SD2 Z Accumulation Error

Thanks grob, I appreciate the help. Here's what I found out:

(1) I'm using the stock 5/16-18" z screw. On connect, the SD2 reports my eeprom steps/unit as M92 X88 Y88 Z2268 E115. However, soon after buying the SD2 I printed a calibration angle and adjust my steps/unit to get me within 0.1mm of actual. this gave me the following : M92 X87.5448 Y88.4222 Z2372.3849. I set this using the initial gcode feature in Slic3r. I attached screenshots of my setup and initial gcode reported after slicing.

(2) I verified there isn't any scaling in RH. I sliced a 20x20x20mm cube at 0.3mm layer as instructed and got 67 layers. So the slicer seems ok.

Thoughts?

Jeff

2268 is definitely the correct value and tnis can be verified with a variety of online calculators or manual calculation of the 5/16-18 threaded rod. The x and y axes are 88 steps per mm assuming MXL timing belts.

Grob is right. Measuring prints and adjusting steps per mm is not appropriate. Aside from repeatability issues like backlash, you are making the assumption that slic3r has actually rendered your object to GCode preserving the exact dimensions. This is often not true, due to the quantization of extrusion space.

The only proper way to empirically determine steps per mm is with a dial indicator, but the theoretical ways are sufficient. Also keep in mind that although steps per mm can be a non-integer, non integral steps will experience rounding error and should be avoided in general. One exception to this rule is with geared drives, but there are no geared drives on a stock Solidoodle.

That tends to be a pretty small effect though. If you look at what it is doing, it is sending absolute height movement commands, so the firmware moves as close as it can to that height and knows where it really went, so there isn't a cumulative rounding error or anything, it just gets as close as it can to the desired height independently for each layer.

I did once print a part that was supposed to be 90mm high and it came out about 100mm, but when I went back to try and figure out the problem, I never saw the issue again. I'm going with the theory that I accidentally brushed the keyboard and set a Z scale factor on that one print :-).

1% is pretty huge really. Should be aiming better than this (in most cases) hehe!

I have to politely clarify here, as Claghorn is right:
* The g-code gives Marlin absolute values in mm (it will ask the z axis to go to "100mm")
* Marlin will calculate, whether you send it absolute or relative mm, the absolute position you've asked for in mm
* It then converts this to an absolute position in steps using the steps-per-mm (i.e. 226800 steps), and all the complicated assembler stuff borrowed from GRBL moves the axis to that position. 226800 steps on a (theoretically perfect) 5/16-18 rod is 100.0125mm - much less than 1%, I'd guess in this world negligible. To make it better, use more significant figures in the steps-per-mm (e.g. 2267.72).

There is still a good reason to use the 0.2963mm layer height: each layer lands on a whole step, and there's not much rounding, so the layer height will be much more precise in reality (microsteps are not very accurate, and even whole steps are only vaguely accurate - check a stepper motor datasheet!). There have been claims of visible banding from this.

In the end, we have a problem where the angular position of the motor is only so accurate - if +/- 0.5 whole steps would lead to visible issues in the print (0.0035mm >= 1% of a layer in the case of the stock rod), then the best solution is probably reducing the pitch of the rod (e.g. M3x0.5 --> 6400 usteps/mm --> +/- 0.5 whole steps = +/- 0.00125mm ~ 0.42% of a layer).

I just like that the numbers are cleaner hehe

.3mm off is indicative of backlash. Check your belt tension and make sure it is even between the two side belts and the short motor belt is tight.

If you do a search around here you'll find all kinds of suggestions for making the system more precise and rigid. Bearings help quite a bit. A stock aolidoodle is fairly sloppy mechanically and needs some work.

Thanks jagowilson!

I started tensioning my Y-belts and noticed the stock Y-brackets were weak and flexed easily. So I printed the "overkill" brackets and replaced them. Then squared everything up and re-tensioned the belts using the SD video. Since I was at it, I went ahead and added in the Z nut-hugger and backlash spring to help with banding, and the thumbscrew wheels to make leveling a bit easier.

Then I was curious what other hardware upgrades I could entertain to improve "accuracy" and "replication accuracy (copies of parts)". So far I only have this:

* Replace XY carriages to use bearings: http://www.thingiverse.com/thing:104059

Thoughts?

Edit: I should mention that I have a near stock SD2 (new Y brackets mentioned above). It printed right out of the box and has really performed good enough for me for 2 years using RH/Slic3r.

Jeff

I gave up trying to keep z steps at 2268,my cubes were always a little smaller than the STL file, I  first I tried 2400 then dialed it down to 2304, I get very accurate  results now.  I suspect my motors are getting older or the v reff may not be optimum, even though it reads 1. 01  still that 100th v shouldn't matter.