Using MDF under the heatbed

Hey guys, just wanting to get some opinions on something. I'm going to design/build a printer and I'm planning to have MDF directly under the heated bed (touching it, not raised above it with leveling screws) for the heated bed to rest on, and to insulate the bottom of the bed.

Do you guys reckon it's a good idea? How do you think the MDF will cope? I doubt it would catch fire or anything, but would it warp, smell? I've seen cork used directly on the underside of the bed, not sure with how it's held up on the printers I've just seen it done.

If not MDF, anyone got any other cheap material I could use? My criteria is it needs to be able to stand up being right up against the bottom of the heatbed and ideally insulate it (definitely not a sheet of aluminium which would sink the heat...)

One of the main reasons for designing and building a printer yourself is to avoid the mistakes that others make. One of the very common mistakes is using underpowered bed heaters. If you're not going to use an underpowered heater, you don't need anything under the bed. 0.4-0.5W / cm^2 will heat the bed up fine without requiring any insulation.

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

I recommend using 0.4W per square cm and putting thermal insulation under the bed.

MDF will certainly warp, so I suggest you don't put it in direct contact with the bed or bed heater in case it distorts the bed.

---

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].

Thanks for the opinions so far.

@the_digital_dentist
I'm designing/building the printer for the fun of it, for me it's a project. I'm also on a budget, so 'underpowered' is what I can afford, building to a price point is better than not building at all (in my opinion). That and insulation never hurts, at the very least it'll stop the heat from going where it shouldn't. My electronics and motors will be going under the bed similar to the Kossel delta, so I think it might be an especially good idea. I could design it to use a powerful heater, doesn't mean I can afford it.

@dc42
Any ideas on what type of insulation? Glass wool or something else? Just before posting this thread I did some research on the heat resistance of MDF. A couple of places said it's quite temperature stable (which came as a surprise to me, and I was skeptical so wanted more opinions). I don't know if they were referring to different grade of MDF that is more temperature stable or what. If MDF isn't suitable, do you reckon some type of HDF like masonite would work? As for the actual print bed, it will be a piece of aluminium over top the bed heater. I'm hoping that will keep things moderately straight (worst comes to worst I can stick some glass over it, I'll be making the print surface removable).

Cardboard.

In that case, skip the heater and the insulation and it will cost less and be easier to wire. You can print PLA on blue painter's tape without a heater.

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

The type of insulation isn't critical. I use cork on one of my printers, and 3-skin corrugated cardboard on the other.

My comments on MDF stem from the experiences that I and others had of the Ormerod 1 printer, which used MDF for the bed carrier. It wasn't stable. Most Ormerod 1 owners replaced it with aluminium. The manufacturer switched to using plywood for the Ormerod 2.

As digitaldentist says, you can print PLA directly on to blue tape. Last week at the TCT exhibition, our small SCARA printer was doing a 400mm long print on to blue tape stuck directly to the surface that the printer was standing on. The PLA+ filament from rigidink is ideal for this, I find it sticks too well to most bed surfaces but it is ideal for printing on to cold blue tape.

Edited 1 time(s). Last edit at 10/03/2017 03:19AM by dc42.

---

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].

While I've seen cardboard used I was looking for a more heavy duty solution that could double as part of the frame, hence why I thought MDF. Would MDF be ok if it were sandwiched between two bits of aluminium?

Also I'm wanting to have a heated bed, not skip it out entirely for the costs. My question was mostly about if MDF was temperature stable and could be used directly under the heated bed as part of the frame. The insulation factor was a secondary concern/benefit. Any ideas for the structural side of things? Whatever it is, the be? will rest on directly so it needs to be temperature stable, not a heat sink and ideally cheap. Do you guys think HDF like masonite would have the same issues as MDF or would sandwiching the MDF suffice.

Edited 1 time(s). Last edit at 10/03/2017 04:07AM by Trakyan.

Experience with the Ormerod 1 told us that MDF is definitely not stable enough to use as part of a frame. I don't think sandwiching the MDF between aluminium sheets would help sufficiently unless the aluminium were thick - in which case you don't need the MDF anyway and you may as well use cork or glass wool insulation instead.

---

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].

What's the problem with standing the bed up on screws? Air is a great insulator- very light weight, and free...

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

Quote
the_digital_dentist
Air is a great insulator- very light weight, and free...

Only when it's enclosed so that convection currents are restricted.

I agree with mounting the bed up on screws - 3 of them, with levelling adjustment.

---

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].

The particular problem for me is that the bolts the bed is mounted also act as a pivot for the movement axis. I feel like the bolts should be clamped down tightly for this reason, rather than having the bed raised on them as 'stilts'. I could use washers between the bed and the rest of the frame so I can still clamp things down nice and tight but still raise the bed up.

That aside, since MDF is out of the question for the frame (really just a bottom plate for everything to screw into), what other materials would you guys recommend? Aluminium would get a bit expensive to have the whole frame milled out of need milled and the shapes I need don't really lend themselves to extrusions. Anything cheap and easy to work with that I could get some plates with some holes/slots cut out of? Also, how thick was the MDF used on the Omerod? Do you think a thicker sheet would have fared better?


Thanks for all the help so far though. I feel I should really elaborate on some of my design goals to make things clearer.
I'm trying to make it cheap, so beefing up the power supply, heatbed, motors and custom milled everything is out of the question for me. 3D printed parts of course are fine and make up a good portion of the printer. I'm aiming to minimize part count. Any part that does not absolutely need to be there will not be(part of the reason I don't really want to add separate leveling screws, springs etc.), this both lowers cost and makes assembly easier. All parts are as standard as possible, and minimal unique hardware parts. Aside from the obvious motors, electronics, hotend, print bed that every printer uses, I'm currently using 608 bearings, one size of M8 bolt (and nut, washer etc to go with it) and (possibly) one size of M3 bolt for mounting the motors. I know it's stupid but I thought it would be a fun challenge for myself, and the parts aren't really over/undersized anywhere as they serve a similar purpose throughout the printer. Like I said it's more of a project than a rock solid work horse, that'll come later when I have the money.

I'll be open sourcing the design once I'm done if anyone's interested, and no you wont be locked into the whole 'one bolt fits all' whack-job choice I made, it'll be parametric so you can tweak to your heart's content.

Edited 1 time(s). Last edit at 10/03/2017 08:31AM by Trakyan.

I don't understand the part about "pivot for the movement axis".

Is this a delta printer?

The bed doesn't have to stand on springs. Springs just make it convenient to relevel the bed with just a screwdriver, if the design/construction end up requiring frequent releveling. One of my printers used to have springs, but after a while I realized that it never required adjustment, so I removed the springs and replaced them with nuts. I can still relevel the bed if necessary, but now it takes a wrench and a screwdriver instead of just a screwdriver.

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

I'm doing a ground up redesign of the GUS Simpson, on that design the bed is held down by three bolts, one at each corner of the bed. The shoulders of the three arms swivel around those same bolts (from what I can tell, and even if not the case in the original design, they will be doing so in my revision). Because the part holding the bed down and the part the arms move around are the same, I don't think it will ever be drastically out of level.

On this machine, the bed is more or less an arbitrary plane because X and Y are not independent of Z and you're moving all three motors for a given movement. There is nothing particularly special about having it 5 mm lower on one side as far as how the arms move. On a cartesian machine, this would mean having to move not only X or Y, but Z to compensate for the level of the bed, so I can understand why leveling screws are useful. On this printer (or any delta really) the axis aren't independent anyway and you gain nothing by making 'level' and parallel to a non existent XY gantry or rails. Since all three motors are moving anyways, "level" is fairly arbitrary up to a point. On a cartesian, "level" means the build plate is parallel to your X and Y axis, and this is not arbitrary because of the reasons stated above. I guess you save in doing some coordinate transforms if you level the bed to be parallel to what the printer recognizes as the XY plane by default, but that's no big victory.

Also, looking at the Omerod, I can see why the MDF would sag/flex and cause issues, but would this still be an issue with a thicker sheet cut approximately into a triangle and supported at all three corners? I'd like to reitterate that my question was less about insulation and bed leveling and more to do with would MDF work for the frame, especially that close to the heat bed. Thick MDF is not an issue if that is what it takes since the bed will be stationary and wont need to hold up to any forces from accelerations like the Omerod y carriage would.

On a linear delta printer, it's very important to get the bed perpendicular to the towers, unless the firmware you use can compensate for bed tilt. Otherwise you will not be able to calibrate it.

Looking at videos of the GUS Simpson, I see what you mean about the arms being attached directly to the bed. I still think that putting MDF right next to the heater is asking for trouble. But I think you could use MDF for the base plate and stand the bed plate a few mm above the MDF on spacers, with cardboard, cork or glass wool between the two. However you do it, you will need to avoid the plastic arms getting too much heat from the bed.

Are there are firmwares that support the kinematics of the GUS Simpson natively? It looks like a fun thing for me to add to RepRapFirmware.

Edited 1 time(s). Last edit at 10/04/2017 03:42AM by dc42.

---

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].

The gus simpson's kinematics are nearly identical to a linear delta. It's the intersection of three circles its just you change the radius rather than center location. Someone did an implementation in smoothie, i think someone was working ok a marlin fork but never finished. I think ill implement it in repetier.

I'll probably go with the cork sandwhich you're suggesting. Also, i don't see why you couldnt calibrate for a tilted bed? Just apply coordinate transforms based on the tilt. Could you elaborate please, i'm interested in knowing more.

RepRapFirmware can calibrate for a tilted bed, but I don't think any of the other major firmwares can.

Does the Simpson kinematics involve trig functions? If so and my experiences with SCARA are anything to go by, 8-bit electronics won't be adequate for running the Simpson unless you are satisfied with low printing speeds. My Robotdigg SCARA printer was limited to 30mm/sec when it was using 8-bit electronics and Marlin, and even then it moved very jerkily.

Edited 1 time(s). Last edit at 10/04/2017 10:32AM by dc42.

---

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].

No trig functions, just like a linear delta its jusy squares and squareroots. I thought bed leveling was pretty widely implemented, or is bed tilt different to this?

Bed tilt on a delta is different, because of the effect it has on calibration (whether automatic or manual).

Looking at the Python preprocessing script, it appears to need to take 7 square roots per coordinate, rather than 3 for a normal linear delta. So doing it on a 8-bit processor in real time may be pushing things rather.

Unless you know of a simple calibration procedure for that machine, I think it would benefit from least squares auto calibration, which isn't available in any 8-bit firmwares.

Edited 3 time(s). Last edit at 10/05/2017 08:57AM by dc42.

---

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].

Unfortunately using a 32 bit board is not all that cost effective compared to an 8 bit board. I think I may be adding a Raspberry Pi Zero for octoprint anyway and I can use it to also pre-process gcode if necessary, and it will still be cheaper than a 32 bit board. As for a kinematics, both GUS simpson and linear deltas are effectively solving the following for each arm.

(x-x₀)²+(y-y₀)²+(z-z₀)²=(r)²

for linear deltas r² is a constant so it can be calculated once and then stored, and to move a point X mm vertically, you just move all three carriages by that amount, so a cartesian value for the Z can just be added on. otherwise you need do the x-x₀ and y-y₀ squares and then the subsequent square root.
That's three squares/square roots per tower.

For a gus, r is what you're changing, so you're not lucky enough to dodge the bullet by storing the value of r² and using that. Moving in the Z direction is also non trivial as the relative positions (or lengths, in this case) of the arms change, they don't just all move X mm up. So you don't dodge that bullet either.

Basically you have three squares, one for each X, Y and Z, and then one square root to find the radius from that, so four per arm. I've looked at the pre processor and from what I understand you only need a2, b2, c2 as they are your arm lengths/positions for a given (x, y, z) coordinate. Finding each involves 4 squares or square roots, so only one more than a linear delta.
The rest of the squares and square roots are there to calculate the distance moved from one point to another and how far each arm extends/contracts for each movement. This is because his script is (i think?) based off segmenting movements into equal sized segments of a given length so he needs to do calculations for the start and end point of a movement (basically do x2 of the kinematics), whereas firmwares usually (from my limited understanding) go off a certain number of segments per second, so the second kinematic calculation to work out the distance moved between points doesn't happen. Please correct me if I'm wrong because it's very possible that's the case.

references
[gist.github.com]
[github.com]

Squares are cheap relative to square roots.

But there are other solutions; e.g. the midpoint circle algorithm calculates the x,y coordinates of a circle without doing any multiplications, divisions, square roots, and only using integer arithmetic. I suspect that this, or something similar from the world of computer graphics, could be developed to calculate the inverse kinematics of printers with rotation-based mechanisms like the Delta or the Gus Simpson.

I'd never heard of the midpoint circle algorithm before, so thanks for bringing it to my attention. I imagine it could be used once for XY and then once more on an orthogonal plane for Z to create a sphere. That being said, I can't work out the specifics for myself. That being said the midpoint circle algorithm seems to be a forward calculation and I'm not sure how easily it could be turned into a reverse calculation for the all important inverse kinematics.