Work in progress. This model has been rescaled to Hamilton’s own factory drawings (it was about 12 % too small) and its parts are being recreated from them with the help of the NAWCC (with special thanks to Thomas Stocker); the parts no drawing has reached yet stay estimates.
Reset brings the camera back to the chosen view; ↶ and ↷, under the tools, go back and forward through the places it has been. Parts too small to tap are listed under Parts. To see how the chronometer works, start the walkthrough under How it works.
Eight steps from the box to the balance, with live diagrams tied to the running model.
The manual’s own order: 58 operations to take the movement down (Sec. V) and 105 to put it back together (Sec. VIII), each part coming off or going on in its turn, with the tools, oils and checks.
Click a name to single it out and read about it; untick to hide it.
A detent chronometer does not start by itself. Once its balance swings too little to carry the discharge jewel past the trip spring and unlock the wheel, it stops, and it takes a quick twist of the box to set the balance going again (Sec. III).
Why the detent was kept to chronometers in gimbals: a sharp jolt can upset it. Against the balance's swing it can check the balance below what unlocks the wheel, and the chronometer sets (stops); with the swing it can carry the discharge jewel round to unlock a second time, and the wheel trips, escaping an extra tooth, so the hands jump half a second. An illustration of those faults, not how a sound Model 21 behaves on its gimbals.
The arm was fitted from 1947 so a chronometer could be shipped with its balance held; the screw keeps the mainspring's power in while the balance and escapement are out for service. With the screw down the balance swings on, unlocking a wheel that cannot turn, and runs down in a minute or two; raise the screw before then and it carries on. Screw it down or lock the arm, then unlock and twist: the hands will have lost the time it stood.
There is no regulator. The rate is set with the weights on the balance rim, each a nut on a screw: the two timing weights at the arm’s ends, or the small vernier weights inside the rim beside them for fine changes. Both of a pair are turned alike, so the balance stays in poise. Out moves their mass further from the staff, which raises the balance’s moment of inertia and slows it: the hands lose.
The manual gives about 40 s a day for a full turn of the timing weights and 2.8 s for the verniers; the model’s thread is set to match, and the moment of inertia comes from the model’s balance and the parts list’s masses. In an overhaul, balance screws and washers bring the rate within 2 minutes a day, then the timing weights within half a second (Ops. 5, 6); a pre-temperature test at 55 and 90 °F follows, and pairs of screws are moved from hole to hole to correct it (Op. 8, Table IV); the verniers make the final changes (Op. 10). The rim has 24 tapped holes, numbered 1 to 13 round each half from the arm’s end, as the adjuster’s numbered balance block numbers them (Fig. 99): the timing weight stands at 1, the vernier at 2, and the screws in 3 to 12, the screws of a pair in the same hole of each half. The standard set is the four pairs a 1941 Model 21 carries; the parts list allows a fifth. Moving a pair along the rim keeps the moment of inertia, so the rate at the mean temperature stays, while the rate in heat changes: the panel gives Table IV’s figure, linear in temperature, and the balance’s own curve on top of it: for the Model 21’s, the small gain in the middle that the factory test card of No. 3390 shows (Sec. IX); for the split bimetallic balance, its middle temperature error, illustrative, the rim’s curl drawn 20 times as far as it goes. The chart gives the Navy test’s three temperature compensation figures and their limits. The model’s clock runs at the new rate, so the hands drift from real time; try 3600×. Zero puts back the weights’ mid-travel, the standard screws and 72.5 °F. The escapement’s settings on the Adjuster’s bench change the rate too, and the panel says by how much; the bench’s own button takes them back. The weights and screws are drawn on the Hamilton balance only.
The navigator doesn't set the hands: the chronometer's error is compared with a time signal, daily, and recorded with its rate, and applied to every reading (Sec. III, Setting; Sec. IX, Table I). Error and daily rate in seconds: + fast or gaining, − slow or losing. Here the signal is the master time, which runs with the model; the hands step in half seconds, so the error is read to the nearest half second.
A debug view: where the parts touch, come close or pass into each other, found on the model's solids, shown through the parts made faint (the box and case too). Pick a part for its contacts alone; hide parts to leave their contacts out.
. A pair shows the worst it does: its overlap, else touching, else its least gap. Click a pair to show it alone; › picks the other part.
How far the balance swings is set by the train's torque: each swing the impulse gives it what unlocking, the detent's springs and its own friction take, and it settles where they balance, 1⅜ to 1½ turns of motion (Sec. II), 248° to 270° each way. No balance is quite isochronous: here the escapement, pushing it on either side of its dead point, makes a smaller swing lose (Airy; the hairspring taken as isochronous). The fusee keeps the torque, and so the swing and the rate, nearly the same from full wind to run down; while the key turns, the weaker sustaining spring drives, and the swing dips.
The isochronism check is the manual's (Sec. IX): wound, the gain over the next 12 hours against half the gain over 24; the Navy's tolerance is 0.50 s, and the factory test card of No. 3390 reads 0.00. The mainspring's pull is the solved spring's, from its measured free shape; the sustaining spring's strength is a design (a split ring weaker than the drive, banked in running, as the period's chronometers had it); the balance's free run-down (1/e in 67 s) lies inside what its parts give (tools/physics.js).
The Navy's test before a chronometer was issued (Sec. IX, Performance Testing): thirty days in six periods of five, at 90, 72½, 55, 55, 72½ and 90 °F, the dial error read daily on Hamilton's comparator to 0.01 s, judged against the Bureau of Ships' limits. Here it is run on the model as it stands (weights, screws, balance, adjuster's bench, mainspring), wound at each daily reading, beside the manual's factory card for chronometer No. 3390. Change something and it is run again. The model has no day-to-day scatter, so its rating and recovery figures come from the comparator's rounding alone.
The escapement's settings, as the repairman adjusts them (Sec. VIII, Ops. 84–88, 97). Move one and the detent, jewels and roller are rebuilt with it, and the figures are measured against the manual's the way tools/escapement.js measures them. The model runs with them: the balance swings as far as the impulse, less what unlocking takes, carries it (Stopping and starting), and the escapement's rate changes with where they push the balance about its dead point (Rate and timing weights). If a setting would stop the escapement working, or leave the balance swinging too little to keep it going, the bench says why and keeps the last one that works. The last slider is the hairspring's own isochronism, the seconds a day it gains for each 10° more swing: the model takes it as 0, but an adjuster can set the spring against the escapement, which loses at a smaller swing, so that the rate hardly follows the swing (Swing and isochronism; about −0.1).
Cut faces are hatched. The plane moves with the movement, so it works in every view, including in the box.
For a maker. Each part's card has its maker's sheet (its lines of the manual's parts list, the fits measured on the model, materials and heat treatment), its drawing to scale (SVG, mm) and its solid (STL). Here: measure between two points on the model, print the whole build book, or take the model's numbers away.