The largest tide machine ever put in brass stood in Washington: the US Coast and Geodetic Survey’s Machine No. 2, completed in 1912. Thirty-seven constituents, tons of bronze and steel, and the instrument that made America independent in tide-table production. I have spent months tracing its drawings and operator notes. Here is what made it extraordinary — and what it cost.
Why America needed its own giant
Before No. 2, American tide tables depended on European machines and hand computation. Rollin A. Harris, the Survey’s tidal expert, argued that American coasts — with their wild variety, from Fundy’s resonance to the Gulf’s diurnal tides — deserved a machine sized for the hardest ports. E. G. Fischer turned Harris’s specification into metal. The result dwarfed everything Kelvin built: where the 1872 prototype carried ten cranks, No. 2 carried nearly four times as many.
The 37, bank by bank
The constituents were grouped the way analysts think (meet them all in Why 37?):
- Semidiurnal core (M2, S2, N2, K2 and kin). The backbone shafts, with the biggest cranks and the finest gear trains. M2’s 12.42-hour period demanded the most exact ratios on the machine.
- Diurnal family (K1, O1, P1, Q1). Essential for Pacific and Gulf ports where one tide a day dominates. Their slow shafts turned lazily — one revolution per simulated day — while the semidiurnals spun around them.
- Long-period terms (Mf, Mm, Ssa). Small cranks that raised and lowered the whole baseline over fortnights and seasons. Individually tiny, collectively the difference between a good and a great prediction.
- Shallow-water overtides (M4, MS4, M6 and compounds). The fast, small, high-speed pinions — the ones I tell visitors to look for first in my museums guide. These captured what estuaries do to the astronomical tide.
Thirty-seven was not numerology: as I explain in Why 37?, it was where signal sank below the machine’s own friction.
Engineering at the limit
Everything about No. 2 fought friction. Thirty-seven crank wires met in an enormous pulley cascade; every extra pulley stole motion. Fischer answered with larger pulleys on jewel-like bearings, a stiff cast-iron frame that did not flex as the room warmed, and a motor drive steadier than any hand crank. Setting up all 37 pins for a new port took a full team-day — the logbooks show operators initialling each other’s settings.
Half a century of service
No. 2 predicted American tides through two world wars — the wartime table surge I describe in the history pillar ran largely through machines like this one — and stayed in routine service into the 1960s, when digital computers finally relieved it. See Why the Machines Died for that handover.
Asked in the gallery
Where is Machine No. 2 now? In American museum and agency custody; display status rotates. Check my verified museums guide before planning a visit.
Was 37 really the world record? For moving constituent cranks summed on one machine, yes. Later British and German machines rivalled it in mass and refinement but not in count — see the Bidston machines and Germany’s competitors.
How accurate was it? At well-analysed ports, high-water predictions routinely landed within centimetres — comparable to the printed tables navigators actually used. Storm surge was never in its remit; that distinction matters and I stress it in the beginner’s guide.
Why size mattered — and didn’t
No. 2’s thirty-seven cranks bought America independence and the world record, but reflect on this: Bidston matched its harbour accuracy with fewer, quieter components, and Hamburg chased silence over count. Bigness impressed congressmen; quietness served navigators. Hold both ideas at once and you understand the whole brass century better than most textbooks. Next, learn to decode the rolls this giant produced in Reading a Prediction Roll.