FirePick Delta is open source. Why nobody built one?

Hi, I am a fan since the begining of the firepick-delta for it's way to avoid linear rods. Now after a few years I've checked back if anyone has built one and can't find anything. I am posting this in delta forum so you,the delta gurus, can tell me your opinion.

Is, as of now, the firepick delta a good option for a delta printer? On their webpage it mentions Marlin for compatible software so it could be built to only take advantage of the 3d printer for now and maybe adapted later for the pick-and-place stuff.

Thanks.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

Link?

mmkey, thought everybody knew about it ... my bad:

Website: [delta.firepick.org]

Few videos:
SmoothieBoard running a FirePick Delta
FirePick Delta 3D printing
axis-probe-calibration
FirePick Delta PnP & 3D printing Robot making parts for itself

And a couple of images:


And GitHub repo:
Firepick Delta GitHub

Edited 2 time(s). Last edit at 10/07/2015 05:25AM by realthor.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

Cool design!
But from a more technical point of view, it has too many mechanical links ( play/backlash )
The arm length is limiting position-accuracy or print height
Too many printed parts.
A common delta has much less parts influencing the inertia ( moving mass/print speed )
-Olaf

Yes, there seem to be quite a few mechanical parts but one advantage is it doesn't depend on the frame, on the perfect alignement of the towers, basically the whole printer is the top center hub containing motors, electronics and the arms with the wheels; you could detach that from the frame ceiling and walk away with it, bolt it to a wall or ceiling, etc.

On the other hand this is created to do so much more than 3d printing but I aggree, from a strict FDM pov there are quite a lot of parts.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

Quote
realthor
Hi, I am a fan since the begining of the firepick-delta for it's way to avoid linear rods. Now after a few years I've checked back if anyone has built one and can't find anything. I am posting this in delta forum so you,the delta gurus, can tell me your opinion.

Is, as of now, the firepick delta a good option for a delta printer? On their webpage it mentions Marlin for compatible software so it could be built to only take advantage of the 3d printer for now and maybe adapted later for the pick-and-place stuff.

Thanks.

Well, they did a beta kit offering back in about April, and now there are around 60 or so of them in the wild, and I have one. They have had limited success with 3D printing so far (a few people have it working), and at the moment the focus is on getting SMT pick and place working, so I would say it is not ready for 3D printing yet. Marlin isn't fast enough running on an Arduino to get it to print very well, so there is an ARM based Motionboard in the works that runs the Smoothie firmware. At this point I would say the project still in a heavy development phase.

Note that this delta has a pretty limited Z height of about 100mm, unlike the linear deltas which can be very large in this dimension.

If you check out the Google group for the project at FirePick Delta, you will find that there is quite a bit of activity, and it is a good place to ask questions.

BTW, the backlash is not as bad as people claim it might be, since it uses zero backlash ball joints.

---

Mike Anton
[manton.ca]
[laserlight.wikidot.com]

The frame's stiffness is very important for this setup.
Four vertical/ parallel extrusions are not the stiffest platform.
If these guys had guts, they´d build the frame like a pyramid...
-Olaf

Quote
o_lampe
The frame's stiffness is very important for this setup.
Four vertical/ parallel extrusions are not the stiffest platform.
If these guys had guts, they´d build the frame like a pyramid...
-Olaf

Don't knock it until you try it. There is little to no load on the actual frame, compared to a linear delta, and this frame is actually very stiff. Those aluminum brackets work really well.

---

Mike Anton
[manton.ca]
[laserlight.wikidot.com]

somebody built a similar machine but said he had a lot of calibration issues.

[hackaday.io]

Quote
ekaggrat
somebody built a similar machine but said he had a lot of calibration issues.

[hackaday.io]

Yes, calibration has been problematic, and that is where a major part of the development effort is currently being focused. Many of us are patiently waiting for a proper method to calibrate the machine, myself included. There are even more geometry issues with this type of printer, compared to a linear delta.

---

Mike Anton
[manton.ca]
[laserlight.wikidot.com]

I have read before about the icepick delta as an alternative but its development stalled. Then I forgot about this type of delta for a year or so until I saw this one made by a romanian guy which I find very similar and quite bold. Of course it's on the cheap side but it's frame is a pyramid made of threaded rods: [www.youtube.com]

It looks to me like the first prusas but no linear rods which is why am am keen on this design.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

A printer is fast , easy to build , very difficult to calibrate and requires a rigid frame . I know from my experience

https://www.youtube.com/watch?v=X9mgDQ-nGEg

https://www.youtube.com/channel/UCuWyirjuE_Fe0zXUW4uCxDg

https://sites.google.com/site/3dneludeltaen/3d-printer/delta-robot-nelu

@altnelu
cool work.. what ratio are you using? why is it difficult to calibrate? I have been long tempted to make a delta like this but have not found any successful examples!

Firepick Delta has some nice calibration routines (i've linked in my previous posts) bult into the code and there are several other calibration procedures they only need to get into the main Marlin code.
One cool calibration which have found is this one: [youtu.be]

One can also look into more complex stuff like "Heuristic delta calibration using simulated annealing": [github.com]

I am sure there are more and Delta guys maybe can help more at this point.

Edited 1 time(s). Last edit at 10/29/2015 11:02AM by realthor.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

You could also use RepRapFirmware, which has a fast and accurate least squares delta auto calibration routine. It needs only one set of between 7 and 16 probe points to fully calibrate a delta if your initial endstop corrections and delta radius are less than about a millimetre out. If you have larger errors, you might need 2 or 3 iterations to do the initial calibration. The video is at [www.youtube.com].

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Does the reprapfirmware support delta mechanisms like the firepick delta?

Quote
ekaggrat
Does the reprapfirmware support delta mechanisms like the firepick delta?

Good point: RRF only supports linear deltas. The segmentation-free approach used by RRF might not be easy to adapt to Firepick because I suspect it would involve trigonometric calculations that might take too long. However, the fast least squares auto cal algorithm could quite easily be adapted to other kinematics and firmwares.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

If the knees of this delta were springloaded ( trying to stretch out the arm ), you could use fishing line/spectra between steppershaft and effector to control it linearly.
No messing around with huge pulleys and closed belts. More or less like the "linear SCARA arm", I try to build once.
-Olaf

Are we still talking Firepick Delta right?

Can you point us to your design? There are geometry errors when using fishing line because of string walking effect. Unless you have a sort of "spectra drive" you will have to compensate in software for the errors, something that is not done anywhere as far as I knowbecause it is difficult to guess the specifics of each design. Even the regular deltas with fishing line have some error because the string walks from side to side on the pulley while its attachment on the carriages stays fixed. Bbut due to the length of the towers, it is insignifiant I guess. The shorter the distance between the walking spectra on the pullley and its other attachment point the greater the errors in the print. There is not much emphasis nowasdays on spectra as the prices for industrial-grade equipment went down .

I am building a SCARA with fishing line and my way of compensating for string walking is to have it walk on both sides at the same time to prevent triangulation errors but that makes the assembly quite voluminous.

Edited 2 time(s). Last edit at 10/30/2015 09:35AM by realthor.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

My remarks was aiming to altnelu's printer, because I see a lot of parallels with a single arm SCARA.
My linear SCARA is still in concept stage.
[forums.reprap.org]

We can discuss the string walking effect in my thread, if you like.
-Olaf

I see someone has mentioned my project here already [hackaday.io]

TL;DR: calibration is very very difficult. Error propagates in non linear ways in 3 space. I spent days calibrating it, in the end I nearly got it spot on, except everything was 4% too short in the Z direction. The arduino can handle a maximum of 40 segments per second. The resolution is bordering on too low with 1.8 degree stepper motors, using a pulley ratio of 1:14, with 160mm long upper arms and 300mm long lower arms. The firmware (marlin) measures the speed at the motors, not the end effector meaning all speeds, acceleration and jerk settings were off by a factor of about 3.5. Thats probably about the worst of it.

If you make a solid frame, 0.9 degree steppers, a 32 bit controller, and use very carefully measured parts and and accurate assembly, I think you could make it work.

I really like where Altnelu took the design. It looks really awesome and minimalistic. I like it.

EDIT: apologizes for the many edits, I just keep thinking of things to add to my post

Edited 5 time(s). Last edit at 11/01/2015 02:14AM by TTN.

Quote
TTN
If you make a solid frame, 0.9 degree steppers, a 32 bit controller, and use very carefully measured parts and and accurate assembly, I think you could make it work.

Quite dissapointing , with that list you can get cheaper with cartesian.
I thought that with a non-linear delta one could escape the need for very specialized parts. What kind of tolerances are you talking when you say "very carefully measured parts"? Is it still doable with off-the-shelf parts and 3d printed parts?

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

@TTN ... any detailed information regarding the 4% deviation along the z-axis would be really interesting especially the strategy to backtrack measured deviations to the root cause.

FYI my frustrating setup ... arms are directly mounted on the stepper motor shaft 0.9deg, 16 microstep, about 40gr weight endeffector just for PnP SMD 0402, all 3 motor motor are mounted on a single printed piece of PLA, 90mm upper arm, 236mm lower arm ... step resolution in the near of 100um ... accuracy error 2500um at worst corner of an 80x80mm square with z=70mm moved down from zero reference (upper arms horizontal).

Edited 3 time(s). Last edit at 11/07/2015 05:24AM by lutz_dd.

Quote
lutz_dd
@TTN ... any detailed information regarding the 4% deviation along the z-axis would be really interesting especially the strategy to backtrack measured deviations to the root cause.

FYI my frustrating setup ... arms are directly mounted on the stepper motor shaft 0.9deg, 16 microstep, about 40gr weight endeffector just for PnP SMD 0402, all 3 motor motor are mounted on a single printed piece of PLA, 90mm upper arm, 236mm lower arm ... step resolution in the near of 100um ... accuracy error 2500um at worst corner of an 80x80mm square with z=70mm from zero reference (upper arms horizontal)

There wasn't any calibration manual to go to, so most of the calibration sympton -> causes I had to discover by trial and error. I spent a few 14hour days trying to get it calibrated to that accuracy level, and that was as close as I managed to get. 4% short in the z direction. Perhaps there was something wrong with my setup, maybe it is something wrong with one of the constants in my firmware, I don't know. In x and y it was accurate enough for my liking at the bed height and at z = 200mm, just not the z height. I never discovered why though.

The error you are experiencing, is it in the xy directions? Check and remeasure all your measurements. Take several measurement, as it is very sensitive (eg +-0.25mm makes a difference) and error seems to accumulate.

The worst case corner (2D, xy) comes along with an unquantified error in z-direction of about 1.5mm. All other corners do not express this big error. I have measured multiple times alle parts and positions, but can't catch the issue. It came in with the new setup I made for better accuracy (single mounting frame for the motors vs. 3 separate frames for each motor ... the previous setup had an error of about 0.8mm)

reference square=80mmx80mm .... measures side lengths of the square are 79.5mm, 79mm, 77.5mm, 78.5mm

But while writing this response I think ...
... the 1st test should use a triangle rather than a square as starting point. Will check it tomorrow.
... maybe my 3D printer is not that accurate as I expect ... need to check the new printed single motor frame

Edited 2 time(s). Last edit at 11/07/2015 05:59AM by lutz_dd.

I was getting similar sort of dimensional issues like you describe. Its a matter of getting measurements measured right and entered correctly. The pulleys are really sensitive to what circumference you enter and the homing angles on the arms are critical too. They are the ones I struggled with most. Really make sure all the arms home at the same angle. Its critical. As for the arm pulleys, I put mine in a lathe and made them perfectly round.

OK ... it seems the single-frame is for one motor 0.4mm off when I measure the triangle of the mounting holes. The motor position have the same error, because the same mounting frame is used ... just 120deg rotated. This needs to be checked in the CAD system (FreeCAD) at the laptop.

Anyway 0.4mm is pretty heavy error at an unepxected place.

And if one thing goes wrong Murphy is waiting for you at the next corner ... harddisc damaged, laptop does not boot. That take a while to recover.

Has someone tried to make a virtual model of his setup? To verify if it's the algorithm or the mechanics that cause problems.
It takes a rock solid mathematical base, before you can iron out mechanical errors.
-Olaf

There are a number of very bright people working on solving the calibration problems of the FirePick Delta. The best place to find out about these is in the forum on Google groups: https://groups.google.com/forum/#!forum/firepick

---

Mike Anton
[manton.ca]
[laserlight.wikidot.com]

The function for the inverse kinematics that I have been using has come from here, as mentioned on the hackaday project log, but I forgot to mention the source in the code: [forums.trossenrobotics.com] My apologies.

On page 3 there, someone claims there are errors in the inverse kinematics function:

Quote
Mecch
I found some typos in the formulations

1. In the first term of Equation (4), x2*x ==> should be -x2* x, also the first term in Eq. (5), x3*x should be -x3*x

2. The last equation, the coefficient of the z term : 2(a1+a2(b2-y1)......)*z should be 2(a1*b1+a2(b2-y1)....)*z

In the text preceding Eq. (7), "From (4)-(5)", should be " By Eq (4) and Eq (5), we can express x and y in terms of z to reduce the number of variables.." to make it more clear

Though others claim it is correct, and that the function presented by OP is correct.

EDIT: fixed link.

Edited 3 time(s). Last edit at 11/08/2015 05:13AM by TTN.