Ninety years, about forty machines, five countries, one slow obituary. The mechanical tide predictor lived roughly one long human lifetime — and like a life, it divides into acts: a brilliant childhood, a confident maturity, a rigorous middle age, and a sudden, almost silent death.
Act I: Proof (1867–1879)
William Thomson — not yet Lord Kelvin — told the British Association in 1867 that tides could be analysed into constituents and re-summed by machinery. The Légé firm built his ten-component prototype in 1872–73; it toured scientific meetings and reproduced past tides well enough to silence sceptics. I reconstruct its mechanism in How Kelvin’s First Predictor Actually Worked.
Success bred commissions. The India Office machine (1879) predicted for dozens of Indian ports from a single setting tradition, and the “second Kelvin design” became the template every later builder copied: crank, geared shafts, crank pins, wire-and-pulley adder, pen on roll. The anatomy is explained in my beginner’s guide.
Act II: The giants (1900–1920)
Rollin Harris of the US Coast and Geodetic Survey believed American tides deserved an American machine. The result, Machine No. 2 (completed 1912, designed with E. G. Fischer), carried 37 constituents — the most ever put in brass — and weighed tons. It made the United States independent in tide-table production and stayed in service for half a century. I dismantle it (on paper) in Inside Machine No. 2.
Meanwhile the Norwegian and Russian hydrographic offices commissioned their own predictors, and the principle spread wherever tide tables mattered: navies, colonial ports, and great estuarine harbours.
Act III: Rigour at Bidston (1920s–1950s)
Arthur Doodson and the Liverpool Observatory (later Bidston) rebuilt tidal science from the mathematics up: longer analyses, least-squares rigour, dozens of minor constituents, and the famous Doodson argument numbers. Their machines — quieter, heavier, more precise than Kelvin’s — embodied the new standards and predicted for hundreds of ports. Profile: The Bidston Observatory Machines.
Across the North Sea, Hamburg’s engineers built the handsome German Gezeitenrechenmaschinen, improved the wire runs, and kept their own analysis traditions alive — the least-known great machines, covered in Germany’s Forgotten Competitors.
Act IV: The quiet death (1960–1968)
Electronic digital computers killed the predictors with shocking speed. A program could hold more constituents than brass ever could, re-run instantly when constants were revised, and cost less per port-year. One by one the machines fell silent: America first, then Britain, then Germany. The last routine machine predictions date to about 1968. Full account: Why the Machines Died.
Where they rest now
Survivors stand in London’s Science Museum, the Deutsches Museum, American collections, and Bidston’s legacy displays — most static, a few still turned for visitors. My verified museums guide tells you exactly which ones you can see, and which still move.
A losses ledger: where did forty machines go?
Nobody has published a full census, so here is my working tally from catalogues, office reports, and correspondence — approximate, and I welcome corrections via the contact page:
| Built for | Rough total | Surviving | Scrapped or lost | Untraced |
|---|---|---|---|---|
| Britain | ~14 | 4 | 7 | 3 |
| United States | ~6 | 3 | 2 | 1 |
| Germany | ~8 | 3 | 3 | 2 |
| Others (Norway, Russia, Japan…) | ~10 | 2 | 4 | 4 |
Two patterns stand out. First, war destroyed more machines than obsolescence did — Hamburg’s losses in 1945 dwarf any peacetime scrapping. Second, survival was an accident of prestige: the biggest and the first were kept (No. 2, Kelvin’s descendants), while ordinary workhorses went to the furnace. If you can account for one of the “untraced,” you will have made a genuine contribution — start with my museums guide.
What the archives get asked
How many machines were built? About forty full predictors plus analysers and demonstration models. My glossary distinguishes predictors from the rarer tide analysers that worked the problem backwards.
Which was the biggest? America’s No. 2 by constituent count (37); some German and later British machines rivalled it in mass and cabinetry.
Did any predict for the D-Day landings? Allied tide prediction in 1944 used Doodson’s methods and Bidston expertise; the mechanical fleet contributed to the enormous table production of the war years. Wartime demand, in fact, kept several machines running around the clock.
Timeline at a glance
| Year | Milestone |
|---|---|
| 1867 | Kelvin proposes mechanical summation |
| 1872–73 | First predictor demonstrated |
| 1879 | India Office machine enters service |
| 1912 | US Machine No. 2 completed |
| 1921 | Doodson’s rigorous Liverpool analysis |
| 1920s–40s | Bidston and Hamburg machines peak |
| 1940s | Wartime table production at full stretch |
| 1960–68 | Digital computers take over; last runs |
My verdict on the ninety years
Three things stay with me after years in these archives. First, the machines were never really about brass — they were about trust: harbour masters betting ships on a wire sum, and winning, for decades. Second, the heroic narrative belongs as much to the craftsmen and the human computers as to Kelvin and Doodson. Third, nothing here is truly dead: the mathematics emigrated into software, and the surviving machines still turn for anyone who visits. For that mathematics, read Harmonic Analysis: Kelvin to Doodson; for the craft, try Setting Up a Prediction Run and Reading a Prediction Roll.