Delta 3D printer assistance

[Leon Grossman]

Someone over on the main [ps1] list suggested I check here.  I've particpated there and feel like I could ask the question without being *that guy* that drops in and asks for help.

Right now I'm completely stumped and any ideas (or someone willing to be an extra brain if I lug this thing down to PS1) would be welcome.

The printer is a Boots Industries v2.5 from www.bootsindustries.com.  I made the mistake of looking at the preliminary work they had done and purchased before anybody had one in hand.  The short summary is that the BI guys had a reasonably solid design hampered by very, very, very poor China quality control.

After replacing a vertical tower strut that was bowed and a horizontal strut that was 2mm shorter than the rest, I have a machine which is sub mm consistent on measurement tower to tower and 1mm difference on the diagonal measurements.  tower X to Y.  X is the lower corner, Y is the upper corner of the other tower.

A to C 1020mm
A to B 1020mm
B to C 1020mm
C to B 1021mm
C to A 1021mm
B to A 1021mm

Diagonal rod radius:
A1 442.5mm
A2 442.5mm
B1 442.5mm
B2 442.5mm
C1 442.5mm
C2 442mm
(NOTE: I swapped C2 and A2 and it had no effect on the skew I am seeing)

I've converted the machine to an all metal hot end and belt drive.  As far as I know, other than a larger than I'd like play in the hot end carriage, there are no geometry problems of significance in the system.  The output looks amazing but there is unacceptably large skew in the part.  I could see 0.1mm of inconsistency along the length as acceptable but 0.5mm is huge.

The full data set is here: https://www.dropbox.com/s/q7ad5ymndqukk0l/Calibration.xlsx?dl=0

But an example is here:

0 Degrees	Corr 0	A corr -1	a corr +1	b corr -1	b corr +1	c corr -1	c corr +1
Part Orientation	0	0	0	0	0	0	0
Rod length	445.102	445.102	445.102	445.102	445.102	445.1	445.102
Radius	264.45	264.45	264.45	264.45	264.45	264.45	264.45
Measurement Direction
Part Length	100.03	99.74	100.14	99.68	100.39	99.92	100.01
-50	3.41	3.59	3.44	3.39	3.61	3.35	3.48
-25	3.41	3.7	3.46	3.52	3.58	3.34	3.61
25	2.94	2.92	3.02	2.91	3.06	3.06	3.12
50	2.9	2.8	2.86	2.86	3.01	2.93	2.89
Max Deviation	0.51	0.9	0.6	0.66	0.6	0.42	0.72
Max difference from nominal	0.41	0.7	0.46	0.52	0.61	0.35	0.61

I haven't had a chance to do an analysis of the data I collected yesterday where I ran a part which is 3mm wide, 3mm tall and 100mm long arranged in a star around 0 at 0, 30, 60, 90, 120 and 150 degrees.  That's why there is no zero measurement in the table.

At this point, i haven't tried correcting any of the individual tower rod radius values and I probably tested this horizontal radius tower correction at a too small value to pick out what I need to see.

I'm not a novice at motion or 3D printing but I know nothing at all about delta mechanics and am about ready to leave this machine on by the trash bin (OK, I won't really, I'd sell it at a jaw droppingly bad loss) and by a Creator 3D cartesian printer I can understand!

So, anything you can do to save me the shame of failure would gain my eternal gratitude.

Thank you,
Leon
Leongr...@gmail.com

[Ryan Pierce]

My $0.02:

Trying to print a part with that small a footprint and that much height relative to it seems like a recipe for disaster. With my 3D printed bagpipes project, I'm sometimes having problems, and my parts' aspect ratios are nowhere near as extreme as yours. The part is really acting like a lever, and forces the print head applies to it get magnified where it touches the bed. If the part shifts or flexes while printing, then even if the head motion is correct to within 0.00000001", you will still get a wonky part.

If it were me, I would take the filament out of the equation entirely. I would try to find a way to play the gcode to the printer with no filament loaded and "air print" it but pause the print head at certain adjacent vertices of the outside of the rectangle at various Z heights. Use a micrometer gauge on a mast perpendicular to the bed to measure the position of the print head. This will tell you if the print head is making one dimension consistently longer or shorter as the Z height rises. If so, then either the gcode is bad, or something is wrong with the machine's alignment and/or positioning. If not, then I'd highly suspect the part moving during the print, and I'd consider time lapse video zoomed in on it as a way to observe possible changes.

Ryan

--
You received this message because you are subscribed to the Google Groups "CNC Build Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cnc-build-clu...@googlegroups.com.
To post to this group, send email to cnc-bui...@googlegroups.com.
Visit this group at http://groups.google.com/group/cnc-build-club.
For more options, visit https://groups.google.com/d/optout.

[Leon Grossman]

Ryan,

* This is a part my other printer would have no problem with.  

* My other printer is a cartesian and this is a delta so maybe I’m expecting too much.

* The skew exists on larger parts (like a 100x100x25 hollow square with 4mm thick walls.

* I’ve used Simplify3D and Slicer (using the manufacturer’s distro)

* I used the original firmware that came with the machine and my own version of Repeater.

* Slow motion video distinctly shows the end effector moving and not the part.  I have the glass clamped to the build surface and the part has to be scraped off with a spatula.  Any bad motion is strictly in the mechanics.

I’m not sure if I follow your micrometer suggestion as all three axes are moving for pretty much any linear motion so I’m not sure what I should measure.

That said, I agonized over my options for a couple of days.  Almost tossed this printer aside and then decided to double down instead.  I think the issue most lies in a defect in the carriage/rod/end effector system.  I’ve got linear motion rails and super precise magnetic carbon fiber rods on the way.

I’ve already got the motion system removed from the current printer in preparation so I can’t run your tests but I will explore this when I’ve got things back together 

Thanks for putting in the thought because I’ve been at wits end given my extreme attention to detail on getting the system setup correctly!

Leon

[Ryan Pierce]

I was talking about keeping a micrometer fixed to the bed to measure absolute position of the head relative to the bed. Something that can give you a cartesian coordinate regardless of what the three delta axes are doing.

But if you're definitely seeing the head move incorrectly, then my bendy plastic theory is dead.

I'm not that familiar with gcode for delta machines. Does the gcode use Cartesian coordinates, in which case whatever parses the gcode and runs the machine handles the conversion to delta coordinates? If so, it may be useful to send gcode by hand to move the head, say, 10mm in the X or Y direction, and repeat this at different Z heights, using a micrometer to see if things shift as the Z geometry increases. It also might be useful to run the job in such a way that you can move the head to a zero point before and after it prints. If the zero wanders, then you're skipping steps.

[Leon Grossman]

The firmware in the delta bot does the transforms to convert everything to cartesian coordinates.

Based on my output and the layer alignment I'm getting, I'm unlikely to be skipping steps.  It is still something I'll have to check on reassembly.

Had I realized how picky delta robots are for symmetry when I started this process, I would have been far less likely to pickup a delta printer.  With Cartesian, you can look and say, "Ah, skew in the X plane and that must mean that the X axis is skewed".  With a delta, it's "Ah skew in the X plane and that must mean some form of quantum relationship and I'm afraid the cat must be dead."

I'll send updates when MonkeyBot v1 comes together out of the ashes of my BI v2.5.

Ryan Pierce

Thursday, January 22, 2015 11:27 PM

Leon Grossman

Thursday, January 22, 2015 9:01 PM

Ryan,

* This is a part my other printer would have no problem with.  

* My other printer is a cartesian and this is a delta so maybe I’m expecting too much.

* The skew exists on larger parts (like a 100x100x25 hollow square with 4mm thick walls.

* I’ve used Simplify3D and Slicer (using the manufacturer’s distro)

* I used the original firmware that came with the machine and my own version of Repeater.

* Slow motion video distinctly shows the end effector moving and not the part.  I have the glass clamped to the build surface and the part has to be scraped off with a spatula.  Any bad motion is strictly in the mechanics.

I’m not sure if I follow your micrometer suggestion as all three axes are moving for pretty much any linear motion so I’m not sure what I should measure.

That said, I agonized over my options for a couple of days.  Almost tossed this printer aside and then decided to double down instead.  I think the issue most lies in a defect in the carriage/rod/end effector system.  I’ve got linear motion rails and super precise magnetic carbon fiber rods on the way.

I’ve already got the motion system removed from the current printer in preparation so I can’t run your tests but I will explore this when I’ve got things back together 

Thanks for putting in the thought because I’ve been at wits end given my extreme attention to detail on getting the system setup correctly!

Leon

Ryan Pierce

Monday, January 19, 2015 10:19 AM

My $0.02:

Trying to print a part with that small a footprint and that much height relative to it seems like a recipe for disaster. With my 3D printed bagpipes project, I'm sometimes having problems, and my parts' aspect ratios are nowhere near as extreme as yours. The part is really acting like a lever, and forces the print head applies to it get magnified where it touches the bed. If the part shifts or flexes while printing, then even if the head motion is correct to within 0.00000001", you will still get a wonky part.

If it were me, I would take the filament out of the equation entirely. I would try to find a way to play the gcode to the printer with no filament loaded and "air print" it but pause the print head at certain adjacent vertices of the outside of the rectangle at various Z heights. Use a micrometer gauge on a mast perpendicular to the bed to measure the position of the print head. This will tell you if the print head is making one dimension consistently longer or shorter as the Z height rises. If so, then either the gcode is bad, or something is wrong with the machine's alignment and/or positioning. If not, then I'd highly suspect the part moving during the print, and I'd consider time lapse video zoomed in on it as a way to observe possible changes.

Ryan

On Jan 19, 2015, at 9:22 AM, Leon Grossman <leongr...@gmail.com> wrote:

--
You received this message because you are subscribed to the Google Groups "CNC Build Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cnc-build-clu...@googlegroups.com.
To post to this group, send email to cnc-bui...@googlegroups.com.
Visit this group at http://groups.google.com/group/cnc-build-club.
For more options, visit https://groups.google.com/d/optout.

Leon Grossman

Monday, January 19, 2015 9:22 AM

[Leon Grossman]

All,

I spent the last few weeks designing, ordering and assembling the MonkeyBot.  It's an unholy union of BerryBot, OpenRail and Boots Industries parts.  I'm about 80% of the way through the build process but I now have a printer which is delivering pretty amazing build quality already.

Also, OpenRail is amazeballs.

Leon

Leon Grossman

Friday, January 23, 2015 9:43 AM

[Sparr]

On Mon, Jan 19, 2015 at 10:19 AM, Ryan Pierce <rdpi...@pobox.com> wrote:

If it were me, I would take the filament out of the equation entirely. I would try to find a way to play the gcode to the printer with no filament loaded and "air print" it but pause the print head at certain adjacent vertices of the outside of the rectangle at various Z heights. Use a micrometer gauge on a mast perpendicular to the bed to measure the position of the print head. This will tell you if the print head is making one dimension consistently longer or shorter as the Z height rises. If so, then either the gcode is bad, or something is wrong with the machine's alignment and/or positioning. If not, then I'd highly suspect the part moving during the print, and I'd consider time lapse video zoomed in on it as a way to observe possible changes.

Don't even do an air print of your part. Send gcode manually and send the machine to specific known and easy to measure coordinates. 

[Leon Grossman]

Status update:

I have completely rebuilt the printer with OpenRail and BerryBot parts.  I put a dial indicator on the head and ran a matrix across the bed.  It seems I found the problem but not the solution yet.  Everything I know about delta printers (which is, admittedly, not that much) indicates that with a symmetrical build, the distortions should be a bowl or a dome based on printer radius or part length based on diagonal rod length.

I'll update here when I've figured out the problem.

Leon

X across, Y down.  Dimensions in mm, bolded values are in front of the towers:

	-150	-130	-100	-75	-50	0	50	75	100	130	150
150						0.102
130				-0.533	-0.330	0.051	0.102	0.127
100			-0.660	-0.406	-0.406	0.051	-0.051	0.089	-0.051
75		-0.864	-0.635	-0.356	-0.318	0.051	-0.076	0.051	-0.152	-0.140
50		-0.737	-0.508	-0.305	-0.241	0.025	-0.038	0.013	-0.127	-0.152
0	-0.432	-0.508	-0.254	-0.229	-0.102	-0.013	-0.025	-0.051	-0.102	-0.127	-0.127
-50		-0.330	-0.076	-0.152	-0.051	-0.076	-0.025	-0.076	-0.038	-0.102
-75		-0.089	-0.038	-0.191	0.000	-0.102	-0.025	-0.102	-0.025	-0.025
-100			-0.025	-0.203	-0.051	-0.127	-0.064	-0.114	0.000
-130				-0.152	-0.102	-0.229	-0.127	-0.064
-150						-0.241

 

Sparr

Thursday, February 5, 2015 1:53 PM

If it were me, I would take the filament out of the equation entirely. I would try to find a way to play the gcode to the printer with no filament loaded and "air print" it but pause the print head at certain adjacent vertices of the outside of the rectangle at various Z heights. Use a micrometer gauge on a mast perpendicular to the bed to measure the position of the print head. This will tell you if the print head is making one dimension consistently longer or shorter as the Z height rises. If so, then either the gcode is bad, or something is wrong with the machine's alignment and/or positioning. If not, then I'd highly suspect the part moving during the print, and I'd consider time lapse video zoomed in on it as a way to observe possible changes.

Don't even do an air print of your part. Send gcode manually and send the machine to specific known and easy to measure coordinates. 

--
You received this message because you are subscribed to the Google Groups "CNC Build Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cnc-build-clu...@googlegroups.com.
To post to this group, send email to cnc-bui...@googlegroups.com.
Visit this group at http://groups.google.com/group/cnc-build-club.
For more options, visit https://groups.google.com/d/optout.

Ryan Pierce

Monday, January 19, 2015 10:19 AM

My $0.02:

Trying to print a part with that small a footprint and that much height relative to it seems like a recipe for disaster. With my 3D printed bagpipes project, I'm sometimes having problems, and my parts' aspect ratios are nowhere near as extreme as yours. The part is really acting like a lever, and forces the print head applies to it get magnified where it touches the bed. If the part shifts or flexes while printing, then even if the head motion is correct to within 0.00000001", you will still get a wonky part.

If it were me, I would take the filament out of the equation entirely. I would try to find a way to play the gcode to the printer with no filament loaded and "air print" it but pause the print head at certain adjacent vertices of the outside of the rectangle at various Z heights. Use a micrometer gauge on a mast perpendicular to the bed to measure the position of the print head. This will tell you if the print head is making one dimension consistently longer or shorter as the Z height rises. If so, then either the gcode is bad, or something is wrong with the machine's alignment and/or positioning. If not, then I'd highly suspect the part moving during the print, and I'd consider time lapse video zoomed in on it as a way to observe possible changes.

Ryan

On Jan 19, 2015, at 9:22 AM, Leon Grossman <leongr...@gmail.com> wrote:

--
You received this message because you are subscribed to the Google Groups "CNC Build Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cnc-build-clu...@googlegroups.com.
To post to this group, send email to cnc-bui...@googlegroups.com.
Visit this group at http://groups.google.com/group/cnc-build-club.
For more options, visit https://groups.google.com/d/optout.

Leon Grossman

Monday, January 19, 2015 9:22 AM

[Donald J]

Any luck yet?

I've started following the Delta Google Group (https://groups.google.com/forum/#!forum/deltabot); maybe they can help.

On Monday, February 9, 2015 at 1:41:28 PM UTC-6, Leon Grossman wrote:

Status update:

I have completely rebuilt the printer with OpenRail and BerryBot parts.  I put a dial indicator on the head and ran a matrix across the bed.  It seems I found the problem but not the solution yet.  Everything I know about delta printers (which is, admittedly, not that much) indicates that with a symmetrical build, the distortions should be a bowl or a dome based on printer radius or part length based on diagonal rod length.

I'll update here when I've figured out the problem.

Leon

X across, Y down.  Dimensions in mm, bolded values are in front of the towers:

	-150	-130	-100	-75	-50	0	50	75	100	130	150
150

...

[Leon Grossman]

Yes, I did.

The problem was that the build platform was made of aluminum that had the consistency of warm butter and was deforming.  I'm waiting on a new, thicker platform to be made before I can completely dial in the machine. 

In the meantime, The newest problem is extreme ooze due to the 1 meter plus Bowden tubes. I'm currently working on a flying extruder design which I hope to complete this week.

If people are interested, I'd be happy to bring the printer in when I have completed it. 

Leon

--

[Andrew Sowa]

Aggressive retraction stopped my oozing but it is definitely not an ideal solution.  I would like to switch mine to a flying style more for torque/speed reasons.  The airstripper extruder doesn't like to push much plastic. 

Count me in, if you decide to show it off. 

[Leon Grossman]

Even aggressive retraction was not enough to save me.  The flying extruder seems to be working but I still have some issues to sort out there with frame rigidity.

I'll bring it into the space on a Saturday or Sunday in the near future after I've received the final parts from Werner Berry for the bed and tower carriages.

Andrew Sowa

Tuesday, March 10, 2015 12:02 PM

Leon Grossman

Saturday, March 7, 2015 4:44 PM

Yes, I did.

The problem was that the build platform was made of aluminum that had the consistency of warm butter and was deforming.  I'm waiting on a new, thicker platform to be made before I can completely dial in the machine. 

In the meantime, The newest problem is extreme ooze due to the 1 meter plus Bowden tubes. I'm currently working on a flying extruder design which I hope to complete this week.

If people are interested, I'd be happy to bring the printer in when I have completed it. 

Leon

On Mar 7, 2015, at 5:33 PM, Donald J <dj....@gmail.com> wrote:

Donald J

Saturday, March 7, 2015 4:33 PM

Any luck yet?

I've started following the Delta Google Group (https://groups.google.com/forum/#!forum/deltabot); maybe they can help.



On Monday, February 9, 2015 at 1:41:28 PM UTC-6, Leon Grossman wrote:

--
You received this message because you are subscribed to the Google Groups "CNC Build Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cnc-build-clu...@googlegroups.com.
To post to this group, send email to cnc-bui...@googlegroups.com.
Visit this group at http://groups.google.com/group/cnc-build-club.
For more options, visit https://groups.google.com/d/optout.

Leon Grossman

Monday, February 9, 2015 1:41 PM

Status update:

I have completely rebuilt the printer with OpenRail and BerryBot parts.  I put a dial indicator on the head and ran a matrix across the bed.  It seems I found the problem but not the solution yet.  Everything I know about delta printers (which is, admittedly, not that much) indicates that with a symmetrical build, the distortions should be a bowl or a dome based on printer radius or part length based on diagonal rod length.

I'll update here when I've figured out the problem.

Leon

X across, Y down.  Dimensions in mm, bolded values are in front of the towers:

-150

-130

-100

-75

-50

0

50

75

100

130

150

150

0.102

[Donald J]

Werner Berry is selling parts?  Cool.

[Leon Grossman]

http://www.shopspt.com/

We've been working together for a couple of months on my printer retrofit.  He's working with my designs for the OpenRail retrofit to my Boots Industries v2.5 printer and will be offering those parts for sale to the ill-fated owners of that printer.  He's mentioned something about creating something for a kit.  I think he might actually be working on a BerryBot kit.  That would be super, amazingly cool if he does that.

Donald J

Tuesday, March 10, 2015 2:38 PM

Werner Berry is selling parts?  Cool.


On Tuesday, March 10, 2015 at 2:08:23 PM UTC-5, Leon Grossman wrote:

--
You received this message because you are subscribed to the Google Groups "CNC Build Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cnc-build-clu...@googlegroups.com.
To post to this group, send email to cnc-bui...@googlegroups.com.
Visit this group at http://groups.google.com/group/cnc-build-club.
For more options, visit https://groups.google.com/d/optout.

Leon Grossman

Tuesday, March 10, 2015 2:08 PM

Even aggressive retraction was not enough to save me.  The flying extruder seems to be working but I still have some issues to sort out there with frame rigidity.

I'll bring it into the space on a Saturday or Sunday in the near future after I've received the final parts from Werner Berry for the bed and tower carriages.

[Leon Grossman]

I finally received the parts from Werner and am shaking down my printer this weekend.

Basically, the majority of the frame, *some* of the electronics, much of the wiring, and few of the connectors remain the same.  My goal was to build a world class printer and I’m pretty close to that goal.  I’ve already ordered a Chimera hot end and will be tweaking and improving as I go now that I have a functional system.

https://www.youtube.com/watch?v=07SX05Y9jtU

Leon Grossman

leongr...@gmail.com