Lord Kelvin and his analog computer | Hacker News Reader

Lord Kelvin and his analog computer

spectrum.ieee.org

142 pointssohkamyung17 comments

I feel embarrassed! Despite being from Glasgow, until now I've never made the connection between the River Kelvin (and the surrounding Kelvingrove and Kelvinbridge), the statue in the park of someone called Baron Kelvin, and the unit of measurement.

You scots are the most extreme combination of a people who have made vast parts of the modern world and people who know nothing about the achievement of their own people. I spent a year doing my postdoc in your country and have to say you guys are my favourite lot on this otherwise wretched rock.

Charles Petzold (author of The Annotated Turing, and Code - The Hidden Language of Computer Hardware and Software) is in the process of writing a book about this. You can see drafts here:

https://www.charlespetzold.com/blog/2016/12/Computer-of-the-...

https://www.charlespetzold.com/blog/2019/06/Computer-of-the-...

There is a very good Veritasium video that talks about analog computers and Lord Kelvin’s tide prediction machine:

https://youtu.be/IgF3OX8nT0w?feature=shared

There's another tide computer on display in the Science Museum London, which uses Cones moving on plates.

And the meccano integrator, from Manchester. And also Marc Brunel's block making machines, and possibly Henry Maudsleys screw cutting lathe (or maybe Whitworths) and his true-plates (always made in sets of 3 to avoid co-dependent bumps) as well as transatlantic cable, wheatstone bridge, you-name-it.

And of course Babbage-engine bits-and-bobs. It amazed me when the IETF had an open-night there, and were told to stick to the space display (ground floor) when it was geeks.. and upstairs were the ur-devices of .. computation and networking.

> There's another tide computer on display in the Science Museum London, which uses Cones moving on plates.

Here's how a mechanical harmonic synthesizer and analyzer work.[1] This is a very clear presentation of Michelson's design. It does forward and reverse Fourier transforms. Reverse (synthesis) is obvious enough, but the trick for analysis is clever.

The hardest part of the problem was the addition. Sine and cosine were easy. Multiplication by a constant was easy. But adding - that was hard. The Michaelson machine has an elegant solution. The positions to be added apply tension to springs, which pull on a common pull bar opposed by a single spring. So the addition happens in force space.

The USGS tide predicting machine #2 [2] added up the terms by running a long chain over many pulleys that can slide. The position of the end of the chain sums the pulley positions. This is a true positional sum, and is more accurate. But the mechanism is much bulkier and heavier.

[1] https://www.youtube.com/watch?v=GyYflzRVu6M

[2] https://en.wikipedia.org/wiki/Tide-Predicting_Machine_No._2

The Pacific Science Center in Seattle, WA, USA has a Tide Generating Machine[1] attached to their Puget Sound Model[2] - it calculates the tides, and then runs pumps/opens drains to make the physical model act like the predictions.

[1]: https://www.flickr.com/photos/ideonexus/5970013589/ [2]: https://pacificsciencecenter.org/exhibit/puget-sound-model/

I remember reading a book about analogue computation, and one fun fact stuck in my teenage brain - you can build a sort 'algorithm' that always takes only one operation. For every 'number' that you want to sort, cut a piece of uncooked spaghetti to a length that represents the number. Take the pieces of spaghetti and hold them loosely in your hand parallel to each other. Then bang one end down on the table, and they will all be at the same position on the table end. Read the resulting order of the numbers from the other end.

Someone could argue that cutting/labelling all the pieces of spaghetti takes considerably more time. I'm sure we are all capable of seeing past such a boring technicality.

The book described how big a pile of spaghetti would need to be to sort large numbers, and I'm afraid I can't remember any of the details, but there's a certain point where you need to bang the pile of spaghetti against something the size of the moon.

Arguably, he also invented the first ink jet printer ("siphon printer").

Here is a list of his inventions: https://collection.sciencemuseumgroup.org.uk/people/cp37484

My note on such things https://win-vector.com/2019/08/06/lord-kelvin-data-scientist...

If he didn’t get knighted, would absolut zero be measured in Thomsons?

His name was still Thomson after being knighted (in 1866), his ennoblement as Baron Kelvin in 1892 was a different award.

The genius that made the transatlantic telegraph cable possible!

That would be better credited to Heaviside, wouldn't it?

the genius that proved the world was exactly 20 million years old!

It's sad. If he had been willing to follow the physics and not try to defend religion against evolution (which was the real motivation for wanting to show the Earth was young), he'd have been able to argue mantle convection was very plausible (as this ruins the simplifying assumptions behind his argument), which would imply one should expect continents to be mobile. Continental drift could have been accepted much earlier than it was.

It's more complicated than that. Yes, Lord Kelvin believed in a young Earth (although one that was still millions if not billions years old, not one that was merely thousands of years old like some religious literalists), but it isn't entirely true that this was out of religious conviction. It's that he (and everyone else) didn't know about radioactivity. He modeled the maximum age of the sun assuming it was a combusting body in the conventional sense without understanding the nuclear fusion which was going on inside it. The people who predicted (correctly) that the Earth had to be billions of years old for the amount of evolution to have happened were of course correct, but they didn't know how the sun could work either.

Reply on news.ycombinator.com