a thoughtful web.
Good ideas and conversation. No ads, no tracking.   Login or Take a Tour!
comment by kleinbl00
kleinbl00  ·  6 days ago  ·  link  ·    ·  parent  ·  post: Pubski: August 3, 2022  ·  x 2

this was also a long time coming

Worthy of note: those numbers are bullshit. The counts per rev on the motors isn't vaguely right, there's a 9:1 gearmotor between the motor and the ballscrew, and the ballscrew calcs aren't even incorporated. And uhh obviously the motor and the ballscrews aren't even physically connected.

But that's the software, cheerfully controlling a servo motor to a tenth of a micron.

The backlash of the gear motor is under 3 arc minutes, or under 0.05 degrees. The backlash of the GT2 belts is 2.7 arc minutes, or also under 0.05 degrees. two of the axes are 4mm/rev ballscrews, one of them is 2mm/rev. .1 degree at 4mm/rev is 0.0011mm, or 1.1 microns.

The machine originally used closed-loop control via Heidenhain glass scales that were totes stolen by the brigand that sold me the machine. With that closed-loop control the machine managed 1-micron precision. I can buy Mitutoyo scales that will work with a module for the servo pack that will get me to within 0.01 microns, or "a coronavirus." I don't think it'll take that. To assume mirror finish for any waveform you need half the wavelength. Visible light starts at around 370nm, so half of that is 185nm, or around 0.2 microns. The motors, for their part, are 24-bit encoders, so 0.0013 arc minutes per pulse or 0.077 arc seconds. 0.073 nanometers per pulse at which point you acknowledge you're measuring absolute fucktons of noise. 4600 pulses just in the combined backlash of belt and gear motor.

But I've taken this creature from "is it possible" to "do I want it."

I got the motors to wake up yesterday. They appeared in SigmaWin and I could jog them. I choked up like I was watching the end of Babe. I've got at least one dead servopak; I paid $190 ea for them because the local guy told me they were $3k and fuck him. I could buy another for $190 used or $400 new out of China or, apparently $1100 out of any scrupulous North American distributor who isn't giving me the fuck-you price. I found this out when I inquired about getting mine fixed and was told they won't fix it if it'll cost more than 70% of the new price or "around $800."

Here's a $4500 mill. Like that surface finish? Here's its stepper motor. A B C D, baby! Mine have 1500 parameters, life-cycle monitoring and not one, not two, but five thousand-page manuals. Which allow fancy moves like this fucking voodoo at 3:30.

I'm literally at "the plane flies." It's not ready for passengers? I wouldn't take it across the Atlantic? But the proof-of-concept has proven out and this fucker IS GOING TO WORK.





goobster  ·  5 days ago  ·  link  ·  

Holy crap. That "voodoo" video is NUTS.

At first I'm getting vibes of the Turbo Encabulator, and having a larf.

Then the vibration damper kicks in and the top of my head blew off and my brains boiled. WAT. THE. HELL.

It's the little things that impress us rubes...

kleinbl00  ·  5 days ago  ·  link  ·  

FUN FACT

When Paramount rebooted Star Trek and put soap opera people like Marina Sirtis and Jonathan Frakes in it, they went similarly low-budget on everything, particularly the writing. Fortunately they managed to find a handful of pie-eyed dreamers who were so stoked to be working on STAR TREK, FFS, that they gave it their absolute all. Not that they could pull it off on their own, of course. Geordi LaForge's dialog, in particular, generally left the writer's room with a block of TECH where his lines would have been because one of the writers had a college buddy working on a Ph. D in electrical engineering at Cornell. They'd write something, he'd call up his buddy and get a bunch of tech nonsense, and it'd be Geordi's lines the next day.

The show did much better than anyone hoped, of course, and oddly enough the guy whose job it was to come up with TECH decided fukkit, let's head out to Hollywood and have a little fun before joining the rat race. The writer's name was Ron Moore and his college buddy was Naren Shankar.

Ron ended up doing pretty well, obviously, going from "fucktons of Star Trek" to rebooting Battlestar Galactica to Outlander to For All Mankind. Naren Shankar? Star Trek to Sea Quest to Outer Limits to Grimm to Almost Human to The Expanse before also ending up on For All Mankind, but not before making an absolute mountain of cash as an executive producer of a little franchise called CSI. So the next time you see a bunch of rubes chortling about stupid jargon?

Keep in mind that to write bullshit you have to speak bullshit fluently, and that the most fluent bullshit speakers are often making their jokes at your expense.

kleinbl00  ·  6 days ago  ·  link  ·  

my bad 2.7 arcseconds of belt lash is with the 52-teeth pulleys, the ones I rejected as being not ridiculous enough. I have 100-tooth pulleys which are going to be preposterous and also get the belt lash down to 1.4 arc min. so not 1.1 microns, 0.7 microns.

The original setup was not one, not two, but three 8mm T2.5 MXL belts, each inducing 26.7 arcminutes of lash for a total of 80 arc minutes or 15 microns. It's abundantly clear that the original design relied heavily on closed loop feedback as that's, like, half a thou. Which is pretty nice for like a Haas VF-2. Which will also churn our a cylinder head, not a watch case.

(haven't opened this excel file in a while...)