CoreXY Mechanism - Not for printer

[Dushyant Ahuja]

Hi Everyone,

Not sure if these posts are welcome on this forum, as it is not related to Printers, but purely the CoreXY mechanism. Please let me know and I'll refrain from posting here.

I'm planning to build a corexy table that simply has to move a magnet around planned patterns. It's not heavy duty, nor does it need to be fast. The end result is similar to the 'Kinetic Art Table - Sisyphus Kickstarter' (https://www.kickstarter.com/projects/1199521315/sisyphus-the-kinetic-art-table); only square instead of circular. 

I want to keep it as cheap as possible and hence, don't want rails everywhere. Looked through the internet and saw a few good examples of what I need. Specifically 

1. ttps://www.youtube.com/watch?v=x782l6KkAqE (http://www.thingiverse.com/thing:648119)
   
2. http://www.corexy.com/botofthecloth/index.html
   

I've been able to modify these designs for what I need, 

After the first run; I've realised that I need to modify the design a little bit to get good consistent results. 

1. Change the geared steppers (BYJ48) to thin nema 17 stepper motors
2. Add a couple of more pulleys to move the cross-over to one side - I was taking a short-cut and thought that the cross across the table would work; but it gets stuck with the mechanism, so will move it up like all the other corexy mechanisms

However, a couple of things I'm stumped on are:

1. I'm using a thin nylon string right now - is there something better I can/should use (don't want to go the belt way, as this is a very light-weight mechanism)
2. More importantly, I'm not sure how to tighten the strings and I need help on the best possible mechanisms. 

Can you please help

Regards,

Dushyant Ahuja

[Ryan Carlyle]

How accurate/precise do you want it to be? That makes a difference in how much you should care about certain design parameters. For example, it looks like your lines aren't parallel to the X bridge, which will introduce some skew error into the positions near the edges of the working area. For a 3D printer, that would be a no-no, but I don't know if it matters in your application.

The usual cable used for these printers if you don't want to belts is braided spectra fishing line. It's super stiff in tension so you get more precise motion and don't need as much tensioning travel or repeated readjustment. Quite slippery though.

The easiest way to tension a CoreXY gantry is slotted motor mount brackets.

[Dushyant Ahuja]

Doesn't have to be super accurate - I'm sure a slop of even 5mm will not matter much; but will redesign the pieces on the X-bridge so that the lines become parallel.Thanks for the tips on the fishing line and slotted motor mounts - will see how I can re-design the motor mounts for that. 

Will wrapping the string around the motor pulley twice help (or hinder)?

[Ryan Carlyle]

What people generally do is anchor both sides of the line to ends of the spool, with enough wraps to allow for the full range of travel. So the same spool is paying out line on one side and taking up line on the other side. Special spools are printed or sold to do this easily, or you can make something yourself.

In principle, you can just do a few wraps around the spool (as a "capstan") without the anchor/loop approach and excess wraps, but you tend to run into issues with the wraps walking and crossing themselves unless the line routing is really meticulous. Some people have tried approaches with grooves and multiple idlers to provide more contact area without the wrap walking issue. But I recommend just doing what you see in the photos above. Friction isn't very reliable for this sort of thing... spectra is really slippery.

[Dushyant Ahuja]

Picking up on this after a long time, as I finally built it. Went to VSlots for the frame as OpenBuilds was having a sale and I got these 600mm extrusions really cheap.

Full Album:

https://photos.app.goo.gl/DhEPv5P0F3c7Wfrl2

Bloopers:

https://photos.app.goo.gl/OlbBB1h7nwLljZci1

Thanks all for your help.

[Mark Rehorst]

Very nice!  I'm going to have to make one of those.  Do you have to manually "erase" to start a new image or is there some sort of wiper that will do that for you?  Are you running it using STL files sliced to gcode?

[Whistler 55]

Looks like you are building the same as this guy.  The Sisyphus art table.

http://www.sisyphus-industries.com/

[Dushyant Ahuja]

Mark,

There's no wiper, the way to erase is to create a tight spiral and that flattens the sand.
I've tried models sliced in vase mode and that works really well. Though these designs use the amazing sandify software. https://jeffeb3.github.io/sandify/
Sandify - GitHub Pages

Whistler,

Yes, this was inspired by Bruce Shapiro's Sisyphus Table. We saw it at NY maker faire in 2016 and feel in love.

[Ryan Carlyle]

That's a really neat build. Thanks for sharing. 

[Dushyant Ahuja]

Hi Ryan,

Wouldn't have been possible without your suggestion on using Spectra line.

Here are the STLs if anyone is interested. I plan to create an instructable, but that might take some time.

https://www.dropbox.com/sh/b1yapl7ybnkou6b/AABFRfURr-Vt-956DQVw8WFta?dl=0

[Damien Borowik]

This is a great project and build, well done!

Could you please share the STLs again? the dropbox link doesn't work anymore.

How did you get on with the Instructable?

[Mark Rehorst]

test- why are my messages getting deleted?

[Mark Rehorst]

Hmmm.  It seems that links to blogs or to github are the problem...

OK, I have several blog posts on the sand table I built.  Search for my blog "Mark Rehorst blogspot" and then search the blog for "sand table" and you'll see everything.  I use Sandify to generate a lot of patterns, but it leaves a lot of useless back and forth motion along the edges of the table in the pattern files.  My most recent blog post is on a post processing program I wrote in Perl to strip out wasted motion along and around the edges of the table.  The result of running it is a reduction in pattern file size (no big deal), and reduced time wasted along the edges of the table, so faster completion of the patterns.  That makes it much more interesting to watch because edge motion is boring.  The program I wrote removes about 99% of the wasted motion.  I'm working on the other 1% now.

I typically run the table with speed set to 500 mm/sec and acceleration at 1000 mm/sec^2. The printable area of the table is 710 x 1600 mm, so it does hit the 500 mm/sec limit a lot.  Other recent posts on the blog have been about making the mechanism run quietly at that speed.

You will also want to look at Michael Dubno's recently released, awesome RPi based sand table control software.  It will accept voice input like "draw an ice cream cone" via your phone, then it will search the web for a picture of an ice cream cone, isolate the edges, convert them to gcode, and draw the pattern on the table.  It does a lot of other great stuff, too, including text, geometric pattern generation, and blinkenlights control.  Search the web for "Michael Dubno Github sand table".

It seems google groups and blogspot are at war- when I previously linked images from groups in my blog, when I opened one of those page it complained about possible malware.  When I try to post links to my blog here, the posts get deleted.  Both are google products...  Hmmmmm.

[Mark Rehorst]

The new version of the sandify pattern trimmer, called trimify2x.pl  is posted here.  It now removes all excess edge and corner to corner motion, and optimizes the remaining corner to corner motion.

[Dushyant Ahuja]

Hi Damien,

Didn't create an instructable as wasn't really happy with the performance. Similar to Mark, I'm moving to belts.

You can find the model here:

https://cad.onshape.com/documents/f580b376c2799721dda7fa3a/w/9626f8665dba7030cdaef92d/e/d5c90f8d9deb1f4225710953

I will be making changes. 

[Mark Rehorst]

Here are some sample patterns made possible by using trimify2x.  The patterns are mostly hundreds of cycles long and would have so much edge and corner to corner motion they'd be unwatchable.  Trimify2x takes out the wasted edge and corner motion and speeds up the drawing process.

This first pattern takes two hours to complete on my table if I run the file straight out of Sandify.  When I process it with trimify2x it only takes 57 minutes to completion:

More, with a link to the pattern files (so you can grab the headers and use these patterns as starting points for patterns for your table) here.

[Steven Butterfield]

Those look amazing Mark and I love what you did with the lighting.

[Mark Rehorst]

Updates:

I messed around a lot and managed to get the noise level down to an almost acceptable level when the table was running at 500 mm/sec, but the steppers were running right at the hairy edge of their performance limits.  Every once in a while a pattern would cause the motors to stall out and it would wreck the pattern and make a terrible noise.  I came to the conclusion that stepper motors aren't very good for running at 500 mm/sec, so I installed iHSV42-40-07-24 servomotors.

It was a huge improvement! The servos have far more torque than is needed, so there's absolutely no danger of stalling them out with any pattern. The table can run faster, with much higher acceleration now- I've pushed it to 2000 mm/sec (yes, two thousand!) with acceleration at 20,000 mm/sec^2 (yes, twenty thousand!). Of course, and that sort of speed and acceleration, the ball throws the sand around and the patterns don't look as nice as when they are run at lower speeds, and the mechanism tends to make a lot of noise, but it sure is fun to watch.  At 500 mm/sec the table is very quiet.

I have a couple more tweaks to the table design to reduce the noise level further

For some reason my posts with links to my blog keep getting deleted, so I suggest you search for "Mark Rehorst Blog" and then search the blog for the term "servo". 

[Mark Rehorst]

Test post - do links to vimeo work?

https://vimeo.com/422175403

[Mark Rehorst]

It looks like Vimeo is acceptable:

speed and acceleration test

https://vimeo.com/423704831

high speed erase pattern

https://vimeo.com/424381250

[Dushyant Ahuja]

Wow. That looks amazing.

[Mark Rehorst]

A new, smaller, quieter table is about finished: I present Arrakis: https://www.youtube.com/watch?v=oN064pGug3A