No single ceremony retired the tide-predicting machines. No harbour held a funeral. Between about 1960 and 1968, one office after another simply stopped booking crank time — because a computer down the hall could do the same work faster, cheaper, and better. Here is how brass lost.
What the computer offered
A program implementing the same harmonic formula held unlimited constituents — no friction penalty per wave, no bronze to cast. Revising one port’s constants meant repunching cards, not a team-day with a spanner (compare the 1910 setup ritual). A year’s hourly heights ran in minutes, and the output printed as numbers directly — no clerk squinting at ink curves (see Reading a Prediction Roll for what was eliminated).
The handover, office by office
America moved first: the Coast Survey’s digital programs took over routine production in the early 1960s, and Machine No. 2 fell silent after half a century. Britain followed — Bidston’s workload migrated to university and government computers through the mid-1960s. Hamburg’s German machines held on longest in routine service, with the last machine-drawn predictions dating to about 1968. The full arc sits in my history pillar.
What was lost — and kept
Lost: the tactile check — a wrong setup showed instantly as a nonsense curve, while a wrong punch card failed silently. Lost, too, the operators’ embodied knowledge of each machine’s personality. Kept: everything mathematical. Doodson’s constituents, arguments, and methods moved into software unchanged; your phone’s tide app computes his arithmetic. The machines died; the mathematics emigrated.
The shutdown ledger, 1960–1968
The handover happened office by office, never announced. My working reconstruction from agency histories and the history pillar:
| Years | Office | What happened |
|---|---|---|
| 1960–62 | US Coast Survey | Digital programs take routine production; No. 2 runs its last tables |
| 1963–65 | Bidston / UK offices | Workload migrates to university and government computers; predictors kept oiled “just in case” |
| 1965–67 | Hamburg | German routine service winds down; motor-driven runs replaced by programs |
| ~1968 | Last offices | Final machine-drawn routine predictions; brass era ends |
Note the overlap: for several years machines and programs ran in parallel, the pen checked against the printout, neither quite trusting the other. Operators from those crossover years describe the strange experience of watching a computer reproduce in minutes what took their team a week — and finding it agreed to the centimetre. That agreement was the machines’ finest compliment and their death warrant in one.
The awkward questions
Could brass have survived as backup? Some offices kept machines oiled “just in case” into the 1970s, but without trained operators and fresh constants they were museum pieces in waiting. Preservation, not production, is their future — see where survivors stand.
Were the computers immediately more accurate? Roughly equal at first, then better: 100+ constituents, longer analyses, and eventually real-time surge modelling no brass could touch. Accuracy was never the drama; economics was.
Is any machine still used for real prediction? No. Every working demonstration today runs for education and delight — which is reason enough to visit one.
An epitaph, not an obituary
Obituaries close books; epitaphs invite visitors. The machines earned theirs: ninety years without a wreck laid at their gears, mathematics that still runs in your pocket, and a handful of survivors that turn for anyone who asks nicely. Go see one — my museums guide tells you where — and when the demonstrator lets you take the crank, remember the operators who did it for a living. The vocabulary on the labels is decoded in the glossary.