BBC Article: The secret D-Day scientists and engineers

Saw an article on the BBC about Jack Darbyshire and Hugh Iorys Hughes’s work to predict the sea waves during D-Day. The article includes this image, showing Jack Darbyshire at work on a Commodore machine.

The article is: D-Day: Secret scientists and engineers behind Normandy landings - BBC News

Image caption: Jack Darbyshire’s work on waves is still used today. He went on to establish the physical oceanography department at Bangor University.

2 Likes

Very interesting… from the article

He came up with a way to model wave patterns using iron buoys in the sea, and measuring their magnetic movements.

His team even developed a computer, the first of its kind for rapid analysis of wave types.

They discovered they were able to predict wave patterns linked to weather events, even storms happening hundreds of miles away.

Over at the National Oceanography Centre “D-day – The starting point for modern UK oceanography” we find this image
image

Left: The drum wave analyser, right: Wave spectra from the drum analyser

Also on that page there’s a bit on tide modelling:

The 53-year-old Doodson was at that time the world’s leading authority on tide prediction. He used two tide-predicting machines: the Kelvin machine, built in 1872 but overhauled in 1942, and the Roberts-designed machine, built in 1906. The two machines were put in separate rooms at the observatory to minimize the chance of a bomb destroying both.

There’s a 1967 paper by Darbyshire:
Numerical prediction of wave spectra over the North Atlantic

A computer programme, written in Elliot Algol has been set up to predict wave frequency and directional spectra over a grid of 100 squares of 200 mile side over the North Atlantic. The frequency spectrum formulae used are those of Darbyshire [1959] and Pierson/Moskowitz [1964].

The programme has run successfully with both the Elliot 803 and Atlas computers and predictions were made for the high waves recorded during the last part of December 1959. The results using both frequency spectrum forms agree very well with the observed spectra

There’s a 1965 paper reviewing the 1959 analysis
Numerical prediction of ocean waves in the North Atlantic for December, 1959

The computational procedure followed here is essentially the same as that developed earlier, first graphically (Wilson [1955]), and then numerically for high-speed digital’ calculation (Wilson [1961]).

The complete program for the numerical prediction of waves according to this system is given in Fortran language in Appendix A

A CDC No. 1604 high speed digital computer was used to accomplish the computations to the general program of Appendix A. All three wind-fields required an overall input and computing period of 6.33 hours of machine time. Of this about six hours was required for the actual computation which meant that the average time taken for a single wave propagation calculation from a space-time lattice point to the ship was about 1¾ seconds. This is amazingly fast considering the complex calculations that have to be made.

The card output from the No. 1604 computer was recorded on an input tape by an IBM No. 1401 machine. Some 16,000 data cards were processed in this way in only 20 minutes.

In the final phase of the program it was arranged that both the significant wave height and the period of the waves, from each single wave propagation path reaching the ship, be plotted against their times by an automatic machine process. The input tape from the IBM 1401 was therefore fed to an IBM No. 7090 computer and the data transferred to a plot tape.

The time required for this was 30 minutes. Finally an SC No. 4020 plotter reproduced the information from the plot tape to visible form in approximately 5½ minutes. The overall machine-time for the complete calculation of three wind-fields, embodying some 17,200 gridpoints, was thus about 7.30 hours.

The times quoted above are not inclusive of the key-punch operation of transferring tabulated wind velocity data to punched cards. About 35 to 40 hours of key punch and verification time were involved in assembling some 1500 input data cards.

In Darbyshire’s 1952 paper “The generation of waves by wind” he mentions Fourier analysis:


which predates computers and the FFT (and would possibly have been done mechanically, if the wartime analyser was classified.)

Bill Hammack has a very good short series of short videos on a particular mechanical analyser - Michelson’s.

I recall seeing an analyser in the Science Museum - turns out it was in 2012

Google Photos

3 Likes