Linear actuator using ClearPath Integrated Servo System @ 72V DC

...and the full version

The first pictures of the parts that I printed in NylonX. I really like the look of it with this rough matt black finish.

It is however not that easy to print (being nylon after all) and likes to warp :-(
If I cannot get it to print warp free, I might actually have to use PET-G, we'll see.


Regardless of the material, the belt clamping mechanism works like a charm No tools needed, no moving parts and it holds the belt securely in place. The teeth of the belt (GT-2) interlock nicely when being fed through the "channel".



Here a comparison between PLA (black) and PET-G (Orange). PLA is simply gorgeous to print! But PET-G is the only serious alternative to NylonX when it comes to impact resistance. PLA (unfortunately) can shatter like glass under load.

Probably also noteworthy: Initially I screwed directly into the plastic part. Just leave a X-mm hole for a X-mm screw and it will give you a nice tight fit. Cutting a thread into it is also possible, but either way I was concerned it wouldn't hold after tightening and untightening a few dozen times. So I switched to using these threaded inserts. Work great, just heat over a flame for 3 seconds and press them in

...yes, of course, not with your bare hands

Dirty

Great work dirty! Looks like you're going to have a serious rig in no time!
Can I ask how you plan to wire your end stops? Will you be using multiple relays and killing power directly or wiring them to the AMC1280 or another way? I see sabertooths have a function to kill as well.. I'm not sure which route I should go.

Also that ball screw spacer is an awesome idea!! I was scratching my head for ages thinking about a way to keep my rod spaced evenly from my ball screw but couldn't think of a good way so I ended up connecting my tube without one. I'm hoping it doesn't rub against the tube too much in operation. But it would also would as a great spacer to ensure I have epoxied my tube to the ball screw straight as well. Don't suppose you mind sharing the cad file by any chance? I wanna utilize the idea on the other 5 actuators!

Hey there

Sorry for the late reply!
Yes, of course I will publish all the CAD files. I was gonna wait a little longer to at least do some more serious testing with it, but so far the mechanical side of things are looking rock solid. Give me some time to brush up the CAD files. Some parts are still named in German

If you have the chance (with reasonable amount of work/cost) then I would definitely add some form of centring device to all six actuators. Without it, I'd expect there to be two major fields that could become problematic:
1. Noise ("Klonk-klonk") when the ball-screw dangles around inside the hollow rod.
2. Wear and tear because when the ball-screw has room to move, the load on the ball-nut will be unevenly distributed and might eventually cause failure.
...doesn't HAVE to happen this way, but since we're talking about a 10ct 3D-printed part here...

So, you glued the tube directly onto the ball-nut? Wow,... interesting... never occurred to me... but isn't that a nightmare to disassemble, maintain, inspect, replace, grease, you-name-it... ?

I have put the ball-nut inside a 3D-printed housing from one side...



...and then added a "lid" on the other side to allow the ball-nut to also transmit pull-force onto that housing.


This housing has a M28 x 1mm thread printed right into it on the opposite side...


...into which I can screw the 28mm hollow rod. I have had a friend cut a M28 x 1mm thread into it with a CNC machine which to me is hands down the most beautiful thread I have ever seen!


The two holes in the side of the housing and the cutout in the top are there to insert two M5 clamping screws that compress the M28 thread after screwing-in the rod and thereby create a reliable force-fitted connection.

this way I get a removable connection between the rod and the ball-nut that will easily transmit the ~500N nominal pull-force and ~1000N nominal push-force that I have planned for.

Very important to watch to not loose some pulse at 4000 rpm !!! Watch my threads ,,,www.xsimulator.net/community/threads/building-a-new-6diy.12321/page-3

Yes, that's what I am still concerned about as well. But the sensors are not mounted directly on the motor axis (4000RPM), but are instead mounted on a belt driven accessory drive as shown in this post:

...this reduces the speed of the sensor to a maximum of about 10 sec^-1. The sensor I'm using will be the AEAT-6012 . At this point I should probably admit, that I have very little (I mean VERY little) experience with sensors/encoders... anything that senses a position. In my childish naiveté I have just assumed that those sensors are absolute position sensors which (I guess) means that they sense their position directly and not by counting steps from a known starting point.

--> Someone PLEASE correct me if I'm wrong here!!! <--

The AMC1280 reads their positions via SSI and can even handle "multi turn". So I was NOT concerned with the question: "Will the AMC1280 miss steps?" but rather: "Will the AMC1280 miss a half revolution?" And since a half revolution at max speed will still take ~50ms, I simply shrugged my shoulders and went "Nah... doesn't sound THAT hard to me..."

But in the end, I guess it all comes down to the question that @Thanos should answer best:
"Up to what RPM is the AMC1280 able to read these magnetic encoders in multi turn"?

@Dirty
The main issue with using encoders be it optical or magnetic even if it as index pulse is that if the encoder turns more than 1 revolution.
Each time you power off the controller will lose center and end positions.
Even if the controller remembers the last position before powering down, imagine if the actuator crept downwards...
In multi turn application like yours you need to for go the index pulse and instead use end positions or lower and upper half signals.
One solution may be that you attach a optical detector to the belt which would indicate bottom/top half of your travel.
At power up you can initilise based on the detector's output, at change of signal would indicate center(index) position.

Or if possible just "park" the rig at shutdown at known positions. Don't ballscrews lock in position? If no then park at bottom positions where they can't go any lower.

well,... it depends Ballscrews with a very low pitch might self-lock when unpowered, but the SFU1605 (5mm pitch) does not. It takes about 5Kg/10lbs/50N of load to move the ball-nut on the spindle, so I am pretty sure that the platform will collapse under it's own weight. Even with 6 actuators in the most (un)favourable geometric constellation.

....which is a good thing! I heard of worm gears which self-lock when back driven, creating a kind of "stu-tu-tu-tuttering" in situations when the platform is being commanded to move downwards slowly. I'd definitely see that as an "immersion breaker" to be avoided. Was one of the main reasons why I didn't go for a crank-arm system.

I don't know, if the AMC1280 has this "park on shutdown" feature...!?! I think however (not sure though!) it has a "calibration on start-up" feature which would be all I need in this case.

I will check @Thanos website...

These are the videos in which @Thanos describes those sensors in multi turn. If I understood correctly, the controller will run the actuator up (or down? ...or both?) against an end stop switch on start-up anyways. So I guess this calibration pattern will run before every use.

Dirty

I finally finished the u-joint



I know there are finished u-joints out there, but they were all small and shaky and wiggled around under load changes and didn't have ball bearings. All in all, nothing I was wiling to spend money on.

Initially I was going for this design:

It utilises these cardanic crosses which are actually meant as a replacement kit for the driveshaft of a 4WD car. The good thing was, that it was a very compact design but since these crosses were made out of one single piece, I had to make the plastic parts as two separate shells,.... and it's already hard to design them strong enough without having to worry about the joint between the parts. So I abandoned it :-(

My new and most probably final design doesn't build quite as compact, but is a lot easier to assemble and strength concerns are no longer an issue. The base and the top can be 3d printed in one piece, no joints! However, I had to de-couple the x and y axis from each other, adding a little bit of extra height to the design. It's capable of doing ±35° in all directions. Design specs were at least ±30.


A little animation showing the assembly


I had a couple of versions before I was satisfied. Especially the fact that I wanted 500N pull force meant that I had to dimension the wall thickness appropriately.

...after all, it''s called DESTRUCTIVE testing for a reason This was an earlier version with thinner walls over the top.

More tomorrow....

Dirty

Those u joints look pretty slick.
If you decide they aren't holding up like you want you might try these.
https://www.surpluscenter.com/Power...ints/1-Keyed-12-HP-Universal-Joint-1-2730.axd

They look plenty strong and compact. I'm going to order up some for my own experiments.

Would be great if you can share the design of the spacer...unless you want to sell it, in that case I don't buy

Holly! Ditry , this Joint is a piece of art...and fragile anyway ;-)

That´s correct...worm gears with ratio more than 1:50 creating a stutter in slow pull conditions. In my case the gearbox even broken down after 100h use.

So,... finally time to write an update on this project.

I've been pretty quiet lately, not because there wouldn't have been enough to tell, but because I really wasn't sure as to where the path I'm on was gonna take me. Or if it was worth pursuing in the first place. More on that some time later in its own thread.

Last November I was about to announce a major milestone reached: "End of mechanical construction"
...at least for the actuator part of the project. I have done some load tests with myself (95Kg) loading the actuator and it didn't make a difference wether it was running unloaded or with all my weight on it. Of course, current consumption was increased, but the motors spun up to the commanded speed regardless of load. That's exactly what I wanted to achieve.

There was a minor drawback, when I learned from @Thanos that the AMC1280 in it's current firmware unfortunately does NOT(!!!) support magnetic sensors in multi turn. :-( Ahhrg,... what a nuisance! I wanted to avoid having to use pots, but that meant, that I had to adapt the design of my sensor housing a little. Now I connect a simple 10 turn pot to the shaft of the pulley.


Not elegant, but pragmatic. And it works OK. I read the voltage from the pot via a 12bit analog-digital-converter. So I get 4096 steps for the ~380mm of travel --> 1/10th of a millimeter resolution. An order of magnitude less then what I hoped for, but I think still good enough. I will need goooood (and probably expensive) pots tough. Guess I'm looking at an extra 6 x 50€ for the pots alone. If anyone can recommend some precise 10-turn pots,... I'd be happy to hear which ones you recommend.

I made myself a drill-jig for the holes in the bottom aluminium plate. I have no clue how I could ever have done anything before I had a 3D printer...



The clamping mechanism is certainly not the most beautiful piece of design on this actuator, but it works absolutely perfect. Even after a few months, it holds the belt in place nicely and is tool-less.


So, that was pretty much "it" for my actuator design. The design works as intended. All specs are met:
- 1000N continuous
- 3500N peak
- 333mm/s speed
- ±35° in all directions on the U-joints



...there are a few areas, where I will continue to develop:
-->I want to "compress" the actuator a little. A total of about 150-200mm can be shaved off through a combination of different measures. Mostly cutting away regions that I had used as buffer zones. And the distance between the two upper bearings can be reduced by about 50mm, maybe more. I even considered mounting the motor on top of the actuator in reverse and transmitting the torque via a belt drive. It would've given me an additional extra ~180mm of room, but since the actuator needs a certain minimum length anyways, there was simply no necessity for it.
--> Wiring will be routed through the hollow inside of the aluminium beams whenever possible and all wires (except the 72V power) shall terminate in an RJ45 connector at the bottom. Then I can hook the actuator up to the controller with a simple Cat6 patch cable. Those are shielded, cheap, flexible, easy to crimp.... what more can you ask for!

I will of course publish all files (Fusion360 & PDFs) here as soon as I had time to clean them up a little. Right now there are still a dozen joint conflicts in them and I am struggling to turn my CAD file into a PDF document that can be used.

Greetings,....

Dirty

Oh men, this accurator looking so nice! Simply exemplary. And the compressed height i would suggest to turn into travelway instantly

propably you know already that clearpath has added the "follow position mode" to their MCPV. I have the feeling that's the best way to go. you forget pots, limit switches etc.

The interface to its absolute positioning should be easy according to @Thanos

I think I want to be a guinea pig for these motors. With the PWM based positioning setup inside the software, that seems to be the simplest setup. They certainly aren't the cheapest (starting at $800 per motor for the 230vAC 1hp), but I think will be the best. If the proof-of-concept works with one, I'll pull the trigger on the remaining servos.

If I am well informed, Teknic offers 3 months return policy for U.S. so you have nothing to loose