Is this the effect of missing steps on a delta?

Hi Folks,
I have a stock sized Mini Kossel built mostly with bits from builda3dprinter.eu only the printed parts were sourced elsewhere for speed in delivery.
I'm using ABS filament and have an E3D V5 hotend and a mylar bed heater from builda3dprinter.
Electronics are smoothieboard 5X and smoothieware.
I'm pretty sure I have calibration down OK and can see flat movement when I jog the head around at low height above the bed using the LCD and selecting X or Y movement.
However when I try to start a print by playing a file from the SD card (in this case the hyperbolid coolling tower pencil holder from thingiverse as it has a nice big flat base) this happens:

The first perimeter is laid down ok, you can see very little variation in the bead size around the circumference which bodes well for flatness.
However as soon as the first layer fill starts (normal straight line crosshatch fill) I'm seeing the bead get far too squashed in the middle of each line and too thick by the perimeter.
After a few lines the head crashes and I have to kill the power.

This happens wherever on the print bed I make the print, so I don't think it's a levelling issue, I think I might be seeing the result of a step or two being missed as the head makes a fast move from the perimeter to the start of the fill.

Would you folks more experienced with deltas agree? What are the results of step skipping in a delta configuration? I'd imagine you don't get simple offsets like a cartesian setup.


Cheers,
Robin.

It sounds like its not calibrated well, but could you post a picture or video of the problem? It would be way easier to troubleshoot it.

A pic is still worth a 1k words

True

I have since tried reverting from the edge branch firmware to stable master ...same result.
Also I've got into the gcode and changed all the rapidmoves (F3000) to slow ones (F300) and no change either... hmmm!

I've attached a photo, where you can see the fill getting squeezed dry by the head hitting the bed (as far as I can tell).

If I move the head using the jog on the LCD to 0,0,0 it is just a paper thickness above the centre of the bed, as it should be. I can then use the jog X and Y to move it over to the area where it would be scraping the bed during a print without it hitting the bed at any point.

Any ideas? What would you do next to diagnose this?



Cheers,
Robin.

Is your DELTA_SEGMENTS_PER_SECOND and what was the print speed?
Very low DELTA_SEGMENTS_PER_SECOND and high print speed can result in something like this.

For deltas you should pass the paper test at 4 points. Positions close to the three towers and the middle. It could be that you have the middle point lower than the outer bed positions (concave surface), that why you can print the outer perimeters ok, but not the infill.
Normally, you correct this by adjusting delta_radius in configuration. You can change the delta radius indirectly by changing the smooth_rod_offset (or any other of the delta parameters) to lower or raise the nozzle at the center. If I recall correctly, in your case you should decrease the smooth_rod_offset a bit and see if it improves things. But really, check if you can pass the paper test at all 4 points, if you cannot, you have a problem with calibration. If you can, you need to take another look at z-height and extruder calibration.

In case you have autolevel probe installed, you can also use the Marlin firmware from Rich Catell which will do all of the calibration for you (its on github, google it). Richs Marlin can adjust most of the parameters by itself, the endstops, the delta_radius and the diagonal rod length as well, and a couple of few other things not in the original firmware.

Quote
imyz
For deltas you should pass the paper test at 4 points.

I assumed he did it:

Quote
Zedsquared
I'm pretty sure I have calibration down OK and can see flat movement when I jog the head around at low height above the bed using the LCD and selecting X or Y movement.

But if not, he should check 7 points and not only 4: [forums.reprap.org]

Hi Folks,
I did check bed clearance at the centre and each tower, I can also scoot the head right across the bed without it crashing if I use the jog so I don't think the radius is wrong.
I'm away from the printer now until I get back to work tomorrow but plan to try some handcrafted Gcode to see what happens. I can check segments per second then too but I'm pretty sure it's the same as the default smoothieware config.

Cheers,
Robin.

Its also possible youre overextruding. Have you checked the z-height?

With a delta, its not where the moves start and end. Its how it calculates the move during the move. If you have not set your arm lengths and radius properly with your delta in the firmware, your moves will be concave, which fits exactly with your problem. Even a little off will cause problems.

---

"Never argue with stupid people, they will drag you down to their level and then beat you with experience."

I've got a printrboard with Marlin too so I'm going to try some autocalibrate (the G30A command that probes to iteratively solve the geometry rather than the G29 that just maps the bed) action over the weekend and see where that gets me, still not convinced calibration is the issue though as the parabolic motion only seems to happen when actually printing. Maybe the PSU is getting overloaded when the bed and hot end kick in or something and so steps are getting skipped.

Can anyone answer my original question: what is the effect of missed steps on a delta robot?

More when I have some better data, thanks for the suggestions folks.

Cheers,
Robin.

What you are getting is unlikely because of missed steps. They would need to be missed in one direction and later in the opposite direction just to compensate.
Missed steps on one motor result in wrong z-position of the corresponding carriage on the tower. You get exactly the same thing when you do not have correctly adjusted end-stop on the tower. The bed z-error because of that looks like this:

hercek,

I am amazed at your bed graphics! I am also very appreciative of your support through all of these bed leveling questions.

Hercek nailed it in his first reply all the way up there and I was totally overthinking it

I'm ashamed to say that I somehow overlooked that I had both segments_per_second and mm_per_line_segment (last one is smoothieware specific I think) commented out.
So basically only the end points of the lines were in the right place and the long straight lines on the fill that were dipping down in an arc and jamming the head against the deck. The perimeter was made of lots of small segments anyway and so didn't show a problem.


Cheers,
Robin.

Great you figured it out at the end