Less precise machines building more precise machines

Hand fit/scraping

BB,

Some of the most precise surfaces in the world, mirrors for astronomical telescopes, can be accurately shaped to on sixteenth a wavelength of light or approximately 1/1,000,000 of an inch. The work is done by hand grinding with simple equipment.

Sal (A lurker here)

Simple: machines don't make machines. You can speed up the process of building a lathe or another mill by milling the surfaces, but if you want a certain precision, that will only go so far. You have to manually reduce the errors, and the rule-of-three method does this quite nicely. What's done is you use the rule-of-three to lap three surface plates flat, then compare these surfaces to the surfaces that need to be flat and remove material from just that surface (aka scraping).

Tim

-- "I've got more trophies than Wayne Gretsky and the Pope combined!" - Homer Simpson Website @

What about curved things like shafts? Is there a similar "by hand" way to achieve good regular curves?

On Mon, 13 Sep 2004 23:23:15 -0400, "Sal D'Ambra" calmly ranted:

Or successfully polished by machinery for several years to obtain a mirror finish at the wrong curvature and focal length, with which the Hubble was originally fit. The retrofit was a set of bifocals. ;)

- - - - - - - - - - - - - - - - - WARNING! WARNING! Dangerous Mailbox Approaching. Evade! Evade! ----- ----- ----- ----- ----- -----

True indeed. It's amazing the degree of precision that can be obtained with very simple tools and techniques ... applied properly. You can do precise work with a saw, chisel, file, and abrasives ... it just takes knowledge, time, and great care. Consider the surface plates and optical surfaces that were (and sometimes still are) constructed by mostly hand labor alone. As you correctly surmise, any machine is capable of doing work MORE precise than it's own level of precision. The difference is in the operator's knowledge and skill.

Dan Mitchell ============

I don't know how the first shaft was made, probably lapped or scraped to the bearings so it happens to turn round.

Once you have a shaft running parallel to the bed (easy to do with yet another bit of trickery), you can advance a carriage along the bed and turn straight shafts. If you just turn the end of the spindle, you can very easily improve the precision of the headstock right there.

The fun thing about turning things, though, is that if you spin one point in space around an axis, it traces a circle, which defines a plane. Viola, you have a perpendicular reference!

Tim

-- "I've got more trophies than Wayne Gretsky and the Pope combined!" - Homer Simpson Website @

I saw somewhere that the 200" Hale telescope mirror was finished by lightly rubbing the high spots with talcum powder on a thumb.

Leon

I wonder. If you scrape two plates flat, then weld them together at an angle to form a trough, filled it with an abrasive slurry, could you drop a rough shaft in there and rotate it, letting it rub against both sides, could you polish a shaft into straightness? I figure it'll rub down the high spots just because they'll hit the sides first. It could probably eliminate taper too since the bigger end would have a higher edge velocity rubbing the sides than the smaller end, and so would rub down more quickly.

Idunno. Axial variations won't decrease very quickly in any case, especially since the plates also wear in the process (depending). Now, if you stroke it back and forth as well as flip end-for-end, that should randomize it good. As for roundness, I doubt that will change much unless you mount it between centers, since with only two surfaces supporting, the bumps can push it wherever it wants.

Another possibility, if the plates are still pretty flat, the angle they form between the contact points and the axis of rotation could invite regular polygons - in fact I wonder if such a system could perfect a rounded octagon if you set the plates such that contact is at exactly 22.5 degrees around the axis.

Tim

-- "I've got more trophies than Wayne Gretsky and the Pope combined!" - Homer Simpson Website @

It will be as round as a twist-drilled hole. In other words it'll be vaguely round, but will be something of a rounded polygon, probably a rounded triangle. I forget the geometric name for the shape.

What would work better would be to place the shaft on a single plate. That would fix random error and would be nice and round. But eliminating systematic error is a puzzle to me at this time.

How about keeping the angled plates, but varying the angle occasionally to avoid a pattern forming?

Given enough time, can make what you can measure.

Mount the shaft between centers and file/sand it to a uniform diameter. This works even if the centers are not precisely aligned if you measure it at many points along its length. You can use outside calipers to compare the diameter at different points. If you are going to scrape the bearing to fit the shaft, you don't care so much what size it is as long as it is cylindrical.

Good Grief!!! Y'all should go to the library and get a book on the history of machine tools.

People made accurately round things long before they made accurately flat and straight. Take a part, put a dimple in each end. Support the part with points stuck in the dimples, and spin. Cutting on the outside of the spinning part defaults to round.

The magic trick to getting a straight-round shaft is to mate reference straight parts to the support points stuck in the aformentioned dimples. The device that does this is called a "lathe".

"History of Machine Tools" by Steeds is a good read.

Ah hah! I just found this: