View Full Version : Ballbars

01-14-2009, 05:43 PM
Evening all,
I have to make (or more accurately get made ) some spherically concave magnets.
Ignoring the magnet part for a second these things have to be *super* accurate. They are to hold Balls for a calibration artifact for the metrology system I work on.
The exact specs are at work, but basically they are for a ball bar (thats a bar with a ball on each end, common metrology artifact) 1 meter between ball centers, which needs to have an accuracy of 1 part in 100,000 (IIRC, might be more accurate than that) so the bar has to end up at 1m +-0.005mm, but with removable balls...:rolleyes:
Ideally we'd like to be able to 'wring' the balls into place so they are repeatably positioned (this is a stepping stone to a more complex artifact - hence the removability) and then the magnet and wrung force will hold the balls in place. IIRC the balls are a few Kg heavy, so probably will end up using neodynium or samarian magnets.
The bar part will be a Carbon fibre tube, and these spherical seats will be fixed to the ends, and set to the correct length by measuring with an interferometer against an Invar standard. Of course that means glue near some very expensive kit...

So, as you can probably guess this has started come from an Academic, and although I think its probably possible Im guessing doing it this way would be expensive.

Any ideas on whether getting spherical concave magnets is even possible, where to look for them, and how to make sure they end up on the ends of the tube positioned to super tiny tolerances very gratefully received :)


01-14-2009, 06:00 PM
I don't know if this is feasible as you envision the part, but the first thing that occurred to me was making the spherical part a laminate stack like a magnetic parallel and installing the magnet on the opposite side, that is, inside the tube. That way you can deal with some of the issues separately from complications of magnetism.

John Garner
01-14-2009, 06:42 PM
Dave --

I don't think you want the ball sockets to be spherical, nor do I think you want them made of the magnetic material itself. Instead, take a look at a magnetic "nest" for a Laser Tracker's "Spherically Mounted Retroreflector" (SMR) and then at the Bal-Tec website.

Hubbs Machine and Manufacturing may be the US's premier maker of SMR nests. http://www.hubbsmachine.com/

Bal-Tec specializes in precision balls, but do know a few things about mounting them. http://www.precisionballs.com/


Just Bob Again
01-14-2009, 06:45 PM
Rare earth magnets are sintered and relatively fragile. They'll shatter easily and have to be plated to avoid corrosion. They'd have to be ground before magnetizing, then plated. Far simpler to machine the actual contact surface from something more reasonable and have the magnet behind it in a counterbore from the back side. You'll need a substantial magnet. A least an inch diameter, maybe 2. You could do the laminated alternating pole stack (harder) or just use a single recessed magnet and have less holding power. Without the laminations, the external field will be much higher and you may have real problems with the standard being attracted to other objects. With the laminations, the field won't be a problem more than an inch away or so.

01-15-2009, 03:56 AM
It seems like you could make a nest for the ball and turn the magnetic attraction on/off with something like a magnetic indicator base.

It's been a while since I handled gauge blocks, but I'm doubtful you'll be able to wring the matching spherical surfaces together without some flat starting points on the balls. If you ever did get a ball wrung into a spherical socket, I'm almost positive you wouldn't be able to get them apart without some sort of release area.


01-15-2009, 06:13 AM
I very much doubt you can have spherical concave magnets made to the required accuracy. Magnetic materials all exhibit an effect called magnetostriction. This is an effect caused by the rotation of the magnetic domains when subject to a magnetic field. In the magnet material itself it causes an anisotropic change in dimensions when the magnet is magnetized. Since no work can be done on a magnet to refine it's shape after magnetization it isn't likely possible to attain the required figure.

Also, the balls will exhibit the same effect and it also will be anisotropic meaning the amount of magnetostriction produced will depend on orientation re the magnet. This isn't a minor technical effect. It's widely used in various types of transducers and is the effect at work that makes transformers hum. It applies to any and all magnetic materials and happens in direct response to the strength of the field, alternating or not.

01-15-2009, 07:07 AM
Did anyone notice the "1 m +/- 0.005 mm" specification? This metal bar would need to be in a 100% temperature controlled environment and nobody should be allowed to touch it.
This requirement is completely unrealistic! Even an invisible dust speck between ball and bar would violate that tolerance.
Academic indeed.


Just Bob Again
01-15-2009, 08:11 AM
Difficult but not unrealistic. A 5 ppm tolerance on length. The tempco of carbon fiber is around 1.5 ppm/degree C. Some special weaves and fillers can achieve less than that, or even negative tempco. It would work in a toolroom with normal care. That's why they chose carbon fiber. It can be more stable than metal by a wide margin. Expensive and hard to work, but a good choice here.

01-15-2009, 09:13 AM
Thanks all, some good ideas for me to investigate further :)

Yes, the layup of the CF is special, it has 0 coefficient of thermal expansion, the only expandy bits will be the endplates and the balls. The Metrology system is also CF, and it also has a very special layup. The accuracy required is high, but this is the artifact used to calibrate the system, and as such the accuracy obtainable from the system depends on how accurate it is.

Ultimately there will be a 'mesh' of these artifacts, as we need to calibrate a system to measure the wing jig of a certain composite airliner, and thats very long. This is (one) of the reasons behind the removable balls, its (theoretically) possible to make a single artifact big enought, but theres no way you'd be able to move it around...

I hadnt considered distortions on the balls due the magnet. Ill look into it and see if that will cause a problem.

The nest idea is similar to the first thing I though of, which was trapping 3 balls between the Big ball and the endplate, which can then be flat, and hence is 'relativley' simple to make and size.

Im really enjoying this job, I write embedded software for a living, but my 'unique' skill set (the recruiters words...) means Im also getting some of the more mechanical items to do as well. :D


01-15-2009, 09:22 AM
The bar isn't metal. It is specified as carbon fibre. If the correct carbon fibre layup is used it will have a nearly exactly zero coefficient of expansion with temperature. This is what I used on my telescope for the truss frame. Pulltruded carbon-epoxy rods have a very slightly negative CLE meaning they expand very slightly as they grow colder. Combined with the positive CLE of the aluminum parts at each end my telescope doesn't change focus over the entire possible range of temperature when taking astrophotos.

Dang, fell asleep for a few minutes before posting this.

01-19-2009, 04:03 PM
Evening All,
Ive been pondering on this, and after a little digging about looking into methods of making hemispheres I found a thread here (http://bbs.homeshopmachinist.net/showthread.php?t=32038&highlight=hemisphere) showing how to do it on a mill.
So I knocked up a seat to fit one of my daughters rubber balls:


Ignore the turned finish, it was a scrap offcut of mystery metal, and was like that when I started.
Looks like the seat part *should* be possible.
Of course they'll need to be cut (or at least finished) in situ on the end of the bar, as the glue will have set then...


01-19-2009, 04:11 PM
Need to borrow my 1 micron per division dial indicator?
Nah, foget it, nobody can measure to those tolerances in a home shop. :rolleyes:

01-19-2009, 04:24 PM
Thanks for the offer, but I hope the Optical Research group at Loughborough Uni has access to the required measuring tools :)
This isnt a home shop job, but there are a lot of practical knowledgeable people round here to pick the brains of.


Peter N
01-19-2009, 04:46 PM
Just had a thought Dave.
You could rough it to near shape, a few thou' or less, using conventional machining, harden it, then send it off for Electron Beam Machining for the finished surface. Capable of nanometer level machining so I've been told.
I've known of the EBM process for some time, but can't point you in the direction of anyone actually doing it.
A friend of mine who works for Dage designing ultra high resolution digital x-ray equipment switched me on to it some time ago.


01-19-2009, 04:54 PM
Cheers Peter, Ill look into that.
I did wonder about spark erosion process as well, but Im not sure of the tolerances obtainable, and essentially its a form tool process, so would need new electrodes for each part I imagine?


Peter N
01-19-2009, 05:06 PM
Yup, sparking is a form tool process, with the electrode being the reverse from of what you need minus the spark gap (around 2-3 thou' a side).
I don't think sparking would give you enough precision Dave, or certainly not a good enough surface finish to get down to the precision you need, as at that level the peaks and troughs from the sparking would be all over the place.

I just did a quick search and found this short Youtube video showing an EBM machine in action, topically enough its machining a sphere :D



01-19-2009, 05:10 PM
I cannot imagine this not requiring lapping with balls. Not that I know about this.

01-19-2009, 05:18 PM
Cool video. That would do (in reverse of course), if only we could get a 1 meter long bar inside the machine ;)

TM, Lapping is a possibility, but if we can get close enough using machines that would be nice (and less skilled). And the closer we can get the less lapping would be needed.


01-19-2009, 06:22 PM
I've been talking to Alwyn

and he says that you're talking impossible tolerances ..because of the 1 metre long bar.

and the only place who would be able to verify them once done is

National Physics Laboratory of London . :)

The plastic you should use should be a plastic called Tuffset

Tuffest is very stable and is even used on the lense positioning of a ZYGO interferometers.

all the best.and good luck ..markj

01-20-2009, 05:04 AM
Yes, I think the length is quite a significant part of the problem.
It seems that the ball seats can be made to the required spec, and I know we can attach them to a zero CTE CF tube, but the attaching is never going to be that accurate, especially as it involves glue...
So I think that either a more accurate attachment method, or a way of finishing to OA length after attachment is required.

It certainly is an interesting project :)


Peter N
01-20-2009, 05:14 AM
How about making the hemispheres, attaching them to the end of the CF tubes (plural), then make a micro-adjustable joiner for the 2 assembled tubes to set the length?
Just a thought.


01-20-2009, 12:09 PM
The glue should not enter into the picture as far as affecting the length. You cut the rod to a rough length and then grind the ends to finish length using accurately positioned stops to achieve accuracy of around +- .01mm. One end is a simple stub fitting that is inserted in the tube with glue. As the glue holds around the circumference of the stub it won't change the length as temperature changes.

The fitting for the other end is special and widely available. It's a differential thread mirror adjuster. The ones in this link are adjustable to 1 micron location accuracy.


01-20-2009, 03:19 PM
Hmm, I like the idea of a differential adjuster. :)
I cant for the life of me see how those ones attach, but the principle should work nicely.
As for glue, there is a possibilty of residual stresses in the glue causing movement as it sets up, or if constrained during curing of it moving afterwards.
Its weird that something that small matters...

Interestingly there is more than a little doubt that this can be made to the required spec (which is of course actually even tighter than I thought!), so it would be really nice to do the impossible here :)


01-20-2009, 05:22 PM
Epoxy resin has a very low coefficient of linear expansion with temperature, about 50% of steel. It shouldn't cause any problems and should be compatible with the graphite/epoxy layup.

I am sure that the company I linked to could accommodate whatever mounting method you come up with. I don't think it would be too difficult to make a socket to trap one end of the adjuster. I also would expect anything that is adjustable to the 1 micron range to be internally temperature compensated.