Bed Leveling/Warped Glass

[mendells 52]

I have always been fighting getting the build platform level and finally made a dial indicator holder and wrote a program to help me level the bed. Now the leveling procedure initially checks positions as close to the adjustment screws as possible, then checks the same pattern but closer to the center of the build platform. After that I have it "scan" the bed in the X and Y directions. I have dialed the bed in as close as possible but it is very apparent that the glass is warped. We have 4x Ultimaker 2+ and all the beds are wrapped to some extent but this one is one of the worst. I have sourced some thicker/flatter borosilicate glass that hopefully should be here in two weeks. I will update again once I have the new glass.

 

[mendells 52]

For comparison here is our extended with a cheap borosilicate glass plate.

 

[Dim3nsioneer 377]

Is that cheap glass really Borosilicate?

I have recently been in contact with a manufacturer of Borosilicate glass quite well known in Europe and they told me that their borosilicate glass is not manufactured with the aim to get even surfaces.

Standard glass is often much flatter than borosilicate glass. And I have no doubt you can buy a 'cheap borosilicate glass' in China which would turn out to be standard glass if checked with appropriate methods.

One more remark to the test method: please be aware you are testing not only the flatness of the glass but also the straightness of the axes (although Ultimaker axes are very straight nowadays).

[mendells 52]

Yes I am aware that this method is susceptible to other factors including the rods being bent or the frame bring off but it does show the difference between the flat glass and the warped glass. I have checked both sheets on a grant slab with an indicator and the ultimaker glass is significantly warped compered to the borosilicate. You are correct in that the borosilicate glass is not manufactured specifically for flatness but is is significantly better than the stock glass regardless. I have also noticed a change in the flatness of the ultimaker glass at higher temperatures (60-100C), but this may be due to the aluminum it's mounted to not necessarily the glass itself.

Edited February 5, 2017 by Guest

[mendells 52]

I finally got the new glass and it is far better than the stock Ultimaker glass. The original glass varied by about .020" and the new glass only varies by about .004" (I'll get a new video when I can).

I have also found that the parts also stick much better to the glass. I'm not sure why exactly but I would guess that it is because there is more even pressure across the bottom layer as the part cools.

[bastienb 12]

Hello @mendells !

So, what about this new glass ? Would you have the contact for your reseller ?

I also have troubles with my glassplates... Either it's already warped when I buy it (UM ones) or it gets warped through time (printing / freezer / printing / freezer twice a day for months). I better understand why Josef Prusa got rid of the glass ! 

[robinmdh 31]

this is one of the reasons why the bigger S5 has multi-point active leveling, it really makes a lot of difference.

Those anodized aluminum build plates for the S5 turn out to be even harder to get flat 😞

 

With glass buildplate suppliers some batches of build plates we had have been outside of specifications for sure! I'd still prefer the safe borosilicate glass to the flatter standard glass IMHO.

Edited May 1 by robinmdh

[geert_2 178]

I wonder what the effect is of heating the glass bed (or whatever else bed material) on the warping? Heating creates two temperature gradients in the plate: one from the center (hottest) to the outside (coolest) which is about 10°C in my tests; and one from the bottom to the top. I think both might contribute to warping, and cause the edges to curl up?

 

Further, since glass is a non-crystaline material if I remember well, it slowly deforms over time. Like honey that is sagging, or like asphalt, although thousand times slower. So repeated heating cycles and clamping cycles might also have an effect, as well as storage after production if stored on a non-flat surface or vertically.

 

@mendells: Maybe you could repeat those measurements with both a cold and warm glass, and then after cooling down again, for all the glasses you have?

 

If you could find ceramics like those used for high-grade computer chips and CPUs, dark brown, these might be more stable? I don't know if "ceramic glass" like used for cooking plates are real ceramics (=baked clay), or rather a commercial name for a different type of heat resistant glass (although non-ceramic)? Real ceramics should have a very small thermal expansion and be way stiffer than glass.

 

[yellowshark 132]

The vertical gradient is around 10c too I guess that depends on how thick your glass is.

[robinmdh 31]

@geert_2 the glass plate warps quite a bit while heating up, so we're changing to starting measuring after the heat up is done, it still moves a little bit afterwards but that is less.

I don't know if the movement of the glass has any effect on warping but indirectly it has an effect on buildplate adhesion since you're laying down your first layer with a slightly incorrect layer height.

The slow deformation over time should not cause any issues with multi-point leveling either, that movement is far slower and so far we level every print.

 

If you're interested I do have measurement data or a script to do active leveling probes while the buildplate is heating up and cooling down for/from the UM3, if you do this often enough you can get an idea of the general movements involved. we did make a movie of this somewhere, I'll see if we can find/post that if there's interest.

[geert_2 178]

@robinmdh: I have no UM3, and in my two UM2-printers the warping creates no real problems. Although I did notice indeed that the edges of the glass plate curl up a little bit in one printer. But I can position my models well enough in the center so that it doesn't hurt bonding to the glass. So I have no real need for accurate tables.

 

It is rather out of scientific curiosity that I would like to know the order of magnitude of deformation: how much does the glass curl up at 60°C, 90°C, 120°C? Is that in the order of 1, 10, or 100 microns? And how much difference there is between different types of glass or other materials? In aluminum it should be less, since aluminum is a very good conductor and all areas should have about the same temp. Glass is a worse conductor. Ceramics (like for CPUs) are good conductors and should have very little thermal expansion.

 

Annealing plastics releases their internal stresses, but it deforms the model. The same is true for metals. So I also wondered what repeated heating/cooling cycles would do with glass? Would they release moulded-in stresses from manufacturing? Or would they rather create additional stresses due to repeated uneven heating cycles, and due to the local very hot nozzle and melt, and the local fans cooling? All of these could contribute to deformations, but I have absolutely no idea which order of magnitude this would be. Although I could imagine that glass plates that are just in-spec originally, would go out of specs due to such effects.

 

[robinmdh 31]

@geert_2: from memory -> It's in the order of 100 microns on some points of the buildplate, on some locations it's below 10 microns, there appears to be a quick and a slow action to the point that the build platform below the bed is likely a big contributing factor. There is more change at the front of the printer then at the back in some cases. So not as much just the glass that that does do something but consider the whole printer.

It's also not really about being at a specific offset at a specific temperature but rather a warping of the buildplate during the warming up, because this does not happen evenly in the whole buildplate.

 

I don't see much deformation after the buildplate has reached temperature and been there for a little while.

[bastienb 12]

This "warped glass" is really a topic... As I wrote, the many cycles hot/cold I have done on my plates warped them (about 6 months before being unusable). And I need to print parts on the whole available printing surface (like hundreds of them) so the flatness is very important ( the tolerance is around first layer height for me).
I decided to change my two glassplates for the official UM2 ones. I received 3 plates altogether :

• 1 was OK
• 1 was bent (~1mm from one side to the other)
• 1 was curved ~(1mm from the top to the bottom of the curve, this one was meant to replace the previous one)

For the moment, the workaround I found is to level the bed so that all the areas of the first layer are either good or either squished. The feeder does not like it but I can print usable parts...

 

Maybe it could be interesting to work together finding someone that could manufacture borosilicate glass plate in Europe with good (garanteed) flatness ?

I personally only know a regular glass manufacturer, but it does not handle the temp change with the fridge...

[geert_2 178]

On 5/7/2018 at 9:52 AM, robinmdh said:

@geert_2: from memory -> It's in the order of 100 microns on some points of the buildplate, on some locations it's below 10 microns, there appears to be a quick and a slow action to the point that the build platform below the bed is likely a big contributing factor. There is more change at the front of the printer then at the back in some cases. So not as much just the glass that that does do something but consider the whole printer.

It's also not really about being at a specific offset at a specific temperature but rather a warping of the buildplate during the warming up, because this does not happen evenly in the whole buildplate.

 

I don't see much deformation after the buildplate has reached temperature and been there for a little while.

Thanks for the info. This less or more confirms what I already thought: 100 microns is about 1 layer of 0.1mm, if I am right.