2DOF Servo Arm + Stepper X-axis

[forums.reprap.org]

I don't think theres much visibility on the 'polarbot' forums so I'm posting this here. I think the key to a low-cost reprap (<200$ cost price) would be low cost hardware (<50-60$) and low cost electronics (<150$). The way to this for me is a 2DOF robot arm sliding on a linear bearing x-axis. Eventually the arm could be stationary and the workbed could be used as a conveyor belt.

I'm thinking a work area of 400mmx300mmx125mm.

Any comments?

So the Y and Z axes would be handled by the arm? The arm configuration will have severe accuracy problems, especially regarding layer height. I don't think it can be made to work with cheap components. Go ahead and run the numbers to prove me wrong, but I think you'll find you need very high angular resolution at the joints to achieve even 0.1mm accuracy.

How many rep-raps do you know that acheive 0.1mm accuracy, repeatably? I would bet, almost none.

In fact as you say the angular resolution is critical. I might switch to worm gears (if I can find some cheaply) or threaded-rods driven by stepper or continuous-rotation servo motors if I need more acuracy then the current servos will allow.

Each servo I am using is 0.35 degrees. Ideally it would be ~0.1 or 0.05 degree but maybe with gearing I could get that. 0.1 degrees over 300mm is ~0.5mm and 0.35 gives 1.3mm which is a bit more than I'd like.

In the 'real world', robot arms are used in many cases to move light loads (like our extruder head) extremely accurately (sub-micron precision). Gantrys are used where they are cheaper (larger build areas are not practical robot arms) and where the effector is heavy.

... if you want high precision with cheap servos, then search for 'parallel scara' in the forums: [forums.reprap.org]

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

jonnycowboy Wrote:
-------------------------------------------------------
> How many rep-raps do you know that acheive 0.1mm
> accuracy, repeatably? I would bet, almost none.

Almost all of them can do better than .1mm

Doing that at 30mm/s+ is another issue entirely.

The reprap mechanics are very accurate from stop to stop, just have some issues at high speed.

---

www.Fablicator.com

Andrew Diehl Wrote:
-------------------------------------------------------
...
> Almost all of them can do better than .1mm
...

Especially the Z axis, where it's most important.

I think something which is lacking on this site is a standard set of "calibrated" parts that each model could print out. Repeatability is the thing I think is missing for the transition of repraps from hobby to commercial-quality status.

Dale, why would you say Z-axis is the most important? Because the model is sliced? I think that if you ask any engineer he'll tell you that all three axis are important but that for each part that you produce, the desired (required) tolerances will be different on each feature.

From what I read on the forums (I am waiting on receiving the Printrbot so that I can print out the brackets for my robot arm - so I don't have first-hand reprap experience), main problems are surface 'roughness', repeatiblitity, and accuracy at speed. I don't aim to fix the first but I do aim to improve repeatibility and accuracy (especially at speed).

Also Andrew, I beleive you are talking about precision, not accuracy BTW.

Good comments, keep them coming!

Since the AX-12 servos has low angular resolution (0.35deg), I'll be switching to the BOSS servos from Biomimetix which basically use the open plans "supermodified" servos with metal gears and a coreless motor. The resolution should then increase to 4096 "steps" (12-bit) so ~0.09 degrees. Since the servos are closed-loop, with rigid arms I should be able to get to around 0.4 or 0.5mm.
The other advantage of these servos is that if I decide I want more accuracy, I could switch them 90degrees and use them to drive worms (20x or 50x slower) to increase the precision by an order of magnitude.

Actually, originally I wanted to use worm gears but they are quasi-impossible to find cheaply enough to drive the cost down.

With regards to the Gada prize, they don't mention anything about resolution/precision or repeatibility (I would be aiming for 0.5mm repeat) so that's why this design could be cheaper and still meet the prize requirements.

thanks!

Uh, I are an engineer... Z axis is more important because you have to have very accurate control of layer thickness before you can control the width of the extrudate. You see, it gets squished against the previous layers, so if the Z is wrong, it will result in amplified errors in X and Y.

Andrew was talking about accuracy. Once a RepRap is calibrated, 100mm is 100mm ± a lot less than 0.1mm every time, if one uses manufactured pulleys instead of printed. Very carefully printed pulleys can probably achieve the same.

For 0.5mm repeatability, you'd better plan on using a huge nozzle by RepRap standards. Ø2mm or probably more. Otherwise your layers won't stack well or even adhere.

How long have you been out of school?

Oh, I absolutely meant accuracy (and only in X &Y) and the Z axis accuracy is even better. The precision is about an order of magnitude smaller on my machines (maybe 2/3x that on many repraps.)

And the big thing is, this is all static. With a system that has practically zero backlash. With mechanics which move in a single axis. Over less than 8 inches. Luckily, the dynamic response of the standard reprap system is pretty good, since it is extremely simple, and the axis are independent.

You are talking about taking two cantilevered beams in series controlled by servos and precisely positioning the tip over 16in of travel. Just the stiffness required to prevent massive vibration of the print head would be extremely high (expensive.)

I'm not trying to say don't try, but I am going to tell you to start in a small incremental way. That way if it turns out things aren't going well you aren't out a pile of $$. I've worked with inexpensive ($100-$200) servo powered arms in college, and they are nowhere near capable of reprap's dynamic accuracy demands from my experience.

Edited 2 time(s). Last edit at 12/10/2011 05:13PM by Andrew Diehl.

---

www.Fablicator.com

Hi Andrew, thanks for your reply.

To be honest, I'm pretty impressed by the numbers you quoted!! I didn't think the repraps were that accurate.

You mention the issues at speed, I expect my arm to have the same problems, especially with the high inertia of my system. I found a cheap worm/worm gear combination that I could use, it would bring my rotational resolution down from 0.09 to 0.00225 degrees. I think that will be good but it will be dreadfully slow moving (actually, that should take care of my inertia issues!).
[www.hobbyparts.com.au]
I do plan on counter-weighing my beam (rigidly) so that it will offset the torque arm but obviously the joints will have to be strong. At least with my design the motors are located at the base.

I already have a 150$ 6-servo robot arm (the Trobot 3 which was on Kickstarter this summer), it's pretty impressive but as you say, not very accurate.

Don't worry I won't spend too much on this robot, I already put my money into the Printrbot and the Microrax beams so at least I should have some capability to reproduce a few different arm configurations cheaply.

Well i haven't been on the forums for a LONG time, due to school. Its good to see that people are fooling around with the robotic arm idea still. If you don't know what I'm talking about search for "robotic arm" on the wiki. I just ordered a printrbot, so I can actually make some of the parts for the robotic arm that I couldn't before. I think the threaded rod in this design makes it very accurate. I have tested the horizontal "wobble", and found that, surprisingly, I can see no vibration when the arm moves, even at high speeds. I will have to work on this some more....

I am just developing a SCARA robot here which may add a little to the debate. I am just printing off the first parts at the moment. I hope to get it working in a couple of months.

The accuracy and resolution of these robots will be less than a cartesian robot, but with careful design it should be good enough for FDM.

It is possible that this type of robot could be the first to self replicate. The SCARA I am designing should trounce the mendel for the percentage of parts produced by itself as it only has one linear rail which is currently beyond the capabilities of a mendel to replicate, but doing it this way throws up a whole new set of problems which the community will have to overcome one by one as they are doing with the mendel.

If you are interested, the arms are 135mm long and geared at 1:5 (8 tooth to 40 tooth). I intended to use 400 step per rev motors, but as I have a few 200 step per rev lying around I will probably use those to get it moving. These should achieve a resolution approaching half of a mendel, but that is still OK for printing parts. There are issues with Z height accuarcy if the two arms are not perfectly aligned which cannot be taken out by simply packing the flat bed, but I might look at a software solution for this.

If you don't think that these types of robot are capable of doing what we ask of a mendel, can I just say that I have used SCARA robots in industry for surface mounting electronics (We used them for putting first generation mobile phones together) and they were very accurate both for surface mount and through pin components. A smaller machine could be even more accurate if engineered well or had parallel supported arms.

heres a video of one in action to show what they can do.
Adept