Say hello to my little friend.

And I do mean little; d is 6mm, D is 19mm, b is 10mm. And at his list price, he is worth two thirds his weight in gold.

And I need three of them.

here's the thing

You don't often see "axial angular contact ball bearings." They have exactly one application: ball screws. The purpose of a ball screw is precise transformation from rotational motion to linear motion. they come in several grades and mine are confirmed to be "there are none more precise" grade through empirical measurement. If you want to go more precise you have to go hydrostatic and my Kern comes from a time before Kern went "we have to go hydrostatic."

The purpose of "axial angular contact ball bearings" is to eliminate all backlash. You've got a bearing race on one side, you've got a bearing race on the other side, and by precisely squooshing them together you make sure that the bearing itself never moves while you're moving other stuff. Within normie machine design your normie angular contact ball bearings end up about tight enough because everything else is about tight enough and you get your tenth of a thou or half a thou or thou or ten thou or are-you-fucking-kidding-me twenty thou without resorting to eldritch magic.

The theoretical repeatability on my machine is a quarter of a micron. I will never hit it? Fuck, I have no way to measure it! But I have to try.

I watched a guy "set up" his CNC machine. He's got stepper motors. He has 400 pulses per unit, which is an important number to know, because it tells you how many signals the controller sends the motor to move by any given amount. His 'unit' was millimeters, so is mine. So his machine has 400 increments between 0 mm and 1mm, or every pulse of his controller potentially moves his ball carrier two and a half microns. My machine has 37 million increments on the X, 74 million on the Y and 84 million on the Z. This is mostly useful for things like active noise cancellation, backlash compensation and other DSP-driven accommodations of the limitations of physics. After all, a coronavirus is about a tenth of a micron. The glass scales capable of measuring to a tenth of a micron will tell you that you need climate stability on the order of a tenth of a degree per meter of length if you wish to maintain accuracy. So - the difference between "part made in the morning" and "part made in the afternoon" can be a micron even if everything is mathematically perfect.

There is exactly one legitimate manufacturer of six by nineteen by ten axial angular contact ball bearings. I'm going to guess they manufacture them at the behest of exactly one ball screw manufacturer. And if you need, them, you bloody well need them. I say "legitimate" because I have sourced Chinese counterfeits for a tenth the price. I will not be purchasing them.

Where things get exceptionally stupid is where you start digging into bearing pre-load. See, my little friend wants exactly one newton-meter of preload in order to suck up the backlash. He'll take 1100lb of axial load, which is good because the Z axis has a constant force of 50 lbs against it - this is probably why I need bearings, because the brigand I purchased this machine from did not understand "axial angular contact ball bearing" and redesigned the machine to BYPASS THEM. As a consequence, one of them is annihilated. The other two might not be? But considering the violence they've experienced (going from 8:1 reduction of a 25W motor to direct screw drive by 750W) I'd be a fool to leave them alone.

So okay. How do you put zero point seven foot pounds of preload on a nut? I mean, your typical lockwasher will do more than that. Your typical Nylock will come off with zero point seven foot pounds of preload. We're talking a truly minuscule amount of torque; I have a torque wrench that will measure all the way down to half a newton-meter in hundredth of a newton meter increments because of course I do, I have an Italian hyperbike. But it won't alarm until 1.5 newton-meters. Even at my most ridiculous I'm in the margins.

Fortunately, the bearing is German. There are procedures.

that little guy? My little friend's little friend? Has precisely machined little sawtooth combs inside that are pushed against by precisely machined little set screws. You torque your nut down, and then you affix it with a 2mm allen wrench via axial force. They're pretty dear, too. List is like a hundred bux. And again, I need three.

Did I say things had gotten "exceptionally" stupid? Well they're about to get thunderously stupid. See, d is six, D is sixteen, h is 8. This guy is teeny, tiny too. And yet, you'll note it is not "nut shaped." It is, in fact, a "DIN1810" nut, subject to ISO 2892.2. Should you need to buy one of these, you buy a "KM wrench" where KMxx is like a dress size. KM wrenches range from a KM72, whose outside diameter is an eye-popping 460mm in diameter, all the way down to a KM00, a minuscule 18mm in diameter.

Do you see the problem?

INA didn't. I called them. When I asked what wrench, exactly, they wanted me to turn a 16mm KM nut with, considering the KM spec only goes down to 18mm, they said they'd have to get back to me on that. I suspect there's some AMS socket no one has ever heard of. I did ask if anyone ever bought these magical little friends and they assured me there was an inventory of 126 of them within the continental US so they keep stock for a reason.

I did reassure the engineer on the phone that at the prices I'd been seeing I was unlikely to purchase said wrench (I mistakenly looked up a "KM16" for a 16mm wrench before I figured out the dress-size thing; they're $800). I was far more likely to fish around in my toolbox for an errant abandoned 16mm socket wrench and maul it with a mosquito grinder. But I wanted to know if such a thing existed. My quest was in vein. My cousin, meanwhile, volunteered to maim the socket because the idea of a rednecked harbor freight socket to tighten a $100 nut one newton-meter on a $650 bearing appealed to his sense of the absurd.

At which point I realized that with a given force of one newton-meter I could 3d print something out of ABS that would be ten times as strong as I need. At which point my cousin pointed out I could glob silly putty on the thing, stick a torque wrench in it, hit the whole mess with freeze spray and rest assured that the force had been fully transmitted.

I've accomplished other shit? The auxiliary spindle works? Everything powers up? I haven't started programming it and I'm still missing bits but we're making progress?

And a big part of that progress is truly wrapping my head around buying two thousand dollars worth of bearings and nuts that I can tuck in my cheek.