The Whitworth Three Plates Method (2017)
ericweinhoffer.com
ericweinhoffer.com
Simon Winchester wrote an enjoyable book about the history of precision that starts with Wilkinson’s boring of steam cylinders to improve upon the efficiency of the first steam engines and rifles, through engineering history to ASML’s nano-level IC fab machines.
In America, it's called The Perfectionists: How Precision Engineers Created the Modern World. Elsewhere, the title of the same book is Exactly.
https://www.amazon.com/Perfectionists-Precision-Engineers-Cr...
A complete dodge & waste of money.
That's cool! My father was an electronic engineer and one of his projects was quite a compact device (as big as 2 decks of cards) that was able, IIRC, to detect a slope of a few microns over a meter, and to take multiple measurements to make a 3d model of the surface. This was paid for by a company [1] that made big slabs of granite, which I reckon must have been these datums referenced in the article.
This was almost 20 years ago, and I was in my teenage years at the time. I had built the C# interface to download the data from the device, but I never understood or even though about the purpose of the whole endeavour until this article.
1: https://www.microplan-group.com/en/
EDIT: it might have been a C interface coupled with Gtk, not C#. In retrospect, neither was a very smart or stable choice 20 years ago.
One thing I'm curious about, is whether this is actually how granite surface plates are still made. Or have they switched to some kind of interferometry.
Edit to elaborate -- to make those ultraflat mirrors, you can also use an interferometric variant of the 3-plate method when, instead of granite or cast iron surface plates, you're making optical flats. Instead of looking at contact patterns with dye, you can observe the interference bands that show up between two optically flat, smooth surfaces to compare them, which is generally more sensitive. But it's still a comparison process. You still proceed by lapping the high spots and checking again...
If you build a flat machine like a surface grinder then that surface grinder can transfer its own flatness onto a surface plate.
Look up a 'sine bar' and gauge blocks for inspiration. :)
It is also interesting to note that functional surfaces are very simple, being composed of flats, rounds and inclined planes. This is why you can make nearly anything with a lathe (and a mill to spare your sanity).
You start with a surface plate and three "pretty good" squares. Then put the squares against each other, and look at the contact pattern, and so on...
Also, IIRC, Whitworth improved the method by introducing scraping to the process, which makes it much faster, but did not invent it; the same fundamental technique was already being done in Maudslay's workshop using lapping when he started working there.
Stacking them (even with rotations by mutlplies of pi/2) will not detect the case when every edge has the same kind of deviation from the straight line, right?
Putting them edge to edge will assure every edge is straight and the same length (within observation limits) but may get you a rhombus which is not a square.
Do you somehow iteratively combine both of the above?
The Michelson interferometer was incredibly sensitive for its day but required heroic efforts to use: it was deep in a subbasement, floating on a pool of mercury, and was routinely unusable during the day due to horse and carriage deliveries a few buildings over. My friend who was a physics major at Princeton said they built one in a day for their physics lab (these days, vibration isolation tables and alignable optics are far, far easier).
I have 3" and 6" cast iron bar stock cut offs that I picked up for this purpose, along with Prussian blue, and something that should work as a scraper... what I don't have is a brayer, but I'm told you can improvise with a rolled up and tapped shop cloth.
If you can find a flat piece of glass then use that as your starting point and then apply the 3 plates method.
It mostly depends on how far you have to go.
Yet several statements in the article make it clear that my understanding is not correct (e.g. "Next, the Green plate acts as the control and the Red plate is lapped against it," or "Next, the Green and Blue plates are lapped against each other in an alternating manner.")
To be clear, I haven't done either :-)
Foundations of Mechanical Accuracy is a great book if you're in to this sort of thing.
There are still people who hand scrape plates, but these days, there are a few flat grinding masters used to make nearly all plates, and those masters can traceability back to hand-scraped plates.
A good surface plate will get you 0.000001"-0.000005"
This method can take you to much more flat surfaces from first principles. Also there will be imperfections in the ice based surface itself too.
Because you could also just pour molten metal into a mold and let it cool to get an initial "flat surface" but there are limits to that.
I dont know what metal would work for that. There's still significant surface tension in an open pour. I was casting some copper a few weeks back. A ~2CM piece had about ~3mm of surface curvature to it. Plus plenty of distortion from unequal cooling where it varied thickness as well.
Differences in rate of cooling leads to uneven stresses and minute bends. Similarly while handling and working the surface plates you need to let them reach thermal equilibrium.
Large forgings which need to be machined to close tolerances are left to sit even for months in room temperature to reach thermal equilibrium.
It's usually molten tin, actually.