwork.js, wempe_estimates.jsThe mechanisms in turn: wear, the key forced, the motion work and its studs, the estimates swept (18)
fea.py, assembly.py + assembly-check.js, dfm.pyThe frame under load, the order of assembly, each part on the machine (19)
wempe_makers.py, book.pyThe maker's sheets' data, and the build book checked as paper (20)
reportdata.js, geometry.py, build_report.pyThis page's data and the page

Sources

The model's sources, as its README gives them:

And for the methods:

  1. Хронометры морские 6МХ. Технические условия, TU 25-07.1533-84 (1st Moscow Watch Factory): the trial and its limits; Instruction 246-54 for checking standard chronometers (1954): the swing for chronometers in general, the banking, the start.
  2. V. A. Shpolyanskii and A. M. Kuritskii, Spuskovye regulyatory priborov vremeni (Moscow 1963): marine chronometers' balances, the 6MX's integral balance and banking screw.
  3. E. Phillips, "Mémoire sur le spiral réglant des chronomètres et des montres", Annales des Mines (1861): the terminal curves' conditions.
  4. S. Timoshenko, "Analysis of bi-metal thermostats", Journal of the Optical Society of America 11 (1925): the bimetal's curl.
  5. National Bureau of Standards, Circular 58, Invar and related nickel steels (2nd ed., 1923), Tables 5 and 13: the nickel steels' expansion by their nickel, the secondary error's quadratic.
  6. C.-E. Guillaume, Nobel lecture (1920), and VintageWatchstraps, "Temperature Compensation by Nickel Steels": the nickel steels and the integral balance's secondary error at the Neuchâtel trials.
  7. M. A. Crisfield, Non-linear Finite Element Analysis of Solids and Structures, vol. 1 (1991), ch. 7: the co-rotational beam element; J. S. Przemieniecki, Theory of Matrix Structural Analysis (1968): the frame element and its consistent mass; P. Wriggers, Computational Contact Mechanics (2006): penalty contact.
  8. NORDFORSK, Assessment of Ship Performance in a Seaway (1987): the operability limit; MARIN (IMDC 2024) and Jachowski et al., KONES (2012): merchant ships' roll periods.
  9. J. F. Archard, "Contact and rubbing of flat surfaces", Journal of Applied Physics 24 (1953); E. Rabinowicz, Friction and Wear of Materials (2nd ed., 1995): the wear law and its coefficients. A. Imrhan and C. Munoz-Vasa, Industrial and Systems Engineering Review 1(2) (2013), and Swain et al. (1970): a hand's torque on small handles. EN 10270-1: patented cold-drawn spring wire's strength.
  10. Mainspring friction and losses: ICMFF9 (gruppofrattura.it), "the torque during unwinding is 10 to 20 per cent less than that during winding"; lubricated steel on steel, the CRC Handbook as cited in US 3,939,081.
  11. The Hamilton's reports, whose tools these are: Would a chronometer made to the model run and keep time? and Does the model behave like a real Hamilton Model 21?

The scripts' outputs that feed this page are embedded in it, so it opens from the file. Built by build_report.py; do not edit index.html by hand.