Ultimaker 3 Support Print Failures

[Responsible 4]

Hi all

I have been using our new Ultimaker 3 with varied success, but have consistently had problems with support structures. What has been happening pretty consistently is that the support tower does not bond well to the buildplate, and is knocked over during the printing, resulting in a big mess of PVA bits.

Is there a way to provide a bigger footprint at the buildplate for the PVA support? I feel like that would help a lot with this problem. But just as likely I am just too inexperienced at this point.

Using Cura 2.4(beta) and the Ultimaker supplied PLA and PVA.

Pete

[gr5 2,266]

Many of the UM3 glass plates are shaped like a hill where the center is higher and the edges are lower and especially the corners.  The bed is leveled at 3 points which tilts this shape and makes the front two corners good and the rear center good but leaves the rear two corners even lower.  Sometimes 2mm lower.

One solution is to just move your prime tower to the rear center.  You can set the position of the tower.  You have to set it in mm but it shows you where that is with a "tower shadow" so you get quick feedback.  But since the platform is about 200mm wide start with X position 100mm and leave Y position alone.

Another solution is to disable the tower.  The new temperature control in 2.4 beta 2 is very good such that there is very little leaking and you don't need the wipe tower so much anymore.

A third solution is to bend up the aluminum base.  Remove glass, bend up the rear two corners "a lot" but without breaking the heater (PCB) underneath.  I did this and it did wonders but later realized I could have broken the PCB.  The aluminum bed compensated and pushed up on the corners of the glass and now my UM3 glass is decently flat when clamped into the aluminum plate.  I bent with close to all my finger strength.  Maybe 50% full strength.  I don't remember. It took that much to make any difference to the aluminum.  When I was done the aluminum was about 2mm *high* in the corners and attaching the glass made them both cancel each other's error.

Edited March 14, 2017 by Guest

[neotko 1,417]

Indeed @gr5 bending the metal can work, but I think if the user really has the glass bend, he should get a new one by Warranty. I say this because while bending can work, is the worst I ever had. I say this because for example, on my umo+ all have close to perfect glass beds, I took out the one from the um3 and tested it on my umo+ (because one of my glassed did scratch) and clearly the glass is so banana is totally useless. But ofc, since I did bend the metal if I get a new one (I'm atill searching for a reliable source of glass with the same kind, no cheap glass) I would have a bed heater deformed.

So, don't bend it. Ask for Warranty, a glass bend on a 3k printer doesn't make sense by any standard.

BUT ofc, before doing that, update the firmware, since old versions get the z and level points less accurate and they have improved that. So if after updating your first layer sucks, ask the reseller for a fix.

[neotko 1,417]

Also you can expand the pva supports, I'm sure someone will post the name of the option, is a hidden parameter on supports that makes them horizontally bigger so they have a better grab on the bed.

[tomnagel 126]

The setting is called horizontal expansion, and is extremely important in e.g. PVA support printing.

[Responsible 4]

Thank you for the replies, I will check out the horizontal expansion settings. I have also discovered that orienting multiple parts so that the support structure 'overlaps' and merges is helpful.

Pete

[Responsible 4]

I tried the Horizontal Expansion setting adjustment on a print that was similar to the one that had failed, and it was a much better result. It was also a lot slower, but I need to play with the setting a bit more to find a good compromise.

Thanks for the tip!

Pete

[gr5 2,266]

You need "pva all the way down" meaning you want a path from all pva down to the glass without passing through pla or air. Horizontal expansion obtains this result for most prints (but not always).