A SCARA (Selective Compliance Assembly Robot Arm) is a type of articulated robot arm first developed in the early ’80s for use in industrial assembly and production applications. All robotics designs have their strengths and their weaknesses, and the SCARA layout was designed to be rigid in the Z axis, while allowing for flexibility in the X and Y axes. This design lends itself well to tasks where quick and flexible horizontal movement is needed, but vertical strength and rigidity is also necessary.
This is in contrast to other designs, such as fully articulated arms (which need to rotate to reach into tight spots) and cartesian overhead-gantry types (like in a CNC mill), which require a lot of rigidity in every axis. SCARA robots are particularly useful for pick-and-place tasks, as well as a wide range of fabrication jobs that aren’t subjected to the stress of side-loading, like plasma cutting or welding. Unfortunately, industrial-quality SCARA arms aren’t exactly cheap or readily available to the hobbyist; but, that might just be changing soon with the Creo Arm.
Slated to go on the market soon as a fully assembled unit, the Creo Arm will also be released as an open source design that you can build yourself. The creator of this SCARA robot, [anfroholic], says the fully assembled production version should retail for less than $4,000. And that’s for an arm that has a claimed positional repeatability of .03mm at a distance of 600mm. Current documentation is fairly sparse, but [anfroholic] has a number of videos showcasing the different functions he has developed, which serves as a nice proof of concept.
Curious about what you can do with a SCARA robot? Check out some of our previous articles on 3D printing with a SCARA, and using one for pick and place tasks.
23 thoughts on “Creo Arm Might Be The SCARA You’re Looking For”
Very lousy video with shitty music.
The robot seems to have some good ideas into it, but it’s hard to tell because of the lousy video.
The (3d printed?) pulley’s probably give it a sub millimetre position capability.
Hackaday says retail version <4000, ??? I'm instantly not interested any more:
This will get you around the lousy kickstarter search box.
This is a better robot for half the price. It was kicked for USD1800 and it really looks like it can do something usefull.
Sorry, i'm in a slightly sarcastic mood today…
The creator says the actual production version won’t have any 3D printed parts, everything will be machined.
(1) Sorry about the video, I wasn’t expecting this to get picked up quite yet. It was more of a test.
(2) This is a prototype, the production version will address these issues
(3) I will admit the price is mildly inflated. We’re a startup and starving for funds.
This isn’t my first time bringing a product to market, I have short supply chains set up already and hope to get the “minimum needed backer’s” arms in time to put under their xmas tree.
(4) I’m familiar with flx.arm. We will be open source and will also be releasing the tool rest portion under LGPL. Meaning anyone can make and sell tool heads if they desire. I would have to say ours can be useful.
(5) No need to be sorry. Constructive criticism builds charector.
In the last few years thare have been a bunch of SCARA’s kicked. all for < than half the pricepoint of this robot.
The video is also so badly made it hurts my mood.
(My first attempt @ commenting was probably removed by Hack a day's spam filter).
I don’t know the creator personally, and just learned about his project in the last couple of days.
Some points to contest what you said:
1) None of those other SCARA products as far as I know are open-source. The creator of this arm also posted on reddit where his willingness to share the design details and provide help to others was already evident. Also, he mentioned that he will share all files on the day he officially launches this product in a few weeks.
2) None of the recent Kickstarter flood of SCARAs (which I believe you’re referring to), as far as I know, show the ability to route/mill things, while this one appears to have done a pretty decent job with wood.
3) Video: Could be improved a lot, no doubt.
I don’t think this hack’s excitement (although the description mentions the admittedly high pricepoint) is about the sold product, so much as it is about a really awesome project that makes experimental steps into a well-established industrial concept whose potential I feel the maker/hobbyist community hasn’t fully explored yet.
Also, achieving 0.03 mm repeatability: This is a fairly bold claim to make when using AMS’s rotary magnetic encoders like AS5048 (which the project creator mentioned he uses), as it appears to have a datasheet-specified 1 degree INL error value, but I’m curious to see how this can/will be fought against.
Thank you for the comments. To address the last part about repeatability, the core of the system is driven open loop with steppers. Theoretical resolution comes out to pi/100mm. From there we can argue about microstepping accuracy but none of this has anything to do with the encoders. The AS5048 datasheet makes a claim of .05 degree accuracy and while I have not been able to achieve that, the ‘ticks’ of the encoder seem to be pretty repeatable. Their inclusion in the project is largely an attempt at missed step detection, zero motion homing and a teach mode where the arm can be dragged around.
My pleasure, great build. Hopefully, you can come out with a more refined video and launch your campaign in style.
Calculation: Does that mean you have a belt+pulley-based reduction of 1:100 on each arm then? Because your Arm-radius = 200 mm. Theoretical resolution = 0.03 mm. Assuming 0.9-degree steppers (400-steps/revolution), I’m seeing you need a ratio of 100 on the pulleys, unless I’m missing something?
There’s microstepping on the drivers. 12.5:1 pulley ratio, 200 step motors, and 1/32 microstepping drivers. 400mm radius for the arm gives you pi/100 at the furthest reach.
On a side note I have caught more flack for that video… I have a professional film crew already lined up for the main video. I film on sunday. And thanks again for your encouragement.
I have addressed a few of these issues above. Let’s not forget that most have not delivered yet, something that I don’t plan to happen. Everyone has a copy of my source files so I plan to be fully transparent throughout the final design and production phases.
Again, apologies for the video, hope your afternoon is better.
I was on board of the hate train until you remarked about the open sourced part, I always liked scara’s and for an open-source hobby project it looks pretty spiffy.
I love the idea, though I worry about the name. On the lower end of my worry spectrum: searching for the product/project is difficult. There’s a well known engineering software suite using the name Creo (http://www.ptc.com/cad/creo) and a lot of people working on arm related projects with the software. On the upper end of the worry spectrum: PTC will protect their trademark. That could lead to confusion with the Creo arm project, or even cost real money.
Thank you for your interest and your concern. We’re in contact with our trademark attorney. Searchability aside, I like the name and would prefer not to change.
I’m sorry I made an error on the reddit post. The arm segments are each 200mm giving a total of 400mm span. 600mm is the distance from the front of the tower.
Great project, looking forward to the launch.
You should be aware of a few things. There are a few open source robot arms out there, that have been out for a long time with no access to any files. The file links are just a ploy to get people to a kickstarter page link. This is upsetting. I can tell this is not what you are up, but to save yourself from internet hate stacks maybe put a beta open source file directory/link somewhere to sooth those doubting minds. It can even be empty with a coming soon comment. Just put a link somewhere so people know it is on the way and don’t waste time and energy asking where the files are. Or just duck down back into the shop bang out the files in a month and post them when you have it ready.
If there is a way you can post prelim CAD files, I would love to see it as I am working on building a hobby scara arm would prefer not to have to many iterations of of a failed pivot joint. My main questions are on bearings used and drive train setup, but I am happy to wait till the files are up officially when the launch kicks off.
Also let me know if you are looking for some injection molded components. I may be able to figure that out.
Thanks for your interest. To address your first concern I said I would release *at* launch. This is not, if(you back me), then(I will release). I’m just putting that out there now. If I don’t then the hate train can come (which I know it will to some extent any way I do it).
When I release, that will be the beginning of beta. I’m just not ready, not so much the messy and ugly aspects but proper attribution is important to me and I would prefer to cite, attribute and honor all those who’s works I have derived.
The main bearings used are 12 and 20mm ID ball bearings. I like to use vxb.com if you’re in the states.
Thanks for the offer with the injection molds. Building and designing them was my first career.
Thank you and likewise
With ball bearings do you notice much out of plane movement? The more I look at it the more I was thinking I need thrust roller bearings. The high end bots use cross cylinder bearings, but those are super pricey. I would like to have enough holding power to keep a steady 3D printer filament lay, but am not trying to do anything a precise as a medical bot. From your videos it seems to work well, but I would love to know what you think about trying a roller bearing set.
With the proper preload I have not noticed any appreciable play in the joints. Even loaded with a few kilograms they seem to hold up very well.
Awesome, thanks! I was probably overthinking it, and the more I can keep the build cost down the better. Yeah I took another look at the router milling video and the resolution seems good even with that huge cantilever.
After a few day’s i’m in a better mood.
The first video I saw was actually the worst by far.
The other videos show this is actually a pretty cool project.
I’ve been looking at a lot of different robot arms and every now and then I’ve been thinking a bit of building one myself.
If i’m ever going to build one I would make a lot of the same choises as you did. Rpm reduction (resolution/torque enhancement) with timing belts is the way to go. Placement of motors, and using different sizes also looks good.
Some of my other ideas:
Making a 2 stage reduction prevents the timing wheel from getting so enourmous.
Using gears for the first reduction would make it pretty compact. The play in the gear would not be a problem. It’s a lot less than the elasticity in the stepper motor itself, and it’s also being reduced by the timingbelt. (But it would make it more difficult to reproduce for hobbyists.)
The big wheel for the timing belt does not have to have teeth. Because it’s not rotating a full revolution the timing belt can be fixed to this wheel.
This would also ease the tolerances for this wheal. Getting timing belts to fit perfectly over 100dth’s of teeth needs pretty tight tolerances .
Aluminimum is lighter than steel, but it also has half the stiffness, which is a disatvantage for robots.
Using hollow profiles increases the weight to stifness ratio a lot (When properly mounted).
@Josh: Regular deep groove ball bearings have reasonable ratings for axial loads. Normally between 25% and half of the axial rating. (And a bearing with 20mm id would have plenty).
It’s easy to loose a lot of the stiffness in the bearings / axles (which is the main reasen a 20mm id is used).
Some note’s about stepper motors.
Using microstepping usually does not enhance accurarcy by much. Actually the worst case accuracy of a stepper motor is 2 full steps under load (max torque).
This is easy to explain:
Take a stepper motoer and apply the rated current. It wil go to a fixed position (No load).
Then gently apply a bit of torque. It will react a bit springy. The harder you push, the more it will move, but if the load is removed it will move back to the last position. If you push a bit too hard however, it will suddenly move to the next position in which it is stable. Because the current through the coils has not changed, this new position will be 4 full steps away from the last position. The torque to angle difference is probably approximately sinusoidal.
If you close the loop (Hall sensors) then you can use microstepping to generate a phase difference (= torque) and achieve the full resolution of the sensors. Instead of the “brainless” steps such as usually for steppers this does of course need full (pid or similar) control and then the difference between a stepper and a bldc is becoming pretty small. (Yes, there are 3-phase stepper motors).
Another cool idea:
These Hall angle sensors are becoming pretty common & cheap. If you use 2 per axle (one directly on the stepper, and one after the reduction gear) you can use the combination to get a pretty high angular resolution.
I’m curious where this project goes…
Thank you for your second input.
Making a 2 stage reduction prevents the timing wheel from getting so enormous:
This was a conscious choice I made to keep the complexity down. And regard to a fixed belt, it would reduce the usable motion by some. For years I have had good results printing pulleys material heavy and truing up in the lathe.
I’m aware of the limitations with microstepping and the springiness of steppers in general, but still thank you for the info. The choice of steppers is because I am using 3D printer hardware. Which correlates directly with my projected pricepoint. I plan on that including 2 complete hardware sets. I’m making the bet people want to start playing immediately. By owning cupcake #1344 I was in the eye of the storm and have no intentions of putting our community through what they did, this allows me to shamelessly EOL Gen0 hardware. And in the interim we can design robot controller proper, with the advantage of time, hindsight and feedback. Being open source the whole controller will be a gift to the open source community, there’s a special piece I will be excited to give to this community in particular. I will be making every effort to do the open source thing right.
With regard to the hall sensors, I’ve been debating whether to use 2. In the period I have been working on this project the price of these have fallen by over 30%. With your comment I will look into what it’ll actually take to implement.
With the regard to the video, most of my others have less than 100 views. It was more of a test I was sharing with my editor. Who knew it would get picked up. Say la vie
Thank you again for your input, always appreciated
Paul, where would I find the axial load ratings? I am not so much worried about failure as I am worried about wondering. The VBX site does not post axial specs for most of the bearings. Have you verified that the stability of deep grove bearings over time? If so I would love to know what applications. Thanks!
Please be kind and respectful to help make the comments section excellent. (Comment Policy)