Hotbed flatness, voodoo or scientific method?

Hey, I have been struggling with my newly built printer hotbed flatness, from another thread I have learned 2 members use 1/4" Aluminum Cast Tool & Jig Plate which is very flat, one of which uses a kapton tape heater on the underside and kapton tape on the top, the other which uses the same 1/4" Aluminum Cast Tool & Jig Plate stacked under a PCB and a plate of glass on top. They both have less than 1 thousandth of an inch in flatness, this is amazing to me, especially since I have been struggling to figure out how to achieve this very same thing. I started this thread in hopes of us all sharing how we built our flattest hotbeds and I hope you will include every detail about them, did you use insulative material between the PCB and the aluminum? Do you use binder clips to hold the stack together? I am hoping this thread takes the voodoo out of achieving a flat heatbed for a lot of us, please help me to make it a reality. I am including a picture of my heatbed which is NOT very flat at all, yet!

Quote
lunarkingdom
They both have less than 1 thousandth of an inch in flatness, this is amazing to me, especially since I have been struggling to figure out how to achieve this very same thing.

Are you talking about being flat, or being level? There's a difference and it's the latter that's usually the bigger issue.

Quote
cdru
Are you talking about being flat, or being level? There's a difference and it's the latter that's usually the bigger issue.

Interesting. It's certainly being level that is more talked about, but being level without being flat seems both ill defined and not particularly useful. My MK2 bed is decidedly non-flat and I've been worrying about how I'm going to deal with that. I was wondering if the MK3s are typically flatter. If so, that might be the easiest way forward. At least standard 200mm sq beds are not particularly expensive.

I have an MK2 bed (definitely not flat) and a 3mm glass sheet on top holded with binder clips. The glass is flat enough to print on it and the rest is simply leveling the built plate and using a good and reliable auto bed leveling system.

Quote
cdru

Quote
lunarkingdom
They both have less than 1 thousandth of an inch in flatness, this is amazing to me, especially since I have been struggling to figure out how to achieve this very same thing.

Are you talking about being flat, or being level? There's a difference and it's the latter that's usually the bigger issue.

I am talking about flatness, being level is another issue but for this discussion lets only discuss heatbed flatness overall, I always have had issues with heatbed flatness, my current design is at least .010 higher on the outer 50% while the center is lower. Obviously my method is not the correct way to achieve flatness of the heatbed, I am very interested in those of you who have solved this problem and I hope you will share how you achieved this with us. I am sure it will help a lot of people get rid of a multitude of issues. The 2 issues I have seen are bad bed adhesion and more importantly when the space under the nozzle decreases enough the filament starts getting chewed up by the extruder. Having said that can we not address any other problems other than:

"What is the best scientific method to achieve heatbed flatness?"

Thank your for all the replies!

Quote
lunarkingdom
...my current design is at least .010 higher on the outer 50% while the center is lower.

.010mm or .010in? Huge difference.
.010mm I wouldnt worry about so I assume you mean .010" (.254mm) That however would make a noticable difference in your first layer of printing depeding on what your first layer height is set to.

yeah .010" or inches, it is a HUGE problem for a lot of us, I hear so much how flat peoples heatbeds are but never hear the how they did it part, it would be nice to know all the different methods people have used, pro's and cons etc.

Whether from out of level or lack of flatness is immaterial, keeping the maximum error within tolerable limits is most important. One easily overlooked thing is that these errors can change quite a lot between the hotbed being cold and being hot - and a little more as the heat reaches the attaching points.

Seems to me that an extra layer or two of tape at the low places, permanently attached to the bed, would solve the flatness problem?

Sounds like voodoo style to me, no offense but for example my new approach will be to use 1/4" Aluminum Cast Tool & Jig Plate that is 12 x 12 x .25 inches (it is in the mail as we speak) which has flatness of .001 or less to start with, I will then add my PCB then my glass on top of the PCB and measure it and see what I come up with. If it is within say .003 or less I will then try and add my glass fiber mat under the PCB to keep the heat on top of but not under the PCB and measure it again and see where it is. If it is less than .004 inches I will live with it.

That 1/4" plate sounds nice, but it will add a lot of mass to a moving print bed design. The simplicity of the Prusa layout is very appealing, but I keep finding myself drawn to a core-xy layout with the bed motion confined to Z. Maybe next time.

I was wondering about epoxying some pieces of aluminum U channel to the heat bed to increase flatness and rigidity, but I have my doubts about how the combination would cope with thermal expansion/contraction. My current plan is to have the heat bed track side up with a thin aluminum sheet to act as a heat spreader and glass on top as the print surface. I guess with that layout I could tolerate some warp in the heat bed itself as long as the aluminum and glass stayed fairly flat and there was still adequate heat transfer. Maybe I can bond the aluminum to the bed with a thermally conductive glue?

I've been pondering on this idea: Print the "shim" ie. put enough layers on the lower parts of the bed to make it flat. Then when you put the glass on this printed raft it shoulds stay inherrently flat as it can be even if clamped tightly. To accomplish this I'd start with zeroing the bed on the highest point (no paper should fit between the bed and the hotend tip there) and then print f.e. circles on the lower spots.

I guess I should now try this myself..

Cast aluminum tooling plate comes milled flat. It doesn't have the same crystal structure as extruded plate so it doesn't warp appreciably when heated. I don't know why you'd want to stack a bunch of stuff on the bed and hold it in place using clips or other clamps. Anything that stands above the surface of the bed is a hazard for your extruder. I use flat head screws that are countersunk into the bed.

My bed is supported by just 3 screws and springs. One screw is a reference and doesn't get adjusted, the other two are used to level the bed. Two screws are at the middle of the front and back edges of the plate. The one in back is the reference screw, the one in front adjusts the pitch of the plate. The third screw is at the middle of the left edge of the plate and is used to adjust the roll. Leveling is simple- move the nozzle to the center line of the plate near the back screw and set its height using a piece of paper. Now push the bed back so the nozzle is near the front screw and adjust the screw until it grabs the paper. That sets the pitch. Now set the roll: move the bed to the center of the left side of the plate near the roll screw and adjust until it grabs the paper. Finally, move the nozzle to the center of the bed and set the zero. Done. Total time about 60 seconds. I can print very nearly edge to edge on the 12 x 12.5" bed plate. Since the bed rides on fully supported linear guides, leveling and zeroing only have to be done if I take the extruder apart, which is very rare.

The only disadvantage to standing the bed up on 3 screws is that it occasionally resonates depending on the speed and makes noise.

Very nicely done - I remember looking at pics of your build. Out of interest, do you have a total cost that went into that?

Quote
the_digital_dentist
Cast aluminum tooling plate comes milled flat. It doesn't have the same crystal structure as extruded plate so it doesn't warp appreciably when heated

Exactly! As far as speed of the heatbed I put my new 17HS19-1684S nema17's on x and y and upgraded my pololu's for those axis to Think3DPrint3D Ice Blue edition's and achieved 100mm/sec last night with my 5 pound bed so I am no longer worried about "speed" with my heavy load. I will let you all know how it turns out when I get the parts in the mail. Anyone else have their experience to share of achieving flatness please chime in! This is a discussion I feel we all need to have! Thank you all for contributing!

Quote
JamesK
That 1/4" plate sounds nice, but it will add a lot of mass to a moving print bed design. The simplicity of the Prusa layout is very appealing, but I keep finding myself drawn to a core-xy layout with the bed motion confined to Z. Maybe next time.

I was wondering about epoxying some pieces of aluminum U channel to the heat bed to increase flatness and rigidity, but I have my doubts about how the combination would cope with thermal expansion/contraction. My current plan is to have the heat bed track side up with a thin aluminum sheet to act as a heat spreader and glass on top as the print surface. I guess with that layout I could tolerate some warp in the heat bed itself as long as the aluminum and glass stayed fairly flat and there was still adequate heat transfer. Maybe I can bond the aluminum to the bed with a thermally conductive glue?

What are you trying to achieve by combining all this stuff? Putting multiple thin layers of different materials together, each with different expansion and warp behaviors over temperature, is a sure fire way to never get the thing flat. An aluminum plate is simple, inexpensive, and works without any screwing around. My 12" x 12.5" x 1/4" plate cost about $30. Are you going to save $5 and 100 gm by using thinner sheet and a piece of glass, and some mystery adhesive?

Quote
the_digital_dentist
What are you trying to achieve by combining all this stuff? Putting multiple thin layers of different materials together, each with different expansion and warp behaviors over temperature, is a sure fire way to never get the thing flat. An aluminum plate is simple, inexpensive, and works without any screwing around. My 12" x 12.5" x 1/4" plate cost about $30. Are you going to save $5 and 100 gm by using thinner sheet and a piece of glass, and some mystery adhesive?

Yes, I'm trying to save money by using materials that I either have to hand or can source cheaply. I guess you're in the US where you have the great fortune of a multitude of suppliers and frequently free shipping. A quick check of my usual sources, the best I can do for 12x12x.25" cast aluminum is $103 shipped. That's USD which translates to $125 CAD. To put that in perspective my total spend so far is less than $300 US, and I believe I have everything I need.
My impression is that more people are printing on glass than on aluminum, but I'll keep your advice in mind. I have some pieces of glass available and it seems like a nice feature to be able to have multiple pieces that can be quickly exchanged, so I'll start with that. It's ordinary glass, not borosilicate, so I'm planning on using aluminum sheet between the bed and the glass to ensure more even heating and hopefully avoid cracking.

What do you use to prep your aluminum plate, kapton tape?

James.

Yes, just Kapton tape. I use 5 mil thick stuff because it is much tougher than the usual 1 or 2 mil stuff and doesn't need replacement very often. I use 2" wide tape so I can replace only the part that is damaged instead of recovering the entire bed.

Cost? Let's see...
Frame extrusions, drawer, side panels, feet, maybe $250
Extruder: $150
Hot end: $60
Y axis motor: $50
Bed plate, heater, power transformer: $100
Linear guides for X and Y axes: $180
Smoothieboard: $140
DSP drivers for X and Y axes: $80
Z-axis assemblies, Y axis ball screw, power supplies, motors for X and Z axes: $0 from the makerspace
Misc other stuff: $100

Total is about $1110.

Edited 1 time(s). Last edit at 05/07/2015 12:19AM by the_digital_dentist.

Yikes we are going off topic, the purpose of this thread was to hear from people who have achieved acceptable flatness as in under 5 thousandths of an inch of their heatbed, I was hoping to gain some clarity for all of the people starting out so we all do not have to reinvent the wheel over and over again when it has already been invented.

Ahh thanks digital dentist, I was wondering as well.

The easiest and less expensive way to achieve "Flatness" is with your glass surface, because at the end it will be there where you are going to print, unless you wanna print on top of your expensive and delicate aluminum plate.

there is several ways to "flatten" your glass sheet, from dirt cheap to high tech ones (I use the dirt cheap method) a flat surface, two glass sheets and fine grinding compound. put one glass on the bottom, spray water and the grinding compound put the other glass on top a rub until the compound is evenly spread on the entire surface of the glass, lift the top glass and apply more compound and water to the center put the glass and repeat 2 more times, then do the same for the opposite side of the glasses.

*Now, even though what I just explained is a waste of time* (glass is flat, really FLAT) it will expand and deform with heat same as Aluminum and other materials alike, and since our heating systems are sooooo efficient with high temperatures at the center of the plate and cooler at the edges, you will always have a higher center no matter what your printing platform is made out off. *The grinded surface helps a lot with first layer bonding to the glass surface and less glue is needed.*

After letting out my sarcastic half and joke some with you. Now on a more serious note, i'm talking with a Chinese manufacturer in order to embed a heating element in between two borosilicate glass sheets. it will make for an all in one print platform.

I like this idea.

But why sandwich the heater between two borosilicate glass sheets?

I'd have thought it would save about 50% of the electricity if the heating element was between a thermal insulator and one glass sheet. Especially if the insulator was a bit rugged/resilient to protect the glass if it was dropped.

Even better would be if the insulator had the same coefficient of expansion as the glass. Or would that just mean that the insulator (and glass if attached to it) would bow if it was heated more in the centre than at the edges?

Frank

People keep trying to squeeze cost out of these machines pretending that it doesn't affect the quality. There's no magic or voodoo. Precision parts cost more than junk parts. The solution to the problem of a flat bed is to spend a few more dollars and buy a flat, thermally conductive aluminum plate. Stick a heater to it, slap some Kapton tape on it, and move on.

If you were going into mass production, you'd spend as much time as necessary to shave a dollar or two out of the cost of the machine, but you're only building one printer. What's the point of spending many hours trying to save $5 on the cost of the print bed? Is your goal to say that you built a 3D printer for $150 from which you can occasionally get a moderate quality print after a lot of messing around? How much is your time worth?

Every hobbyist printer and kit maker is working to drive the cost of these machines down, taking out quality and reliability with each dollar they manage to shave out of the cost. First they make crappy frames and use under sized end supported guide rails that flex if you look at them sideways. Then they add autotramming and call it a feature when in reality it is a band-aid solution to a crappy build. They do things like zip-tie bearings to extruder carriages. They use threaded rods where lead screws should be used and people have all sorts of Z axis problems as a result. They use two motors to drive those threaded rods instead of a single motor and belt and the result is that the X axis alignment can go out any time someone touches one of the screws, or you have a wiring problem, or you have mechanical binding on one side of the z axis. They sell this as a feature: it's really easy to align the X axis! Every compromise they make to take out cost adds to the time you're going to have to spend trying to coax the machine into producing a print.

Why not make your goal building a better printer, instead of a cheaper, crappier printer? Don't copy the compromises made by the people trying to sell $300 printers. Look at professional/industrial machines to see how to do things right and copy them instead.

To go back (almost) to lunarkindom's original question: How much error in flatness/level is tolerable? What is the effect of too much error?

My own observation is that a very low error seems to be most important is single part prints covering a large area and printed in ABS. Accurate and consistent control of the first layer seems to be as important as anything else in getting good adhesion and minimizing curling. For the standard fare - Yodas, Owls, vases etc. a quite astonishing variation in the hot end to build stage gap seems to be O.K.

The effect of the error has been experienced by everyone here- if the nozzle is too close, the extruder jams. If it's too high the part doesn't stick.

Ignoring flatness for a moment, adhesion depends on the surface material of the bed, its temperature, the size and shape of the object, the material being printed, hot-end temperature, and environmental factors like temperature and air movement. You also have to consider mechanical issues in the printer- is it able to maintain a constant Z elevation throughout the range of motion of the X and Y axes? If the bed and extruder carriages move in arcs because the the end supported guide rails sag, you may to have a hard time getting prints to stick anywhere but near the center of the bed.

In the slicer you specify the first layer thickness, with a number like 80% of nozzle diameter being considered the maximum (in Cura, anyway). If your nozzle is 0.25 mm that means maximum first layer setting should be no more than 200 um. Obviously, a 200 um deviation from flatness is bad, but if the nozzle is less than 200um away from the bed the plastic will get squished out and it will work as long as the extruder doesn't jam. An extruder jam is going to be a function of the material and the hot end temperature. If the first layer is printed too close to the bed the excess plastic tends to make ridges. On the next layer, as the nozzle bangs into those ridges they may be knocked loose. So there are many variables on the "too close" side of the problem.

Going the other direction, if the nozzle is too high the plastic will appear to be under-extruded and there may be gaps between the lines laid down. This will affect how well the model sticks to the bed. Straight lines will tend to pull off the bed more than curves. Small objects will pull off easier than large ones. The shape of the nozzle may also have an effect- if the hole is smaller than the brass area at the tip, the nozzle may push on the previously laid down line and tend to either remelt or knock it loose. So there are many variables on the "too high" side as well.

So how much deviation from flat is acceptable? Your guess is as good as anyone's, but it is reasonable to say it should be less than +/- nozzle diameter.

All this assumes you've leveled the bed properly, meaning adjusting tilt so that the deviation from flatness measured at the nozzle tip is uniform, and that your printer can maintain that setting. The best thing to do is minimize the errors that are most easily controllable- start with a flat bed, minimize temperature variation over its surface by using a thermally conductive material (aluminum, anyone?), regulate the temperatures of the bed and nozzle using PID control, build a solid printer that's able to maintain the level setting throughout the range of motion of the X and Y axes, calibrate your extruder, and control air movement and ambient temperature by enclosing your printer (hey look, all that stuff that the $300 printer kit people took out!).

Edited 3 time(s). Last edit at 05/07/2015 08:42AM by the_digital_dentist.

Quote

Even better would be if the insulator had the same coefficient of expansion as the glass.

This is exactly the reason why it is in between two glass sheets. If any of the sides acts differently, there will be deformation and stresses that could lead to separation of the sheets or cracks in the glass.

Quote
ggherbaz
This is exactly the reason why it is in between two glass sheets. If any of the sides acts differently, there will be deformation and stresses that could lead to separation of the sheets or cracks in the glass.

Sounds good - look forward to hearing how it goes!

Quote
the_digital_dentist
build a solid printer that's able to maintain the level setting throughout the range of motion of the X and Y axes

So what you're saying is that using a pink granite surface plate with a tolerance of .000050" might be overkill with my printer made from threaded rod and cheap Chinese linear bearings on drill rod with a hobbed bolt extruder made with a tap with a hand drill?

My .02, just use a piece of plate glass and maybe a piece of tool plate. Buy a decent quality piece and spend your time and money elsewhere. If we needed .0001" tolerance, we wouldn't be in a RepRap forum building printers out of threaded rods, motors ripped out of discarded electronics, and old ATX power supplies...

As the_digital_dentist points out, another error is in the nozzle height is sag in the guide rails - if you add a 6mm aluminium heat spreader plate which is 0.025mm flatter than a glass plate but its mass causes 0.1mm of extra sag then the improvement has not been worth it.

On one of my printers I do not use a heat spreader at all, just a headed bed clamped directly to a glass plate. Although there is movement during warm-up and I have to level the bed quite often, I have not found any problems which could be put down to either lack of flatness or poor thermal profile.

Edited 1 time(s). Last edit at 05/07/2015 10:12AM by leadinglights.

Quote
cdru

Quote
the_digital_dentist
build a solid printer that's able to maintain the level setting throughout the range of motion of the X and Y axes

So what you're saying is that using a pink granite surface plate with a tolerance of .000050" might be overkill with my printer made from threaded rod and cheap Chinese linear bearings on drill rod with a hobbed bolt extruder made with a tap with a hand drill?

My .02, just use a piece of plate glass and maybe a piece of tool plate. Buy a decent quality piece and spend your time and money elsewhere. If we needed .0001" tolerance, we wouldn't be in a RepRap forum building printers out of threaded rods, motors ripped out of discarded electronics, and old ATX power supplies...

So where would we be if money was no object? This is the largest 3d printing forum I have found, is there one with more resources? If there is let me know and I will take my $1300.00 (so far) build over there too. On a more positive note I really like how the subject is progressing, lots of great knowledge shared! It seems to be really going somewhere now, if we only had some pictures ;-)