SF Lash module

Hello,

Perhaps you can help me with this:

Recently Woodsmoke, my printer, struggles to replicate stls involving circular edges accurately. Attempting to print overtly circular components results in scruffy, egglike nonsense. The symptoms are already discussed under another thread:
[ forums.reprap.org ]

Some alteration made to Woodsmoke's mechanics has introduced backlash. Prints of the Wades Extruder body no longer accept the 608 bearings without excessive post production. This seemed like an opportunity to learn about the Lash module in SF. The next step was to design a calibration piece: [ thingiverse.com.]

Intuitively I understand how Erik's original gcode path identifys the problem, but struggle to describe the process with words. Essentially the lashmaze allows you to measure the print head's positional inaccuracy at a point, when it is traveling in one direction along an axis compared with when it is traveling in the other direction? Half the difference is the amount of backlash on that axis. Great. Fine. I took the theory for granted and entered the lash values measured. Circles seemed better.

Raldrich was thorough enough to reprint the maze with his corrected settings and good enough to post me his findings. He had found that there was still measurable backlash after software correction. More specifically He reasoned that the lash value should be the full difference between positions, rather than half. I believe his exact halving of the lash value with a corrected print to be a fluke and I get the impression that he didn't remeasure his second set of corrections. Raldrich can confirm. But clearly I should have been that thorough. If the lash module (and the lashmaze) works, one should be able to print a maze with no measurable backlash.

The trouble is I can't.

I was going to delay this post until I had time to print a proper array of tests and write up my best guess at what is going on. But it seems several people are having problems with the method. Has anyone else had positive or negative experiences attempting to measure their backlash with the lashmaze? For my own part I have found measuring with a digital caliper particularly finickity, with care needed to get repeatable results. How accurate could one expect their final lash results to be?

>0.1mm ?

Now hopefully the topic is open to discussion. I will follow this up with some testing when I can.

woodsmoke

Edited 3 time(s). Last edit at 08/25/2011 07:03AM by Woodsmoke.

Recently did your lashmaze test with my pirated cupcake.
Showed good alignment of the ridges, but sensibly reduced gaps in the Y axis as compared to the X axis. Will write down the values.
And yes, looking at printed circles, they come out slightly un-round. (sigh!)

Have not yet played with the settings in skeinforge, that will be the last resort. First I am going to check :
- Y belt tension (notoriously difficult to access in my setup), feels rather normal and similar X belt tension to me, but maybe a bit in the low? BTW what would be the optimal belt tension?
- Y idler adjustment on its axis (hadn't thought of that, but yes, of course, it does matter), I think there is some play there
- Y motor current (could it be too low? also need to check that motor's temperature - it is hidden and out of reach under the carriage)
...
any other sources of backlash one could control?
The rods are outstandingly smooth, friction is at a minimum, well oiled/greased...

Lash is a fact of life when you do any kind of CNC, as well as dopey ballscrews with +/- 0.15mm of positional error per 300 mm, let's face it: few things more common.
But people get emotional quickly when it is question of 'lash (cf. other thread).

Other things that cause it are: loose pulley set screws allowing a little slip in each direction from the flat, broken wires in the belt, loose bushings allowing the carriage to rotate slightly if it pulled from one side, pulleys with teeth that are narrower than the gaps in the belt.

I think belt tension should be high as possible to reduce lash on Mendel, but on Prusa it is limited. As discussed elsewhere too much on Y bends the motor bracket. Too much on X bends the X rods and makes the top Z bushings come away from the bars. The Sells Mendel is a lot better in that respect.

I think one side effect of belt tension is I think it can change the resonant frequency of the axis, which limits the maximum zig-zag rate when doing small infills.

---

[www.hydraraptor.blogspot.com]

Another side effect of belt tension is it must affect the XY calibration a little as it must stretch slightly, even with steel or kevlar reinforcing.

---

[www.hydraraptor.blogspot.com]

I wonder what tension the belt pitch is quoted at?

---

[www.hydraraptor.blogspot.com]

nophead Wrote:
-------------------------------------------------------
> I wonder what tension the belt pitch is quoted at?

I have been foraging for similar information, but nothing definitive yet.

An interesting document for modelling the Mendel and similar:

[www.gatesmectrol.com]

As for the Prusa I think I got the Y motor bracket problem solved with a modified mount. Not yet tried.

Should help, the Y axis gets to push quite some mass if doing the heated bed plus glass setup, so it needs more tension.

As for the X axis: have been thinking of doubling the smooth X rods with a pair of M8 rods, 15 mm outside, coplanar.
Not sure it is the best idea.

Not yet arrived to fine measurement of bending in the bars, dealing with the gross and most obvious stuff first...

The tension on the belt is like the tension on a guitar string. A tighter belt should move the natural resonance to highter frequencies. But can't we also move the resonance frequency up by forcing the belt into a harmonic mode?

I'm thinking a ball bearing held at the right point on the belt should damp out most of the lower resonant modes.

Menawhile here is a spreadsheet with the measured distances and calculated values.
Woodsmoke, given the discussion in your blog, should I enter the (averaged A - averaged and (averaged C - averaged D) values in skeinforge, or halve them as you recommend?

Confirms what could be oogled: X lash is 0.067, Y lash is 0.184

@all: What values of lash are you getting in your machines?

Edited 2 time(s). Last edit at 08/28/2011 07:31AM by Lanthan.

Holy Thread Revival Batman!!

Just did the backlash test.

My results...

backlash x: 0.046mm
backlash y: 0.107mm

This is on a Prusa V2 (pretty much) with Syncromesh cable rather than belts.

Ok.

I tried putting the above measurements into the lash module and ran the test again.

On the Y axis I saw an improvement, 0.019mm. That's quite a difference.

The X axis got worse though 0.098mm. I am wondering if there is a natural limit to what you can improve on in software?

Edited 1 time(s). Last edit at 04/16/2012 07:09PM by Justblair.