GUS Revision

Hi All,

Some of you may have heard of or seen the GUS Simpson printer from a couple of years ago. Personally I think it's awesome. However it suffered from some issues. I'm currently working on a revision of the GUS, trying to fix or improve anything I can think of.
Just wanted to start a thread here for anyone who wanted to collaborate, suggest ideas or improvements, or point out where I've done something stupid.

I have a few major goals.
-User friendliness
-Usability
-As many standard or readily available parts as possible, and as few different parts as possible.
-Improve mechanics and printability
-Make it cheap, make it scale-able. Using only rotating joints is a big plus to the GUS style printer.

I have a bunch of improvements I've implemented or plan on implementing, but more is always better. Right now I'm trying to decide on the build area for the printer, since the GUS style printers have an a typical build volume (approximately a trigonal base pyramid with bulged sides) it's not as simple as saying "200 mm cubed". Besides, who uses all of that build volume (especially the Z?) to begin with?

The design will be fully parametric and open source. I'm hoping I'll be able to slim things down to ~5 parameters that the end user can tweak to customize the machine to their needs (desired build volume, how large of a printer they have available to print the parts), input the details of the hardware they want to use (if it's different from what I'm using), and everything else will be calculated to fit automatically.

Feel free to ask questions/make suggestions. I'll document as I go if people are interested. If not, I'll just release the files when done.

EDIT: Should this be under deltas? I'm not too sure, if so could a mod move it/someone tell me how to move it? Just had a closer look at the delta section and noticed Simpson was listed there. Whoops.

Edited 2 time(s). Last edit at 07/03/2017 04:27AM by Trakyan.

have you seen the improvements made to it on the conceptforge forum. I have built one of the improved one s 1 year bak but have yet to commission it properly.. We can collaborate if you want .

the biggest head ache is the string walking error. i just love the way it moves and i wish it could be a practical printer..

Edited 1 time(s). Last edit at 12/29/2017 02:06AM by ekaggrat.

Yep, I've read through just about every bit of material I could find on the simpson online a couple dozen times haha I love this printer design. The most recent revision I've seen is jfriessen's one on conceptforge. I had concerns about string walking too, so one of the changes I decided I'd do in my revision is to use large pulleys to eliminate or at least minimize string walking.

Also, I'd love to collaborate! Haven't had much interest so far from others but I can understand why. So far my little project is just smoke but the design is close to completion, at least for the first prototype. I hope to print it by the end of the month but knowing how my timelines are a bit 'flexible' and I tend to get bogged down in the design details it could be until mid/late feburary.

My design is fully parametric and pretty simple to use. It's written in OpenSCAD and you can just give it a list of your hardware you want to use, the build volume you want (either a base size and end up with gus's "natural" print volume that's more or less a pyramid with bulging sides, or a cylinder with a radius and height that most people and slicers are more familiar with) and it calculates everything else for you like clearances and stuff.

Anything you'd like to see in a new GUS ?

I will be printing Gus 2.5 arms soon. I think the string walking issue is only in the 1.5 version . But I will report back as soon as I am done. I think a better approach would be belts and it has been done by rustynoob but even he left it half way though . So I don’t know if that would a better aproach . I got my 1.5 arms printing only recently but due to string walking a cube is printed with slightly curved walls.

If Gus can be solved using belts . It would be a very good candidate that can be injection milder and made into a mass produced machine . It’s just too cool when u see it working. I wish someday it becomes a proper printer.

Injection molding gus would be complicated due to the herringbone gears and all the nooks and crannies of the gears, it's really tailored to take advantage of additive manufacturing. As for belts, I considered it briefly but routing the belts is a bit difficult I think. The belts need to terminate around the same radius pulley as they leave from, and this will need to be offset from the departure point so the belts don't overlap. This offset would introduce triangulation error similar to the string walking as you've got two belt widths on the drive arm side, from the departing end of the belt and the returning end, but only one width of belt where it loops back on the driven arm. This also means that the belts must "skew" for a lack of a better word, where one side of what should be a straight belt path is shifted sideways, but the ends of the belt stay orthogonal to the belt path instead of rotating to match the new belt path.

As for 2.5 solving the string walking, I'm not sure, there still seems to be very little space between the drive pulley and the first redirect, giving you triangulation error. However, there is a higher mechanical advantage from the bigger gears and more pulleys, and he says he's gotten +-0.1mm I think which is as good as most other printers you'll find.

to be clear they are cycloid gears.. what i meant was eventually once it runs smoothly it could be simplified to the point of being either injection molded or resin casted.. if there is one cool printer i want on my desk to show off it would be gus..

as far as error is i can report back once i get the new arms printed.. even the current arms which i have can take out good prints ..

u can see it here :

instagram 1


instagram 2

instagram 3

Edited 1 time(s). Last edit at 01/10/2018 09:06PM by ekaggrat.

I know they're cycloid gears, I think the issue would be more with the face they have a helix, and even more difficult, a double helix (herringbone). Things get even trickier with all the hollows and enclosed areas. I have a few ideas to replace a herringbone gear but keep it as a self aligning gear that could be more injection molding friendly.

From your post/picture there and what I've heard from others, calibration seems to be a royal PITA, I've got a few ideas for auto calibration that could help. I still need a machine in front of me so I can code and test if the method actually works, but it's sound in theory as long as there isn't too much slop, the reference I'll be using (the bed) is machined to close dimensions and the basic assumptions about the geometry is correct (shoulder and wrist offsets are equal, the actuator is linear). The calibration routine should in be able to figure out steps/mm for each arm and (as a bonus side effect) if the arms aren't all quite in the same plane like if one of the shoulders is drooping a little.

i created a virtual model of gus in grasshopper 3d to see wwhat causes the domical effect ... seems that it happens if you get the max arm lengths wrong in the firmware that effect is created.. so i will be trying to add a offset in the arm length in the firmware ( i amusing marlin as i am using cheap ramps for now ) and see if i can get that corrected.. will report in a day or two.

which firmware would u be using?

I plan to auto calibrate max arm lengths as well. Currently I plan on using smoothie as the cheapest board I can find with built in drivers and software settable current is the mks sbase.

Overall I'm trying to eliminate the hassle I hear people have had to go through with calibrating Gus.

the mks is ok for testing i guess but people dont have good experiences with it . i would go for a re-arm with a ramps or a proper smoothieboard..

I'm going budget to start with, and I'm not fussed about 32 bit, it was just the cheapest board with software settable current for the steppers, so rearm doesn't really work for me.

i got gus 2.5 up and running using the latest files and marlin.. NOw i need the damn thing calibrated.. A major pita.. i am thinking on investing in smoothie or moving to teacup.. marlin is a major PITA to insert any code in it..

instagram

I got some smoothie clones for relatively cheap.
I've finished the design stage and now I just need to find the time to print all the parts and assemble, from there is on to writing the firmware and calibration routine. The calibration routine will have to wait until I get the bed CNC cut out of aluminium (aka, wait until I finish designing and building the cnc to do it). The whole project might be getting shelved for a while until I finish the CNC to be honest as it's an easier project, and the enormity of designing the printer and working with the firmware is fairly daunting for my first project.

The printer is as of yet untested, so I can't guarantee my design works, but I'd be happy to share the source files with you or some STLs if you don't want to mess with OpenSCAD (I haven't cleaned the files up at all so you'll have to do a little bit of digging to alter some parameters), if you'd like to have a go at building it. I made some changes to move the motors off the arms and onto the base, so they're stationary. The change is fine in theory, but it's untested and could be flawed.

i think right now it needs a better way to calibrate.. the hardware verison 2.5 is just fine and very easy to put together. The hardware has a lot of compliance due to the geometry which limits the speed. I can have a look at your files and suggest changes. I am bad at openscad as my background is more of design than programming. I think a better way to making the files would be freecad or inventor.

i hope you get your gus running soon to test the calibration routine or i could do it for you once i get a smoothieboard ( clone ).

My version should also help with rigidity since there is less moving mass, and most of the bearing joints (except the effector) are preloaded from the tension of the string.

I spent quite a while making it in OpenSCAD so I'm unlikely to scrap it now . I can export it as a CSG if that's more useful to you. I wanted to use a free and open source software to design it, so that didn't give me too many choices. FreeCAD is a bit clunky and awkward to use, so I ruled that out. OpenSCAD let me do everything in one tool, whereas doing something like cycloidal gears is difficult in most other CAD programs. That and if I made it in another CAD program, anyone wanting to customize the machine for themselves would have to be experienced with CAD and know what they're doing. With my OpenSCAD script they tell it the build volume they want, what hardware they're using (bolts, bearings, motors) and it takes care of everything else. I don't think it's possible to constrain the machine this thoroughly in CAD programs like inventor because all of the possible collisions and constraints occur in different positions of the actuator. At the very least it would be very messy.

For the auto calibration you'd need a conductive build plate milled to decent spec. From some test calculations the dimensions don't need to be super accurate, a couple of mm shouldn't make too much of a difference to the resulting steps/mm. The allowable error should actually grow for bigger machines, it's small ones that will need to be precise. That being said, I'll have to test to see if it even works, and to do that I'll need a mill to cut the aluminium.

ok cool.. i can live with openscad.. I knew the base needed to be accurate so i built it out of 2mm laser cut steel sheets.. that way the hole spaceing is super accurate and the plates are super flat. Also as a bonus they are conductive.. In hind site i think a better approach would be carbon fiber 4 mm plates..

i am leaning towards teacup now till i get a smoothie. But i will have to test it on my delta and learn all the details of it before i change it for gus.. overall it looks a lot cleaner to change than marlin.

right now i am fighting the bowl shape of my 0 layer which i cant get flat in spite of modifying a lot of parameters. Also i calibrated the steps per mm very accurately and got the MAX arm lengths to be equal. But even after all this i get a bowl shape with the corner near the left front arm lifted up.. Completely stumped for now

Are you using manual leveling or using a touch probe?
I'm hoping to get rid of all those steps/mm and max reach woes and have it all auto calibrated. Downside is it needs a specially shaped, accurately milled conductive plate (so no carbon fibre). The actual base/frame/body of my design is 3d printed, it's just the build plate that'll be metal. The base had numerous features on the top and bottom surface, it could be milled but would need multiple setups and it just seemed easier to 3d print. Besides, if I'm relying on the arms to 3d print accurately, I might as well say the same for the base. Besides the base is probably the easiest thing to correct for (just measure the three distances between the corners and input them into the kinematics).

right now manual leveling.. the prints i have are very acurate as the mother printer is acurate and not a crappy printer.. The thing is strings are not super acurate and tend to stretch . so if you change to belts it will be better, as then the steps per mm calibration will not be required which is another headache and depends on the accuracy of the vernier used.

i will be adding a probe and get bed points soon .

Even if I switched to belts, there would be some steps/mm calibration issues due to how accurately the hole for the pulleys on the arms are printed. Even with a good printer you're looking at +-0.1mm and that error is magnified by the lever of the arm. Belts would solve the string walking issue but I have some ways to potentially solve this or at least minimize it to where it doesn't make an appreciable difference. The reasons I'm sticking to strings are mainly twofold. For starters, I'm on a strict budget and I've really had to downgrade my plans for this and another project of mine because of it. Secondly, with my arrangement that moves the motors off the arms, the belts would take up eight times their width along the bolt. That's 24mm with 3mm width gt2 belts, more if you account for the idler bearings not just the belt width. It doesn't seem like much but the remaining space is split between the arm and shoulder, so you start needing close to 90mm bolts to get a decent arm thicknesses for rigidity. These sorts of bolt lengths are quite clunky in a relatively small machine and require larger shoulder offsets which means less rigidity. If I moved the motor back onto the arm this wouldn't be as much of an issue (I'd only need one belt width if I also split the arm so I could run the belts in the middle of the gear like the original GUS Simpson did with the strings) but I'm reluctant to do so, at least in this design. I'm not convinced switching to belts will be the silver bullet for steps/mm issues, so I'm looking for a way to auto calibrate it.


The more I dive into its design, the more I realize the GUS Simpson isn't ideal as a small printer, its advantages really shine at scale, and the proportions start to look nicer too. Right now with my first planned prototype, the effector is a bit over 100mm in height below the wrist of the arm for a build volume that's 100mm tall.

The design is currently on hold until I can find or build some way to CNC mill an aluminium plate for the printer, and acquire some more PLA for all the parts. This should hopefully be soonish as I'm almost done with the design for something like the MPCNC and should have some more disposable income after the next couple of weeks. Who knows, I might even be able to incorporate these projects into one of my uni courses.

Why no Pics?