Curved bed.....

I don't know if anyone can help. I have little experience of 3D printing - but, I am a scientist who routinely deals with microscopy and electronics.

My son and I have put together an Ormerod and have managed a few prints (including a passable orthogonal test piece) but we are struggling with both the Z-axis height and bed compensation.

It took us quite a time to realize that the Z-height probe (infra-red) only works reliably with the room blacked-out and the lights off. However, even with correctly functioning Z placement at the origin (over a white sticker) and with automatic bed compensation we cannot get the head to track at a reliable height over the bed. I had first thought that this was a problem with the compensation routine. But, I now think that is unlikely. We replaced the five screws (with their double nuts) which act as feet for the metal bed with manual adjustment screws. This has allowed us to get the bed closer to level (I wanted to do away with the need for bed compensation altogether). But, the head tracked too close to the surface in the middle of the glass plate (resulting in splutter) and too far from it on the outer edges (unattached cylindrical piping). It seemed like the glass plate was curved upwards in the middle. A little investigation revealed that the glass and metal plate are both flat when not attached to the heater bed. The heater bed, however, has a very noticeble curve up in the middle. We then added a washer to each screw between the heater bed and the metal plate to allow clearance for the curve. This has improved things a bit, but we still are not yet in a postion to print a first layer that is uniformly 'squashed' across the entire bed. We may need to put more washers in. I am however concerned with the thermal transfer since the metal plate is already hoovering above most of the heated bed. So, the question: Is this normal? Am I expecting too much by wanting a base layer of PLA that is uniformly piped onto the surface rather than a head spluttering in the middle of the glass plate and threads of unattached cylindrical extruded material at the edges? Would a new plate solve the problem? Or, would I be better-off getting a thicker glass plate which would resist curving. Or, are we experiencing some wierd non-linearity in the automatic bed compensation routine? Any help or experiences would be gratefully received!

Even if the heater PCB or heater spreder is curved, since the glass is so much stiffer it would not cause more than a tiny curve in the glass surface (and mostly the heater would be bent to the shape of the glass). What looks like a "bump" in the glass surface is probably actually the hotend "drooping down" a bit. If the acrylic x-arm is a bit uneven or flexes (from the pressure of the bearing on the X carriage), that would have the same effect. There is an aluminium replacement X arm kit, but I am not entirely sure what to do about it without that. It could probably be stiffened up a bit by gluing it together, or attaching something rigid to it.

Hi ChristofSchwiening

I have updated the instructions to show how to put the bed together to make it much more adjustable, and if there are problems with the MDF warping. There is a limit to the amount the bed compensation can do; it is better to have the bed physically level, then let the compensation do the final adjustments. See: [reprappro.com]
and (second and third step): [reprappro.com]
and: [reprappro.com]

With your 'bulge' problem, check that the x-axis arm is straight (ie not twisted along it's length) and flat (ie not bowed). If they are not flat, the nozzle will move up and down relative to the bed.

Ian
RepRapPro tech support

Hi ChristofSchwiening

Read this post and the rest of the tread

Bed leveling issue - how is this even possible?:
[forums.reprap.org]

Erik

Many thanks for all of the replies. I like the three-point support solution Erik. I will have to see how it works with the curved heated bed. I will try putting a washer between the glass and the heater at each clip - that should reduce any bending. I agree, jstck, the glass can't flex that much. So, I need to consider the x-arm. Am I right in thinking, droftarts, that if there is a flex/twist in the x arm then the angle of the extruder head will also change? Surely since the head moves on the rod it can't actually describe a downwards curve - any up or down movement must be associated with a lateral shift. So, if I print a long x-axis straight bar shouldn't it appear curved in the y direction where the head is moving? I will report back later!

Quote
ChristofSchwiening
Many thanks for all of the replies. I like the three-point support solution Erik. I will have to see how it works with the curved heated bed. I will try putting a washer between the glass and the heater at each clip - that should reduce any bending. I agree, jstck, the glass can't flex that much. So, I need to consider the x-arm. Am I right in thinking, droftarts, that if there is a flex/twist in the x arm then the angle of the extruder head will also change? Surely since the head moves on the rod it can't actually describe a downwards curve - any up or down movement must be associated with a lateral shift. So, if I print a long x-axis straight bar shouldn't it appear curved in the y direction where the head is moving? I will report back later!

As has been mentioned, what you believe is a curve in the glass is most probably the hotend tilting slightly at the middle of the X axis. The tilt moves the nozzle closer to the bed and so gives the impression that the bed is curved. If you clip the extruder assembly onto the first position on the X arm (closest to the Z support), it usually improves matters by relieving the weight moment on the X arm and alleviating the push of the Bowden tube at the centre of the X movement. Replacing the acrylic arm with a less flexible (twistable) aluminium part is a better solution - though a small difference is not at all serious, just adjust the Z zero position to be correct at the edges of the prints and put up with an overly squashed middle for the first layer. When slicing your own designs, set the first layer height higher than the other layers so that there is some leeway - the thinner the first layer, the more accurately the bed must be levelled (or compensated).

Try putting a straightedge (metal ruler) on the glass and see how much it is able to rock, which will tell you how curved (or otherwise) the glass is.

Dave
(#106)

OK, guys I have had a few hours playing. So, the X-axis arm was rather too easy to twist - so I have tightening it a bit. I can't tell which position is level, but I am now fairly convinced it isn't the problem. This morning I wrote a VBA script to control the Ormerod via the serial port and create a 3D infra-red reflectance test routine (or a proxy for height if reflectance remains constant). It was very easy to do. I traversed the complete bed by scanning X lines (initially at 20mm increments and then down to 5mm resolution) then shifting in Y. I polled the IR sensor 20 times in each location to get a decent average value. The plots are very revealing. There is a very clear dip down in the glass at the front left corner where the clamp holds the glass (stickers removed and glass-cleaned with Z height that gave a sensor reading around 650 for maximum sensitivity - windows blacked out with 'tin'-foil and lights off). I then removed the glass sheet and scanned the metal bed. It showed exactly the same profile (except worse). I then looked again at the heated bed. It has a 3 mm bend in it dropping down in the front left corner . I have now cut a full hole in the cardboard to make sure the thermocouple definitely isn't curving the bed but it makes no difference. I have attached the Excel sheetZplanescanning.xls with the scan of the metal plate in it - it also has the necessary macros for the scanning (although I have not bothered to make a user interface). If there is any desire for such a scanning tool I can clean it up and add an interface - since it currently has the x,y box coordinates, step sizes and averaging hard coded into it - so, please use with caution. It doesn't mess with Z height, so if you must click the button to get a scan, set your Z height initially well clear of any clips etc and check that the bounding box is compatible with your setup.
So, how do I go about getting a flatter heater bed? I have just bought this from RS. And, my power supply switch doesn't work - which means that I have to pull the plug to turn it on and off. Does someone cover that?
I really can't see how this could be the X axis twisting and drooping - why would have depend on the Y position?......

Quote
ChristofSchwiening
..There is a very clear dip down in the glass at the front left corner where the clamp holds the glass......

Have you adjusted the y-axis-end-plates?

Adjust the 2 inner screws so the bed is perfectly level seen from the nozzle in the inner position (X30 or such)

Run the nozzle out to the outer corners of the bed, any difference in these corners (while moving the nozzle in the plus minus y-direction) is caused by the outer y-rod, this misalignment if any, is mirrored in the outer bed corners

Use nozzle measurements to adjust the y-axis-end-plates so the rods are at the same hight in both ends

Erik

No, Erik I haven't adjusted the Y axis end plates. It is a good idea. To get that right I will write a Z height calibration routine for my bed level scanning scripts and then I will try and scan the MDF/wood bed and get that level. Then I will drop the heated bed on top of it (with cardboard) and clamp it tight and scan that. Earlier this afternoon I clamped the metal base to the MDF and ran a scan with bed compensation active. It produced a profile which had the bed compensation points level (i.e. all four corners) but a raised central section. So, I think the bed compensation works but the heated bed creates an upward curve in the centre of the metal plate which propagates through to the glass plate.

Christof, how much higher do you think the centre of the bed is than the corners? It's very difficult to bend glass. I measured a very small doming effect on the bed of my Ormerod, and I bought some 4mm replacement glass beds from my local glass supplier to reduce it (stiffness varies with the cube of thickness). But I'm back to using 3mm glass now because I can't see any difference in print quality, except perhaps on very large prints.

I'm currently using photo frame clips to hold the bed near to the corners, but I may shift them to be 1/4 of the way in from the X=0 and X=200 edges of the glass. That should reduce doming in the x-direction.

I always set Z=0 at the centre of the bed, using a differential IR sensor. If you know how much higher the centre is than the average of the corners, then you could allow for it even when using the standard IR sensor.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Dear DC42,
I don't yet know. The bend in the glass (when all 5 screws were attached) appeared to be enough to induce sputter in the middle of the bed and leave unattached PLA being pipped at the edges. To address the question I have just finished making a fully calibrated Excel VBA x,y,z scanning routineZplanescanningv0.2.xls (which I have now packaged in a more user friendly fashion). Would anyone like to give it a whirl? It can be used on anyone of the bed elements and produces some nice images with little user knowledge or intervention!

I have just run out of puff this evening and need to head to bed. The Tour is heading past our house on Monday, so I am putting-up scaffolding at the weekend. I may have to leave it all for a few days.
I will post some scans early next week!
Greetings,
Christof

OK, this is embarrassingly bad construction. I have strapped the bed to the front slider (cable tie) and made a 'calibrated' scan of the MDF bed.
It is not level! There is a 2 mm slope in the Y axis (at maximum X) although the linear bearing side (X home) of the Y axis is flat. The X-axis also slopes by over 1mm. So, the bed is twisted. Before making any adjustments I will attach the drilled aluminium plate on it and do another scan. Off to work now!

Edited 1 time(s). Last edit at 07/04/2014 05:08AM by ChristofSchwiening.

Just before I disappear. Attached is a scan of the heated bed resting on the "poor man's" aluminium support modification (i.e. I have placed the aluminium sheet on the MDF, cardboard on top and rested the heated bed on that). I have placed screws in each corner, but no nuts or clamps - it is just resting. The heated bed has about a 2.5 mm bend in it.

It's not the bed that's twisted, what you actually have is that the 2 y-axis ground rods are not parallel but have a twist between them. This causes the angle of the bed to change as it moves along the y-axis. To fix it, loosen the screws that hold the acrylic y axis end plates to the printed parts, then move one of the end plate up or down relative to the other to eliminate the twist, then tighten the screws.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

DC42, yes that makes sense - I had not thought about that. I can see that a pair of non-parallel rods would cause the far end of the bed to shift in Z as it moves in Y, independently of any Y bed slope - no wonder I couldn't get a flat bed compensation. It is all more complicated than I imagined! I am away from the printer at the moment, but I am looking forward to seeing if the twist is visible to the naked eye. I will be slightly disappointed if it is since I should have spotted it.......once it is flat I can concentrate on turning my scanning array into STL so that I can make copies of things.... A visible-surface object copier may make some of my tasks easier. Is Meshlab what most people use to do triangulation etc from x,y,z data?

Hi Christof

Sorry you are having problems with the bed. But great job on the probe! We had also thought of using the IR probe as a scanner, but there's a patent on that idea... nothing to stop you, though. However, I think the IR probe maybe a bit too unfocused to do accurate scans, and there's the problem of bumping into things with the arm, too. But I'm ready to be proved wrong... forum members are a very creative bunch!

Ian
RepRapPro tech support

Yes, I was wondering about the accuracy and focussing. The bumping into things is relatively easy to avoid if one starts scanning high and one knows what the 'head exclusion zone' is. The probe could then negotiate downwards in the safe areas. With a bit of interpolation, smoothing etc it might just be good enough. I have now aligned my parallel bars, added the spring adjuster and I will try some more plots of the glass.

Quote
droftarts
We had also thought of using the IR probe as a scanner, but there's a patent on that idea... nothing to stop you, though.

Ian, can you provide a reference to that patent? I'm wondering whether it covers scanning with a differential IR probe such as the one I am now using

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Well guys, thanks for all of the help. I now have a single layer of PLA beautifully stuck to the glass bed with no compensation at all - absolutely perfect. In the end it came down to a few tweaks. The metal plate to support the MDF, removing the screws on the corners of the bed at the far end of the x axis and using a single spring loaded screw, replacing the countersunk Y-axis end plate bolts with 'non-centering' standard bolts and washers to allow for the bars to be made parallel. Top notch support from the forum - now all I need is a power supply with a switch that works! My leveling routine for the Y bars was to place the head at the far end of the bed and with the motors off push the bed along the Y-axis whilst looking at the IR value. With the Y-axis plate slackened-off and the centering bolts gone it was relatively easy to tweak the acrylic into a position where the level was almost constant......then I used the bed screws to level the bed and checked the y-axis again. A bit iterative, but I didn't want to have to use any bed compensation.

Spend a bit of time tweaking as you have, and maybe a few tens of pounds on a few sturdy replacement parts and you will end up with a machine that works better than many costing well over 5 times the price. You also get to learn more about the way the machine works and mechanics in general than you probably wanted to know! Next comes the freedom of realising that you can create complex designs and have them in your hand in hours at little cost or effort, which I found really stokes the creative fire. Choose a 3D CAD design program with care. I can highly recommend the free downloadable "OpenScad" program. It uses text-based input for the design and so appears at first glance to be a poor choice next to programs with nice GUI front-ends, but once you learn the basics you will discover that it is far more versatile than the expensive applications, especially when it comes to making small changes to your designs. I used to use a professional CAD program at work that cost several thousand pounds, but discovered that I prefer OpenScad, and can design faster with that than the expensive program after making a few libraries to perform tasks that I use frequently (the libraries effectively take the place of the "tools" found in expensive programs for doing things such as applying draft or fillets etc. but additionally give you customised "tools")

Dave
(#106)