Rep2: Difficult leveling problem (no, really. It's difficult.)

[David Kessner]

So I'm having a devil of a time leveling the build plate on my Rep2.  Something is way out of whack and I don't think it's me.  Here's what I've done (sorry, it's a long one).

I have two glass plates, and the acrylic.  I tested each one for flatness using a brand new steel ruler and a set of feeler gauges.  I made sure that the ruler was 90 deg to the platform and not bending.  The worst plate (glass, surprisingly) allowed a 0.1mm feeler gauge to slip under the ruler.  The best (also glass) plate barely allowed a 0.04mm gauge to slip under. I redid the test several times to make sure it was consistent.

I took the best plate, cleaned it really well, added some extra "shims" under it using three layers of blue tape in addition to the felt pads already on the bot, and put the plate in the bot.  Note, the felt pads are right next to the pre-existing rubber shims.

I told the Rep2 to level the platform (Stock 7.0 firmware).  Once the head was in the "first position" I canceled the print so that the motors were off and the platform was in the correct z-position.  Using the 0.10 mm feeler gauge, I moved the head over each of the adjustment screws and adjusted.  

Next, I clipped on whpthomas's dial indicator (http://www.thingiverse.com/thing:41090) and measured the height of the build platform in several places.  Then remeasured 2 or 3 more times to make sure the measurements were consistent.  In all cases the measurements were consistent to within 0.02 mm.

So, here is the "topo map" of my platform:

Rear row, left/center/right:  -0.19, (+0.01),  0.00

Middle Row, left/center/right:  -0.08, +0.09, -0.02

Front Row,  Left/center-left/center-right/right:  -0.07, (0.00), (0.00), -0.13

All numbers are in mm, and represent a deviation from "nominal".  Numbers in parenthesis are from positions directly above an adjustment screw.  

What this is saying is that I have a 0.28 mm deviation between the left rear corner and the center of the platform!  It is absolutely no wonder that I can't get 0.1mm layer prints to stick.  If the nozzle is pressed against the platform in the center then it's still floating 0.28 mm off the glass at that corner!

Just for grins, I put in the other glass build plate and quickly (less methodically) redid these measurements and they match almost identically.

My conclusion from this is that the bot's gantry system is warped.  

Does anyone have any thoughts?  Ideas on how to fix it?  Etc?  Other than making a build-platform that is warped identically, I have no ideas.  Well, I can always yell at MBI...

[Dan Newman]

If I understand correctly, a negative value means that the extruder
carriage/plate are closer than "nominal" and a positve value means
further apart. With that undestanding,

So, some how the extruder/plate corners are closer in the back left
and front right corners (-0.19 mm and -0.13 mm, respectively)? Since
the plate is flatter than that, if it's tilted up in the back left, we'd
expect it to be lower in the front right for a positive deviation,
not a negative one. So, yes, something seems off with the gantry
system. It almost looks as though the X axis rods might be bowed
upwards a slight amount? That might account for less of a gap on the
left/right ends and more of a gap in the center?

If anything, I'd expect them to bow downwards a tad… So, perhaps
I have a sign error in understanding your numbers.

> Does anyone have any thoughts? Ideas on how to fix it? Etc? Other than
> making a build-platform that is warped identically, I have no ideas. Well,

> I can always yell at MBI…

With everything installed, it's hard to judge if the long X axis rods
are bowed or not, but I'm wondering if that's the case. What to do about
it? Ask for straighter rods? I wonder if they are within the spec of
what MBI ordered? These rods usually come with specs on deviation from
straight per some unit of length.

Dan

[Eighty]

I'm sitting in a hotel room 5 states away from my printer, so can someone answer a question for me?  What's the diameter of the X axis rods?  I'm going to run some deflection calculations on these parts. 

[David Kessner]

Sorry.  A positive deviation indicates that the platform is higher (closer to the extruder) in that area.

The weird thing is that we can't easily conclude that the X rods are sagging in the middle.  If you look at the Middle-L/C/R data (-0.08, +0.09, -0.02) it does look like sagging in the middle.  But the rear row data (-0.19, (+0.01),  0.00) shows a different kind of sagging.  Looking at the "left side data" or the "right side data" shows something different yet.  

I almost think that if I loosen all the bolts on the metal frame, twist the frame, and then retighten the frame then I could get the four corners to even out.  After that I would be left with just the sagging of the rods (both X and Y rods), but it would be less warping than I have now.  I'll have to think about this for a while before I attempt such a thing.  And probably do some math too.

Both the X and Y Axis rods are 8.0 mm in diameter.

[Dan Newman]

8mm

Dan

[Dan Newman]

On 18 Mar 2013 , at 8:17 PM, David Kessner wrote:

> Sorry. A positive deviation indicates that the platform is higher (closer
> to the extruder) in that area.
>
> The weird thing is that we can't easily conclude that the X rods are
> sagging in the middle. If you look at the Middle-L/C/R data (-0.08, +0.09,
> -0.02) it does look like sagging in the middle. But the rear row data
> (-0.19, (+0.01), 0.00) shows a different kind of sagging. Looking at the
> "left side data" or the "right side data" shows something different yet.

I was allowing for some "slop". You had mentioned the plate being off
from true by nearly 0.04 mm and then there's some additional measurement and
reading error with the dial indicator. So, I was seeing some chance of
it being mostly rod deflection/non-straightness. But yes, this needs
paper and pencil.

All made difficult since you don't have an obvious reliable plane against
which to reference. Reminds me of what folks go through when they buy
a low end home mill or other CNC equipment. They then spend a month
relapping, truing, etc.

Dan

[Eighty]

Ok, this is going to be a very long answer.  First, I'm a structural engineer by trade.  But my expertise lies in big heavy things like concrete buildings, not teeny-tiny things like build plates.  Having said that, I took a quick stab at what kind of deflections we would expect on an idle printer.

 

I'm not going to bore you with too many of the calculations, although I'm willing to share them if anyone is interested.  I took the generally-accepted moduli of elasticity for acrylic and glass.  Figured up the moments of inertia, and modeled them as they sit on the platform.  Dead load only, assuming the plate weighs 1 lb.  Yes, I work in imperial units.  The Brits may have given up on it, but we Americans haven't!

An empty acrylic plate should deflect 0.029mm at the ends (less at the front, more at the back, due to the pad configuration).

An empty glass plate should deflect much less - 0.001mm +/-.

 

A pair of 8mm rods (at the gantry) weighing a total of 1 lb, plus a 1 lb extruder at midspan should deflect 0.05mm at the middle.  Dead load only, no dynamic load.  This assumes the standard "steel" modulus of elasticity of 29E6 psi.  May be stronger for hardened steel.

 

Long story short, the gantry should sag in the MIDDLE, where the build plates should sag at the ENDS.  So let's say we tram the center of the platform as our 0.0 point.  Theoretically, we should see as much as 0.8mm gap at the outboard ends (for acrylic).  More like 0.05mm for glass.  This assumes no long-term creep of any members.  But as we all know, acrylic and glass will creep.  Steel, not so much.

 

As I said before, these are some quick numbers, based on some generic material properties.  But even if we cut these deflections in half, that's a considerable amount - given that we're trying to print at 0.10mm layers.

 

These calculations marry in principle to what you're seeing.

 

Twisting the rods would be a good way to see if you can counterbalance any permanent deformation.  But the more likely culprit is the build plate.  It's not nearly as stiff as the rods.

 

I've been pondering a way to adjust the corners of the build plate.  Independently of tramming (over the adjustment points).  My thought is that you'd tram over the bolts, then adjust some arms at the outboard corners that would "flex" the build plate into a perfect match.  This would require some wings that hang off the adjustment platform at a steep angle, so fairly high relative forces would be involved.  But if you could take "any" build plate and make it match your gantry, then the prints would be perfect.  But my lazy-man solution is to just push up on the plate as the first layer goes down - everything sticks better, and I can move on with life...

[David Kessner]

Here is an interesting thought...  First, twist the frame of the bot such that the left rear and right front is moved +0.06 (up), while the right rear and left front is moved -0.06.  Next, adjust the screws so that the left side of the platform is raised by 0.03 and the right is lowered by 0.03. 

When I adjust the original data to have 3 data points on the front row instead of the original 4 (to make the data easier to manipulate), and factor in the above procedure I get the following table:

             Left     Center   Right

Rear      -0.10   +0.01    -0.09

Middle   -0.05   +0.09    -0.05

Front     -0.10   +0.01    -0.10

Now the data looks more like "typical rod sag".  But even then there is a 0.19mm difference from the lowest to the highest spot.  It is way better than the 0.28mm that I started with, but still isn't "good" (relative to a 0.1 mm layer).

Has anyone ever had to "twist" the frame like this?

[David Kessner]

Eighty, I sure appreciate the calculations that you've put into this.  I'm an electrical engineer, so when it comes to the structural calculations I quickly get lost.

So if I can summarize what you are saying, and factor in my latest stuff, what we can conclude is:  The predicted deviation from center to corner for a glass plate+rods is 0.05 mm.  Assuming the frame twist + re-tram procedure works then I'm predicting that the actual deviation will be 0.19 mm.  From this we could assume (?) that the rods are warped.  A twisting of the rods would be the next logical step to confirm or eliminate rod warping as a cause of the problems.

Is this right?

[Dan Newman]

Thanks for doing those comps. Interesting to have reasonable ranges to put
on effects we know are there but don't know the order of magnitude of.

> I've been pondering a way to adjust the corners of the build plate.
> Independently of tramming (over the adjustment points). My thought is that
> you'd tram over the bolts, then adjust some arms at the outboard corners
> that would "flex" the build plate into a perfect match.

We just need to borrow adaptive optics techniques from the 'scope people!
Put piezo stress sensors on the underside of the plate and then apply currents
to them to actually cause a deformation….

Dan

[Eighty]

I guess I'd have to think about the racking part for a bit.  Might cause other problems, like y axis rods binding up.  Granted, we're not talking about a lot here, but still...

 

How repeatable are your measurements after removing the build plate and putting it back?

 

I'd be very interested to hear what you find by rotating the X rods by 180 degrees.  Also, have you flipped the glass plate over?  Could be that it's "just enough" out of plane to make a difference in the opposite direction.

 

For what it's worth, my new acrylic plate (yes, the replacement for the original warped one) has developed a sag at the ends.  After my anchor goes down, I get a loose noodle until it gets to the center.  So my acrylic plate has been creeping over the past few months.  I have been meaning to flip it over, and see if it helps.  I suppose I should also store it flat when not in use, rather than sitting cantilevered like that. 

[Eighty]

Ooh, Dan, I like that!  Kinda like torturing our build plates into submission!

 

Would water boarding work as well?

[Dan Newman]

> Has anyone ever had to "twist" the frame like this?

Sort of. My Rep 1 with its wooden frame has actually twisted a little after a few days of
high humidity. I was having a levelling issue, but I no longer recall the details other
than it was vexing until I noticed that the two Z rods were not parallel: relative to each
other, one appeared to be leaning ever so slightly forward, while the other ever so slightly
back. I held the bot by its left & right ends and twisted and that sufficed for printing that
afternoon. A few weeks later, when the humidity had been back down for a while, I loosened
a good number of bolts and retightened and I've not noticed a recurrence (visually, or in
levelling issues). But, that was a Rep 1 with a wooden frame.

Dan

[Dan Newman]

I'll have to get an opinion from my attorney. Do we have to worry about
the People for the Ethical Treatment of Acrylic?

Dan

[Eighty]

Well, these steel frames aren't exactly built square. I take my printer back and forth from office to home, and always have to shim the front right foot. If I don't, it rattles while printing. Many others have posted little adjustment feet...and guess where I always see them? Front right foot.

[Eighty]

> Do we have to worry about the People for the Ethical Treatment of Acrylic?

Maybe under normal circumstances. But this herd has been infected with Acrylic Spongiform Encephalopathy, aka Mad Build Plate Disease. We gotta put 'em down, Dan.

[David Kessner]

After removing the plate, and reinstalling it I had to re-tram.  I'm not sure why, but the nozzle was touching the glass.  After that, using the same procedure as before, I took another set of measurements.  These measurements were within 0.04 mm of the previous measurements and the general pattern of highs and lows matched.  

I also flipped the plate over (swapped left and right sides) and repeated the measurements.  It didn't match quite as well, within 0.08 mm, but again the pattern of highs and lows was identical.

Tomorrow I will try rotating the rods and maybe racking the frame.  I can't do it now because I'm so tired I'm almost cross-eyed, and that's never time to do something potentially permanent.  

What is the correct procedure for rotating the rods?  I'm thinking of finding a section that the bushings do not touch and just using pliers on the thing.  That way if the rods get roughed up it won't effect the operation of the bot.  Maybe some sort of cork/rubber/leather padding on the jaws of the pliers?

[Jay]

You mean like a Handy Man with those POS Sony Controllers...gah! Was my first CNC Mill....

[David Kessner]

I tried pliers, both with and without masking tape on the jaws, and it didn't budge.  I marked them with a sharpie so that I could actually see them move, but they didn't move.  It might have been a leverage thing with the angle I need to hold my hand/wrist inside of the bot, so I'll try vice grips next.  Of course, I now need to run by the hardware store and buy some vice grips...

On Tuesday, March 19, 2013 5:09:51 AM UTC-6, Jetguy wrote:

The rods are at least case hardened. It takes some effort to mess up
the surface finish. A paper towel folded over a couple of times and
pliers would likely work fine.

[Joseph Chiu]

How is the R2 chassis put together?  I'm wondering if it's possible that your chassis is "out of spec" (assuming there's a spec) and was made just slightly undersized.   This is a totally wild guess, but if they are so tightly held that they can't turn, could it be that the rod ends are being squeezed together so tightly that the rods had to deflect?  

--
You received this message because you are subscribed to the Google Groups "MakerBot Operators" group.
To unsubscribe from this group and stop receiving emails from it, send an email to makerbot+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Enginwiz]

Hello David,

the X-Axis rods in my Rep2 were hanging down in the center of the build plate.

With a dead flat glass plate I could only print in the center area. Further out

the first layer height got too high and the first layer would not stick to the tape.

I was quite angry after this discovery and used brute force to bend the X-axis

rods upwards. You can actually bend a single 8 mm steel rods with your bare hands. 

Put your dial gauge on the X-rods, grab the first x-axis rod in the center with three fingers  

and lift this rod upwards. Hold the frame of the Replicator down with your

other hand and bend the rod about 8 to 10 mm upwards, hold for a second and release. 

Repeat with the second X-Axis rod to get equal bending.

Check the permanent change with the dial gauge and repeat until

the sag in the middle under the weight of the extruder in the center

position is compensated.

My Replicator 2 took no harm and now prints also on the left and right side

of the glass plate. 

[David Kessner]

So, I have an interesting result.  I turned the X-Axis rods 180 degrees.  Using a pair of vice-grips with some masking tape on the jaws I was easily able to do this without marking up the rods at all.  Unfortunately, the results are almost exactly the same as before!  So, apparently the rods are not warped at all, and something else is responsible for the approx. 0.3mm deviation from the middle of the build platform to the left edges.

I think I'm going to re-do the tests using this version of a dial indicator holder:  http://www.thingiverse.com/thing:23257.  My thinking is this:  if the weight of the extruder is making the rods sag, then removing the weight of the extruder should help.  I will remove the stepper motor, heat sink, and a few other things.  Shove the extruder to the side where it will have less of an influence, and then redo the measurements using this indicator holder and see if there is a difference.  If there is then the problem is with rod sag.  I also need to do some thermocouple "adjustment" so this might be a good time to tear the head apart.

On Tuesday, March 19, 2013 11:39:01 AM UTC-6, Wingcommander whpthomas wrote:

I can confirm that my gantry is at least 0.15mm higher in the back right corner than any other. It hasn't,t caused an problems so far, because I just avoid that corner, however I was planning to pull the side panels off and have a look at how one would level the gantry - relative to the build plate  - probably by loosening the retaining boosts and adding a shim.

[Eighty]

Interesting. But what's the point of removing the extruder? You can't print without it, so you might as well resign yourself to the fact that there will be some minuscule deflection. Strain shares a linear relationship with stress, so in order to bear a load then everything must deflect.
In my opinion (and its only that), you'd be better served by finding a way to compensate for this down below on the build plate, and moving on with printing. You don't really have any options up there, short of a massive upgrade to bigger rods.

[David Kessner]

You are absolutely correct in saying that the pragmatic approach would be to just fix it on the platform and get on with building.  But in this case I think there is value in knowing where the problem lies, even if it can't be fixed directly. 

The point of removing the extruder and remeasuring is to confirm or eliminate rod sag as a possibility.  Because if it isn't rod sag, then there are more things that need to be looked at-- like glass plate sag or maybe something that I have not thought of yet.

I am also intrigued by Enginwiz's suggestion that warping the rods on purpose might be a solution.  If it works then it's a good thing, but at the moment I think the risks cannot be ignored.  I will proceed with extreme caution on this one, for now.

As for fixing the Y Axis, I now think that the best solution is to loosen the bolts on the rod ends and then inserting shims and tightening the bolts back up.  This sounds much more controllable than loosening the frame bolts, torquing the frame, and tightening the bolts back up.  I'll probably do this work this weekend.