What's new
What's new

Current state of the art reference planes

rhb

Aluminum
Joined
Apr 27, 2019
Location
A small town in central Arkansas
Who makes arbitrarily large reference planes?

What level of accuracy can they achieve?

By what methods do they test the accuracy of such surfaces?

How is loading and temperature addressed?

I am looking for very specialized information. The sort of thing google can't find. I did every type of search I could think of today and got nowhere.

I'm a retired research scientist. I just obtained a personal copy of "Foundations of Mechanical Accuracy" by Wayne Moore. Reading that spawned my questions.

I read it years ago via an interlibrary loan, but could not find a copy at a reasonable price. I recently discovered that Moore Tool Co. keeps this masterpiece in print. For $150 I received a sewn binding, sixth printing done in 2017 on fine glossy paper. A refreshing change from glued binding, print on demand books for as much or more.

At the time I ordered it from Moore Tool, a pair of Amazon sellers were asking $4000 for a used copy!

I am interested in the technology of doing something very difficult.

I am, naturally, aware of L.S. Starrett and Doall. Does anyone go beyond them?

Reg
 
Last edited:
A surface plate is a small reference plane. By arbitrarily large I mean just that. If someone had an application that required a 10' x 12' surface which was extremely flat, how would such a thing be made? And what flatness is achievable?

"Foundations of Mechanical Accuracy" was written 50 years ago. I'm curious how much technology has changed since then. Moore describes a process which was developed by Joseph Whitworth in the 19th century involving the creation of 3 plates.

For an arbitrarily large flat plane making 3 is not practical. So how would it be done? What are the limits of modern technology?

Amateur astronomy has been transformed by the development of CNC grinding and polishing machines. Is there a similar development in the making of surface plates?

My understanding is that a laser interferometer based system is employed in modern production of surface plates. But I've never seen any details about such machines.

I came across a thread somewhere which asserted that Starrett had investigated buying Chinese plates and resurfacing them as a budget line but concluded they were not sufficiently reliably made to bother with. Unfortunately, I stumbled across the thread and don't know how to find it again.

I'm a metrology nut. I am fascinated by national lab level metrology.
 
The tolerances achievable for surface plate accuracy are generally based on size, driven to some extent by practical considerations of handling, as well as the more fundamental material properties that can have some effect on flatness, such as CTE and naturally-occurring variations in composition of the material. If you look at a grading scheme tolerance chart, plates rated AA (lab grade) that are in the 8' x 10' or 6' x 12' range show overall flatness tolerances in the neighborhood of 0.001", with more "normal" sizes, 18"x24", 24"x36", down closer to 0.0001.

For measuring the flatness, the standard practice is to use an autocollimator to track the angular deviation of a right-angle reference mirror as it travels across the surface. The federal standard governing surface plate specification gives a procedure for doing this, as I recall. The use of interferometric methods is likely to give higher resolution results, but in many cases that perceived accuracy increase is highly dependent on the integrity of the plate support, and a host of other variables that could affect results in the nanometer-scale measurements typical of interferometry. An electronic autocollimator is probably "state of the art" for production of large plates these days.

Generally, you can make a granite plate just about as flat as you want, but it requires several orders of magnitude increase in time and expense to achieve reduction of flatness tolerance variation by an order of magnitude.
 
For smaller parts optical flats would be pretty accurate. Surface plates look like mountain ranges compared to optical flats.
But they are pretty small and pricy: 305mm round optical flat is close to 10 grand. 1/20 lambda (0.03um) flatness, something like 100 times better than AA laboratory grade surface plate.
 
For big plates I would look at active controls, usually laser based. Similar to what's done on the large telescope (or multiple element) mirrors, where a series of sensors and actuators continually adjust surface positions to maintain a target shape/planarity regardless of orientation or load.

Done right, you can have multiple small plates with active mounts giving planarity within microns, regardless of load. Not saying it's cheap, but it's doable.
 
Thanks for the link to Renishaw. That should feed my metrology habit for a few days.

In stumbling around today I ran across mention (Tru-stone FAQ IIRC) that overhead lights could cause a temperature gradient which warped the surface of a large black AA grade plate. I'd expect that would require incandescent lighting and that LEDs would not do that. However, it does suggest that the limitations are driven by environmental rather than technical issues.

I was bothered when I bought my black 9" x 12" Chinese plate 20 years ago that there was no indication of the correct support points as per the US standard. In contemplating the subject this morning, I've come to question the 3 point support as it causes the plate to droop in a manner which is dependent upon where the loads are placed.

I've not done any mathematical analysis yet, but it seems that supporting the plate on a thick pad of very soft elastomer would avoid or at least minimize deformation due to point loading.

The mounting that comes to mind is a very stiff steel pan with a 1" cast elastomer pad so the plate is essentially floating on the elastomer pad. I need to get a 24" x 36" plate so I'll try that out when I do. I'll have to upgrade my metrology capability to check it though. I'm currently contemplating the design of an electronic level with fractional arc second accuracy based on this concept.

NasaAmes-SimpleTiltmeter.pdf | DocDroid

The angular resolution cited is 1/4 millionth of an inch per inch. Experience reported on the EEVblog forum suggests that with a 2"/div vial the resolution is much finer and far exceeds an LVDT design such as the Talyvan.

I'm a geoscientist. I started out in igneous petrology, so calling a gabbro granite is weird, though not all that surprising. The criterion for classifying an igneous rock as granite is >10% quartz. But there was no work when I got my MS so I moved into geophysics and towards the end of my career was concerned with things such as the deformation of the earth caused by a locomotive pulling into the train depot. The water level in a well was observed to rise and fall when trains passed and reported in 1892 by F.H. King of the University of Wisconsin to vary with the weight of the train.

So issues like the surface warping because of the overhead lights are surprising in that I had not considered that as a source of error, but otherwise very familiar.

Thanks to all.

Have Fun!
Reg
 
You can find really large plates pretty easy for the cost of hauling them. Optics are a good option. They can be as accurate as your wallet is deep.
 
I was looking at the deflection models in Machinery's Handbook. Support on a pad would reduce the deflections by a factor of 2-4. However, this morning it occurs to me that one might simply swap less mechanical deformation for more thermal deformation.

I have Watson's "A Treatise on the Theory of Bessel Functions", Morse and Feschback and many of the usual suspects. I'd not got that serious yet. I am delighted though to be referred to such as it speaks volumes about the intellectual level of the forum. I'm sitting in the middle of rural Arkansas, so finding anyone nearby who has any conception of mathematics at that level is pretty near impossible.

The concept I'm considering is using a pair of fractional second levels attached either to the plate or a 3 point support and another pair mounted on on a small plate with 3 point support and using a small MCU (e.g. Arduino, STM32F103 "blue pill") to do differential surface calculations and determine the flatness of the plate. The vials are so sensitive that just walking around the plate will produce a measurable tilt. Hence the need for differential measurement.

The question of the day is, will a Wein bridge oscillator reach 300 KHz with decent performance? And are 1n4148 diodes a suitable substitution for the diodes referenced in the NASA Tech Brief.

I use my Chinese plate and silicon carbide sandpaper to do lapping. By wetting the paper on the back it adheres very nicely to the plate. I was able, using worn 100 grit paper, to remove the hollow from a Ouachita stone in about 10-15 minutes. I really should have dressed it when I bought it as it had saw marks.

Thanks for the links to the earlier discussions. I skimmed through the metrology topic list for a while, but it's often hard to tell from the initial title where they go.

Have Fun!
Reg
 
A bog standard Hilger & Watts autocollimator will resolve 10 micro inches per foot of slope. That can be used, with the aid of simple 1/20 wave mirrors to lap a surface table to a similar level of flatness, if you have the patience and the temperature/radiation controlled environment to do it in...

A Wein bridge oscillator can operate from single Hz to MHz, but is tricky to keep linear. The gain control part of the circuit is quite important. An LC resonant circuit, or the equivalent of a crystal or ceramic resonator can be simpler to control, with a more stable output.
 
I think the idea of putting a surface plate on top if an elastomer on top of another flat plate is purely theoretical and outside the realm of practicality. At that point, the stiffness of the lower flat plate comes into play and you may have well have just had a thicker surface plate. You could make a surface plate to be any thickness you want. The laws of practicality say that the normal thicknesses are good enough. To put one on an elastomer on top of another plate is merely a more complicated way of just having a thicker top plate. Keep it simple.

Sent from my SAMSUNG-SM-G920A using Tapatalk
 
A Hilger & Watts autocollimator would be lovely to have? Do you know where I can get one for $100? That's the BoM for what I have in mind.

The NASA Tech Note used a constant frequency, so tuning is not an issue. The whole point of using a Wien bridge oscillator is to suppress harmonics at the source by amplitude control. I was studying the NASA circuit today and don't understand it on close inspection. It looked familiar, but is not what I thought. The diodes are an array of 1 pF capacitors below full scale of 0.7 V. So they are being operated as precision matched 1 pF capacitors.

Jim Williams wrote a stunning paper on the subject, "Max Wien, Mr. Hewlett and a Rainy Sunday Afternoon". I *highly* recommend it to anyone with an electronics background. And even without that, Jim's presentation is a work of joy and beauty.

As for the practicality of the elastomer, look at the deflection equations for simply supported beams with constant loading and compare them to the point loading case. Just flip the diagram upside down.

This thread will give you an appreciation of just how accurate a sensor you can make out of a $10-$20 vial and a handful of electronics. JBeale and branadic posted the important results.

suggestions for high-resolution tiltmeter (inclinometer) sensor? - Page 1
 
gedankan of support/table, this thread is pretty cool, has me not paying attention to work.
I would think a natural rubber sheet between rock table would be best: It can be of hardness to be at 1/2 compression with empty rock top, has dampening qualities that most synthetics do not, and would form to a table sub-bed that is pre cambered opposite sag, leading slightly more upward support - via the rubber- to the middle of the stone top. The table legs would be 4, until near ground then magically (not sure how yet- still a few hours left here) to three. Three is fundamentally more stable and easily adjustable to rough out initial plane.
Having two surfaces uni planer seems like it would get tricky fast. Level 5 ft from me is not level where I am, and is it perpendicular to theoretical center of non squishy earth (stationary) or cmass of Earth right now (measurable, but dynamic)? Math, you can define level as anything you want, to anything in universe.
The light deformation is tough, at first it would seem heat makes since - but if mirrors (optical glass/AL) deform then I would guess it is more the excited electrons in the metal from blue light trying to jump ship. So, does the stone plate need to be grounded electrically to slow the electrons roll?
The support table, would it be mixed length angles over wood to dampen and hopefully have self canceling harmonics?
 
The high accuracy the OP wants and 100$ being costly do not go together.

There are large 30m long machinery tables and measurement systems being made, accurate in travel to 1 micron or less to 1 or 3 planes, aka length plus xyz tilts.

OP:
Look at optical tables.
More accurate and much more rigid than surface plates.

Anyone can make one - given lots of hours and a few bits of kit for 300$ or so for metrology.
Optical flats, mono light source.

Amateurs routinely lap glass to 0.1 micron or better flatness or complex curves for astronomy diy.
You could, as well.

The state of the art in industry is probably in microprocessor fabs.
I believe 13 nm feature sets, so about 1 nm repeatability and accuracy locally.
 
A little clarification. I got the idea for a low cost, high precision means of testing a reference plane using two sets of orthogonal precision levels as outlined in the NASA Tech Brief and extensively discussed in the EEVblog thread. One pair attached to the plate or plate support and the other as a movable test probe with a small computer to calculate the changes between the two planes.

My initial thought was to seek a patent for the idea, so I was a bit obtuse in the initial thread. Subsequently I concluded I do not need the money enough to justify the work and expense of getting a patent and putting instruments into series production.

I'm aware of traditional practice for such work. The instruments are too costly for intermittent hobby use. Substitution of a 2"/div vial for a 5"/div vial gives a resolution of 1 nm/m. At that resolution *everything* else becomes a problem. Lapping for a few seconds will raise a thermal bump.

The direction of gravity is a convenient reference that one can readily use to establish an orthogonal plane. The direction actually changes as you walk around a plate mounted on a stand. However, that effect is even smaller than the height effect. The height effect has no bearing on the direction of the acceleration vector, just the magnitude. Other masses in the vicinity will change the direction but the effect is *extremely* small unless it is a mountain range as was discovered during the British Ordnance survey of India during the 19th century. Deflection of the floor due to a human walking around is orders of magnitude larger.

The elastomer pad is just a convenient way to introduce uniform distributed support. Same concept as floating in mercury, but cheaper, safer and more stable. Pads and castables of Shore 00 20-40 with plate motion constrained to vertical by guide bearings is an inexpensive way to create uniform support.

If you consider the deflection of a supported beam (i.e. not fixed) with uniform loading and turn the figure upside down you get the case for two point loading on a uniformly supported plate. Then compare the case of a point load for the same supports. All I've done is look at the denominators in the equations in Machinery's Handbook. For the various rectangular beam cases I looked at the denominator is 2-4x larger for distributed loadings vs point loads.

I might get lucky and find the deflection of a uniformly loaded and point loaded round disk supported at 3 points as solved problems, but for my purposes, just the information from Machinery's Handbook is sufficient.

I'm primarily concerned with mounting a 24" x 36" x 4" B grade Chinese plate and measuring the errors accurately.

I'm also curious about some work Wayne Moore mentions in "Foundations of Mechanical Accuracy" with regard to the deformation of "granite" plates as a consequence of wetting. I worked for several years in the field of the mechanics of porous media in the oil industry. In particular the stiffness and deformation properties of rocks as a function of pore fluid properties and pressures.

Based on that experience, I would not expect a plate to expand upon wetting. If it does, the only mechanism that I can think of is surface tension increasing the pressure within the pores. I plan to discuss this with a friend who retired from a career as a research scientist doing laboratory measurements for a supermajor once he returns from wherever it is he has gone to visit. He built a very specialized machine for such work 20 some years ago. To the best of my knowledge only two such machines have ever been built and the costs were well over $1 million for each.
 








 
Back
Top