Head rises when moving towards the centre?

Hi all.
I'm making a Delta printer, based on Rostock Mini design.
Basically, it works fine, and i've calibrated it to move flat and to scale.

However, there is a slight problem.
When the carriage moves away from the centre, it moves flat.
When it moves towards the centre, it tends to rise a little, especially closer to the edge of the build area.
If you order the head manually 10mm, 10mm forward, then 10mm, 10mm back, this is what you would get:


I.e. if you print a thin rectangle from centre, then the side will be flat to the edge, flat acrosswise, then rise and fall in a curve as the head moves back towards centre, then flat again.
We are only talking 0.2-0.3 mm rise at worst, but it wrecks chaos on larger builds with fine details on first layer, and might result in a nozzle crash with resulting offsets ruining the build.

What could be causing these rises?
It feels suspiciously like some sort of backlash, but i'm using GT2 belts with machined pulleys, and tightened to a note, so it's not obvious where to look.

Deltas require symmetry, parallelism, and squareness.
Error of movement comes from these deviating from being perfect, (assuming your software variables are not mucked up).

Length dimensions can be out of blueprint specification, as a new length can be accounted for with software.

Quote
A2
Deltas require symmetry, parallelism, and squareness.
Error of movement comes from these deviating from being perfect, (assuming your software variables are not mucked up).

The distances between the towers measure within 1mm from each other, the heights are as well, the rods are identical by way of assembly.

Other reason i don't think it's a problem of geometry or assembly - the problem is directional, not an absolute deviation or something you would expect from a bodged geometry.
If it goes one way, it curves up, the other - goes flat.

I'm using Marlin firmware from GIT of a week ago.

try this....




adjust .5mm at a time

Quote
xboxhacker
try this....

Nope, that works perfectly.

I was watching the carets on the towers - they stop in different positions for going one way, and for going the other way.
It's not steps being skipped (or if it is, it's peculiarly symmetric and repeatable), since the positioning is always perfect if you start from the centre.

So, there is something wrong in the software - the positioning error of some sort is direction-dependent.
I've been trying tweaking settings at random, with no measurable effect on the problem.

Maybe i'm not explaining it too well, let's try another way.

Let's say i command the head to position (0, 40, 0.1).
If it was in the centre, or in position (0, 30, 0.1), then it would arrive at (0, 40, 0.1).
If it was at (0, 50, 0.1), then it will arrive at approximately (0, 40.1, 0.3).
This seem to be an actual offset in the motor positions - i'll try to print out the raw data going to them.

Simply speaking, the position the head arrives at for exactly the same given coordinates differ based on from which direction it was arriving.

How did you find out what location did the head come to?
Did you use M114 (Get Current Position) command?

Quote
hercek
How did you find out what location did the head come to?
Did you use M114 (Get Current Position) command?

At first i could see the carets stopping in different positions - about 0.25mm off - depending on direction.
I've since added printing of delta coordinates to the LCD, then added step counters to stepper.cpp, with similar printout.
However, it's not a easy software issue - the numbers match up on each run and direction.

The problem is somewhere past the software - either mechanical or related to Pololu drivers (running at x8 microstepping).

Maybe i want too much print area with too much perfection?
The problem is only meaningfully large near the edges, which in my case are half way to the tower, with rods almost vertical.

How could be rod almost vertical when the head is half way to the tower? Ok, maybe you can call about 14° agnle almost vertical.
Anyway it looks like you are geting about 0.25 mm Z-error when you move half the maximum possilbe distance (ie 50%) from center to a side. I do not know about rostock mini, but a clasical old rostock can be calibrated to move withing 0.05 mm z-height area when moving 75% possible distance to a side. So it should be possible to calibrate better. The point is that the more to the side you are the more precise your calibration needs to be to keep given maximum z-height error.

Since you have different Z-height when you arrive to the same spot from different locations, your problem is primaryly not a calibration problem (and probably not some software bug either). I actualy observed this on a rostock here. But the problem was only temporary and never caused a bad print. Just after homing when I moved head down to the bed and moved from center to a side, measured z-height, returned back to center and went back to the side and mesured the z-height again; then I got a different z-height sometimes. The difference was typically about 0.1 mm. The difference was not always there. Whenever I measured this, the problem disapeared when I was moving for a while just above the bed (mostly I just did a circle around outer side of the heat bed). From that time on, the data from z-probe were consistent withing ±1 microstep (±0.0125mm). I did not find out why was it happening. My guess would be that belts somehow needed to settle to the puleys or maybe ball joints needed to settle somehow (they are quite tight). GT2.5 belts are used.

Edit: only spelling.

Edited 2 time(s). Last edit at 01/10/2014 01:05PM by hercek.

Quote
hercek
How could be rod almost vertical when the head is half way to the tower?

Cause it's quite small - 14 cm centre to tower rails, with extruder location at 7 cm towards tower, the horizontal distance covered by the 18.5 cm rod is (7 cm - 3.3 cm for platform -2.2 cm for carriage) = 1.5cm, almost vertical.

Quote
hercek
My guess would be that belts somehow needed to settle to the puleys or maybe ball joints needed to settle somehow

Maybe.
I can't quite move the platform by hand to stay between two states - no obviously slack anywhere.
It seem to get less noticeable as i keep tuning and running my machine (or maybe just stops to matter as much), so it might be an interference pattern of several suboptimal tuning issues, or something like that.

From: [reprap.org]

Quote
reprap wiki
If your print head is too high or low in the middle of the print surface, adjust DELTA_SMOOTH_ROD_OFFSET by half mm and try again.

DELTA_SMOOTH_ROD_OFFSET is the horizontal offset from middle of printer to smooth rod center. I don't own a delta (yet) so I'm just going by what the wiki page says.

See Johann's blog entry on how to fix this: [deltabot.tumblr.com]

Edited 1 time(s). Last edit at 01/11/2014 05:04PM by TheTechnicalNoob.

What you have is backlash. Since it repeats every time, it's not a skip. Look at things like belt tension and play in the rod ends.

One classic is the belt drive gears slipping back and forth on the stepper shaft because they aren't quite tight enough. If your steppers have flats on the shafts they will slip over a very narrow range. The result will be what you are seeing.

If it was a measurement issue, it would not be directional (one error forward, different error in reverse).

If it was software, everybody on the planet would go nuts about it.

Edited 1 time(s). Last edit at 01/20/2014 04:24PM by uncle_bob.

you could also check my blog post on [www.tripodmaker.com] to get more knowledge on how a delta printer moves around

Did you ever find a solution for this? I am having the exact same issue.