Bearing ball free play and expansion question.

using 316 stainless steel 1/16" ( .0625")bearing balls. I have designed my racings at exactly .0625" space for the balls. To allow for imperfection, I feel like I should spec to .064" +/- .001" for the racing area. Does this sound right?

I have designed a device with .015" clearances as a standard throughout. It may produce some heat but nothing "hot to touch". Is there a rule of thumb for expansion? I'm thinking this is a complex formula which I will likely just have to avoid by direct prototype and see if it locks up when hot. Any advice on figuring expansion and clearances?

A few hundred individual parts. Nearly 50 individual magnets. All whirling around with 1/64" clearances. Designed by a retired tech, not an engineer. The work needs precision to the point where I am farming the work out because my blind self just can't get that anymore. I was trying to simulate an imaginary cosmological event and it turned into an alternator. Have you ever found yourself wondering "Why do I have such weird hobbies?"

I'm not sure how to respond to this post. If you want to experiment, have fun, but be aware you could be throwing money at something physics may frown on.

For some unasked advice, don't use 316ss as a bearing ball, if you need non-magnetic properties then there's plenty of cheap (especially at 1/16") ceramic balls you can use, either Al2O3 or Si3N4. Otherwise, use 440C balls. And thermal expansion is a well defined subject, search for CTE of metals and you'll get lots of hits with values for different materials.

But if you come up with something neat be sure to let us know about it.

If you're making something like a rotec bearing or large thrust bearing I agree that stainless is a real bad choice. We use plastic balls on anodized aluminum races with good results for products we make. The race grooves are a larger radius than the ball diameter and the races are "angled" so the centerline of the radius's do not have to line up perfectly.

If you try to make them line up perfectly it's a bear. You will have to make special tools to measure the race groove locations.

" I have designed my racings at exactly .0625" space for the balls."

Your using the wrong ballpark. Modern bearings are designed with tolerances measured in microns, not
thousanths of an inch. Have you considered using a stock part for your machine?

This one is just a proof of concept prototype. I absolutely intend to use full bearings eventually but if you've never prototyped, everything changes so go cheap. Heck, the original design was 48" in diameter and it is down to 9". Theoretically, the ultimate design is infinitely large. Compromise, compromise, compromise.

As jim said.
Really.
0.01 mm or 0.0005" error is pretty poor, and will skate/burnish surfaces and not work well at all as a bearing.

You need to get the size error down to low single microns.
Often "bearings" of any quality are squeezed and in compression, by less than 1 micron in differential size.
I think it is quite a bit less, but am pretty sure they are squeezed.
Ballscrew nuts are one common cheapish example.

You did not tell us what you need, speed, load, precision, low backlash etc.

If what you need as I understand it is better quality, lap the races with a brass lap.
Make a good brass or ci lap.
Use a suitable abrasive to charge the lap.
Lap the surface for maybe 2 minutes for 9".
You wont be able to make the races (width, depth, profile) to size, but You can sure make it smooth and flat and very uniform, just like 100k people before you.

I would try with one lap for bottom, and one-2 thin one-s for the side(s).

Get a set of graduated balls, in 1 micron sizes, and test them, for contact/load.
This tells you the size You have.
It is not all that hard to make smooth surfaces quite uniform, to approx 1 (2) microns.
What this means.
If you make 50, their sizes will vary, somewhat.
They will need different size balls.

It is very very hard to make lots of same size in width + depth + profile, to same size, cheaply and quickly.

For reference, gage blocks are supposed to be lapped to 0.01 microns (Moore or a commercial manufacturer or a us standard, best as I remember).

Don´t dispair.
Double the accuracy from half a thou or 0.01 mm, will work for all but precision apps.
Sanding and flexible abrasives of any type have no place in precision apps.

(Special exemption. Grin).
It is possible to machine-scrape with flexible pads on small grinders, die grinders, dremels etc.
This makes lots of small 0.01 mm divots for oil holding, when crossed, for overall accuracy that can be as good as 1 micron.
Laborious and slow, like all scraping.
Potentially very accurate.