MiXley 3D printer - Designing a heated print bed

Door mux op donderdag 28 april 2011 23:10 - Reacties (10)
Categorie: 3D printer, Views: 18.395

MiXley 3D Printer - Designing a heated bed

There we go, the second part of our blog on a 3D printer design. Today we will be discussing the heated print bed - from the first design goals down to the actual design. First a little intro on what a heated bed is and why we want it. Oh yeah, the 3D printer is now called MiXley - a portmanteau and general mess of our names: Michiel (MiX - pronounced like TeX), mux and Wesley.

A 3D printer melts thermoplastic filament and then squirts it out onto a surface or product so it fuses together with the rest of the plastic that makes up your product. This is all fine and dandy when you've already got a bit of your product laid down, but what if you're just starting? The object needs to be laid down somewhere. That's your print bed. Well, most printers don't really print directly on the print bed at the start of each print. They first lay down what is called a 'raft' - a criss-cross pattern of plastic that elevates your product a few mm above the print bed. Why? Well, for starters, it thermally decouples the product from the print bed. You see, when you print the material it's pretty hot and as it cools down and solidifies, it contracts. When the bottom part is in thermal contact with the heated bed, but the rest of your product (obviously) isn't, it will cool down unevenly and warp or deform as a result. Another function of the raft is easy removal of your product; it has relatively little surface area, so it's much easier to peel off your print bed than a flat-bottomed product.

However, you can fix this another way. If you heat up your print bed, your product won't be chilled upon printing and consequently will not warp. This also improves the bottom surface of the piece. Of course, heating needs some form of power and we do want to heat the bed quite evenly. Also, we need to regulate the temperature with... well, temperature sensors. How do we know how to do this best? What are the design considerations? And will there be cake?
Current state of the art and problems
Any good engineer knows that before you start designing something, you need to do research. Either that, or you're lazy and actually hope other people have done your work for you. In the world of RepRap there are many people who just go their own way. That leads to infinite diversity, but also lots of wheels reinvented. I dove into the internet and the IRC to look what people had to say on the matter. And what I found was surprisingly useful.

One of the people in the reprap community who seems to be at the forefront of reprap innovation - nophead - has played quite a bit with the phenomenon heated bed and produced this here:

Nophead's heated bed
This is Nophead's mark 3 heated bed. Source

This is a type of heated bed I have seen executed quite a few times; a pretty solid chunk of aluminium with a few power resistors stuck on the bottom. These resistors are run at 48V - already considered an unsafe voltage - but they are operated from an isolated supply. So no real problems there. This design does have a few caveats; it's fairly heavy so for a Prusa Mendel-type machine (with translating heated bed) it limits your acceleration rate and maximum linear extrusion speed. Because of its weight, its takes a while to warm up, also since the power has to be converted, a good 20-30% of power is wasted. The bed isn't quite 100% safe; the chassis isn't grounded so a ground fault might cause a safety or fire hazard. There is no passive overtemperature or overcurrent protection. And lastly, at maximum heating power the heat isn't spread very evenly. The last point is largely mitigated by using a very large chunk of aluminium which acts as a heatspreader. The effect would be more pronounced with a thinner bed.

There are more designs. Prusa jr. himself uses a PCB heated bed. This design uses power resistors or just pcb tracks to heat up.

A copper trace resistance heater

The main upside of this design is that it is very light. The resistive tracks are spaced pretty closesly so the heat distribution is also very good. However, most designs don't implement proper safety (although Prusa has some very nice silkscreen on the back). Although grounding the heated bed isn't necessary anymore (because it is already electrically isolated), you still need overtemperature and overcurrent protection at the least. Unfortunatly the designs with pcb tracks are very hard to adapt to mains voltage. The copper resistance is just very low, causing way too much power to be dissipated in the tracks if you put 230VAC on them. Also, the gaps between the PCBs - if not properly taped over - can cause very slight surface quality degradation on the bottom if you print right over an edge. This design, though, is good inspiration.

Other designs range from using candle lights under the bed to using nichrome or constantan wire to heat it up. These options are low-tech but have fundamental drawbacks. For instance, taping heating wire do the heated bed the low-tech way doesn't provide a very reproducible design and causes safety and resistivity problems when adapting to higher voltages.

Now, I keep hammering safety and mains AC operation as important points. What I would like to achieve with the 3D printer design as a whole is a design that can hold up to consumer electronics standards. It should be safe to touch and even poke with a screwdriver, it should be forgiving and should have a good enough performance to use without having to plan too much. It should also be cheap. This is why I want to make the heated bed - the highest power component in the printer - work on mains voltage. Running on mains voltage means the power brick doesn't need to be rated a few hundred watts, but just the 30-60W that the stepper motors require. We don't necessarily save on heated print bed (HPB) cost, but we do save elsewhere, and significantly.
Consolidating the design requirements
Alright, we've seen some of the designs out there. We now have an idea of what we want to build. I've also spoken to a few people on IRC about the matter, and there are a few design requirements stemming from what I'd like to call 'good design practice'. Here we go:
  • The heated bed should have an area of 200x200mm
  • The heater should run off 230VAC (mains voltage here in Yurop)
  • Power and signal should be electrically isolated and on separate, safe and suitable connectors.
  • Heated print bed should be connected to mains earth
  • Overtemperature protection
  • Overcurrent protection
  • Protection should be resettable or able to be swapped without tools and with readily available components
  • The design should work with both PLA and ABS (require different HPB temperatures)
  • The print bed including all moving parts should weigh less than 0.35kg
  • The design should heat up in less than 60 seconds (at least for PLA)
  • Design should be easy to assemble
  • Design should cope with thermal expansion
  • Design should implement thermal sensors
  • Design should be reasonably cheap to make
Mechanical construction
Yeah, that's quite a list. Let's start with the basic construction and tell you exactly why we did what we did. Here's an exploded view:

Exploded view HPB

you see that in essence, the heated bed is built up around just three pieces: a bottom plate, four PCBs and a top plate. Let's start with the PCBs and top plate. This is basically inspired by Prusa's PCB heater, but with a twist. The PCBs have 24 1W 2512 size resistors each, with a bit of copper area to spread the heat. All the boards are interconnected. This means that the bed is heated very evenly. Since getting 200x200mm PCBs is hard, especially if you're on a budget we use 4 smaller PCB's. The four boards fit into each other like puzzle pieces. The PCBs are the thinnest and largest we can get cheaply; 0.8mm FR4 and 100x100mm - we can get these for under $3 a piece at iteadstudio.com. The aluminium top plate is there to provide an even printing surface and spread the heat a bit more. It is just 1mm thick so it adds very little to the (thermal) mass. The bottom plate - both to save weight and provide thermal isolation - is milled out of nylon. The PCBs are thermally connected to the top plate with some thermal compound and - this is the nice part - as the nylon bottom plate is bolted to the top plate, tightening the screws actually pushes the PCB against the top plate. If the bolts are then bonded with some Loctite, the construction will have good thermal performance as well as mechanical stiffness and immunity to thermal expansion.

Now we can roughly calculate the heat-up time by taking the thermal mass of 1mm of aluminium, 0.8mm FR4 and a bit of margin for the components and copper to arrive at a figure of about 200J/K of thermal mass to heat up. If we want to heat this up to 55C in 60 seconds (for PLA) assuming 25C starting temperature - we need 6kJ of energy delivered in 60 seconds, or about 100W continuous power. With 96 1W resistors in total, that is exactly what we're getting. If we'd want, we can choose slightly lower-value resistors and probably pump out twice that heating power without overheating the resistors (though they will go over spec, the 'junction'-to-copper resistance is low enough to keep the film temperature within spec).
Safety features
Next up is safety. First of all, one side of the PCB is connected to the aluminium, potentially forming an electrical connection with it if the thermal compound isn't spread evenly. There must, under no circumstances, be mains voltage on tracks or copper areas on the side facing the aluminium. The 0.8mm FR4 is rated 600Vrms isolating, so the PCB can handle 230V on the other side. A small solder connection on the aluminium is connected to the earth circuit on the mains connector, forming a positive connection between ground and heated bed. For overcurrent and overtemperature protection, a PTC resettable fuse is placed right in the middle of the heated bed, in a PCB cutout. This device trips at 1.25A or 85C, whichever comes first, and opens the circuit. After it has cooled down, it closes again. A separate glass fuse is not implemented.
Connectors and connections
Then come the connectors. For mains power, Molex Microfit jr. was chosen because both sides of the connector are shrouded (i.e. no open contacts), it is available in SMT, it has a positive mechanical connection, it is keyed and it is dissimilar both in pin pitch and form from any other connector on the build. This means even the biggest idiot won't be able to misplace this connector and put 230VAC on a part that is not designed to handle such voltages. This connector, via bendable cut-proof silicone wires, goes to a standard panel mount IEC connector (europlug) elsewhere on the static chassis.

We haven't actually gone into the PCB design proper. Here it is:

MiXley Heated Print Bed - schematic
The MiXley HPB schematic. Click on the picture to get a readable version. There's a lot of interconnects to the other PCBs here, it makes reading the schematic a bit uh... sketchy. Dude, I'm on a roll here!

PCB layout

Green is the top layer, red is the bottom layer. On top you mostly see a lot of resistors and copper areas. You can now see what I mean with the boards fitting together like puzzle pieces; picture a second board like this, rotate it 90 degrees to the left and try to fit it to the top of this board. Hey, that fits! On the left and top edge of the board you can find a few solid green areas; these are exposed solder pads that are used to connect boards to each other with a small soldered bridge. The NTC connections and 230V connections need to be connected through to each other. On the bottom-right you see the two connectors; one 230VAC connector on the left and one signal connector on the right. The signal connector breaks out to the NTC connections (on the bottom) and a few exposed pads on the right side of the board. Those last pads are used to connect the endstops to. And that's pretty much it; all those red wires on the bottom are just to wire the NTC connections through to the other boards. Oh yes, on the bottom-right you can see a bit of a red and blue oval shape; that is an isolation slot for the optoisolated SSR. Safety first!

The design has four PCBs, all the same, but with only one PCB outfitted with connectors and active components. The other PCBs only have the resistors and are loop-connected with the rest of the PCBs. On the middle of each PCB is an NTC thermal sensor, so on the whole of the heated bed are four separate sensors. This enables monitoring of the temperature in each sector. Furthermore, the mains power through the resistors is not constant; it can be modulated with the use of an SSR; essentially a solid-state isolated switch. The SSR has an optional snubber if testing reveals significant inductance in the circuit. The switch signal and NTC information is passed through to an SMT 10-pin shrouded standard IDC connector with strain relief.

Also on that connector are three inputs to the main electronics. These can be used for integrated endstops on the HPB. This saves on cabling and workmanship, and as such on total BOM cost. The third input can for instance be used for a user-operated button on the HPB or, for instance, an autocalibration point.
Odds and ends
Another deviation from the standard heated bed is the use of linear ball bearings instead of printed bearings. This should remove the need for spring stabilization and improve printing accuracy and speed.

One major problem with this design is that it requires both milled aluminium and milled nylon. Whereas the PCBs are pretty easy to come by these days, milling equipment is still expensive. The nylon bit can probably be printed (albeit in parts with a little bit of redesign). The aluminium bit is a bit harder. If one would need to make this without milling, one possibility would be to glue M3 standoffs to a 1mm aluminium sheet with the well-known pink 3M epoxy glue.
All this costs money, and it's interesting to see what this cost us and what such a design would cost you if you were to buy it in volume.

First there is the electronic components. We're getting them from Farnell:

Costing breakdown for MiXley HPB - table
This table is actually for three HPBs - divide the sum and parts by three and you end up at the actual HPB cost

Costing breakdown for MiXley HPB

That's on a total budget of about 30 euros for the electronics. The PCBs take up about half of the budget; mainly because we have to order 20 at iteadstudio.com and we're just using 12. At higher volumes, the PCB part drops to 35% of the cost and the total cost for the electronics is about 22-25 euros. This includes connectors, wiring and thermal interface.

Then there's the rest. The top plate is milled from 5mm thick aluminium sheet, while the bottom shell is milled from 8mm nylon (PTFE). Materials and milling don't cost us anything; Wesley takes care of that. For any other person, CNC milling this will cost about €70/hr plus material cost. Material cost for the aluminium boils down to about €5, as does the nylon. This drops to sub-€1 figures as you get to mass production. Milling time on these things is pretty quick; combined about 15 minutes a piece. For a one-off you will probably pay double, so say €40 for the milled parts. Small series go down to at most €20.

Then the bearings. We're probably going for four $2.50 LM8UU bearings; that's 7 euros total for one machine (i.e. 4 bearings).

This brings us to a grand total of 37 euros for us in small quantities, and about 50 euros for a complete HPB against material cost if it were to be produced in 10+ quantities. Not bad, but certainly not cheap either. It's all for science, though.
This concludes my treatise of the heated print bed design. What you've seen today is a very high-tech, mostly safety standards compliant, mechanically sound design for a 100W Reprap heated print bed that heats up in about a minute. Just this week the PCBs have been sent off to iteadstudio.com and I'm expecting them to get here around the 1st of May. Then we'll see how the quality is and do some safety testing!

Volgende: Een adapter van $4 - is het wat? 05-'11 Een adapter van $4 - is het wat?
Volgende: Building and improving a 3D printer (part 1) 04-'11 Building and improving a 3D printer (part 1)


Door Tweakers user Bjornski, vrijdag 29 april 2011 10:34

Very interesting reading Mux! Interesting stuff and very well written; easy to follow for an ICT nerd with only minimal electronics background.

I've been casually following the RepRap project for a few years now. This seems to be the ultimate adult-playing-with-lego kind of hobby project. :D (if only it wouldn't cost so much time...)

I do have a few questions though. Nopheads printbed incorporates 9 magnets in the printbead. Is this to hold a metal surface down, on which to print? Are you planning on doing the same?

Also, you're talking about milling the aluminum and it should be 1mm thick. Later on you're talking about milling it from 5mm. Are you milling-off all that aluminum to get proper bushes? If so, wouldn't it be easyer to cut the alimunim in such a way that it has (9 or so) flaps that could be bent into u-shaped brackets? Those u-shaped brackets could then hold an m3 or m4, which you could glue in place.
This seems to me as a mechanically sound solution that would save the relatively high cost of milling. What are your thoughts on that?

Nevermind, I only now noticed the 16 screw-holes with corresponding bushes in the top plate. Brackets wouldn't work as they could only be around the edges, which could cause the aluminum to warp when heated. :)

Anyway, keep up the good work!

[Reactie gewijzigd op vrijdag 29 april 2011 10:38]

Door Tweakers user mux, vrijdag 29 april 2011 10:50

I believe nophead uses the magnets to release a (steel) thin print bed so he doesn't need to put any force on the machine itself to pry the product loose. I.e. after printing he can take off the whole print bed including product and, independent of the machine itself, pry stuff loose. He can even put another print bed on and continue printing while he's fiddling with the previous print. I don't plan on doing this (my partners in crime might), because 1) it would add a lot of weight and 2) I think I can try to experiment with either flash-heating the bed to loosen the product or just designing for proper removal. It is an interesting concept, surely very handy for larger production volumes.

The actual assembly will be much clearer as we get the components back from manufacturing. And indeed, clamps around the aluminium are a total no-no because of the thermal expansion. The top surface is totally unrestrained in XY-direction. It is held down very statically undetermined by those 16 bushes in the top plate which are pulled into the nylon supports with M2.5-3 bolts so the individual connections are statically determined (and thus need only a dab of loctite to be held in place indefinitely).

[Reactie gewijzigd op vrijdag 29 april 2011 10:51]

Door Tweakers user delirium, vrijdag 29 april 2011 12:54

I am following this project with anticipation!

Door SA007, vrijdag 29 april 2011 14:09

I wonder about the thermal sensors.
Is there a place milled in the aluminum for them to fit info of are they crushed when you screw down the nylon?
You don't mention it in the replacement for the aluminum plate.

Also, it seems that the jumper wires for the ntc's are very near on of the resistors, is that within the 2mm required for double insulation? It looks like the air gap is smaller than 2mm.

Also, as far as i know you need to buy the contacts for micro-fit plugs separate from the plug itself. I don't see them in you component list. Also the tooling for these plugs is quite expensive and tricky to do it without the tool.

Besides that it looks like a very interesting design that also needs a very high powered soldering station to put together :)

Door Tweakers user mux, vrijdag 29 april 2011 14:27

SA007 schreef op vrijdag 29 april 2011 @ 14:09:
I wonder about the thermal sensors.
Is there a place milled in the aluminum for them to fit info of are they crushed when you screw down the nylon?
You don't mention it in the replacement for the aluminum plate.
Good point. There are 0.4mm recesses in the aluminium where the NTCs fit in, with a dab of thermal goo.
Also, it seems that the jumper wires for the ntc's are very near on of the resistors, is that within the 2mm required for double insulation? It looks like the air gap is smaller than 2mm.
closest points:

X1 50.500mm Y1 47.705mm
X2 49.350mm Y2 46.755mm

That is technically a UL fail, but I was working with IPC-2221 as a guideline, which (arguably) says 1.25mm for uncoated-to-coated clearance. I will need to put a bit of kapton tape over there to comply. Will put that in my workflow, thanks.
Also, as far as i know you need to buy the contacts for micro-fit plugs separate from the plug itself. I don't see them in you component list. Also the tooling for these plugs is quite expensive and tricky to do it without the tool.
Fortunately, I've got a few friends over at Formula Student who are willing to lend me the Microfit tool and a few contacts for this cause

Usually, those contacts are €0.08/pc in minimum order quantity 100 pcs, which would hurt the BOM cost. And yes, those tools are ridiculously expensive (and the hand tools are pretty slow to work with as well). poo
Besides that it looks like a very interesting design that also needs a very high powered soldering station to put together :)
700W hot air station, on its way to my home :D

Door Tweakers user breitsma, zaterdag 30 april 2011 23:54

How about using hot air or a hot fluid? Just make sure it moves around to spread the heat.

Door Tweakers user mux, zondag 1 mei 2011 08:37

Hot air is very wasteful - you're continually heating up cold air and blowing out hot air. If you'd want to make a closed circuit, it will be hard to find a fan that can reliably operate at elevated temperatures. Also, it will make for a heavy and bulky solution. A hot liquid [gas and liquid are both fluid] also has the agitator/fan problem, although arguably you could use a magnetic agitator. Still, bulky, heavy and I'm not convinced it is cheaper.

Another problem here is that I don't have experience with this type of design. You can't just put a liquid inside a closed box and call it quits; in order for it to be safe (i.e. not leak hot fluid or in case of a breach, cause further damage) the box needs to be able to deform under heat stress and not crack. The fluid inside needs to be something like oil, but that has stability issues (you can't really put standard vegetable oil in there, it will slowly deteriorate and produce precipitation, possibly clogging up the agitator). And on the other hand, you need to choose something that does not chemically erode your seals (which are probably either double-lip metal or silicone o-ring seals) or the fluid compartment themselves. Those are all areas I have absolutely no expertise in, but I do have sufficient knowledge about electromechanical design, especially if I can orthogonalize (decompose the design problems into independent problems) the mechanical design like I did here. That is why it didn't even cross my mind to use fluids for distributing the heat.

Door allert, dinsdag 16 augustus 2011 21:29

dutch CT issue April 2011 featured an article about the makerbot cupcake cnc, they mentioned that either keeping the bottom warm like you do, or cooling the lot with a fan so the plastic will cool fast enough to avoid tension between the layers. They ruined there's using another type of plastic.

Door Chris White, woensdag 27 maart 2013 13:23

Excellent write up - no conclusion?! How did you get on with the prototypes? I'd love to know if it worked out, as I was thinking of trying something similar myself...

Door Tweakers user mux, woensdag 27 maart 2013 14:41

Unfortunately there is no conclusion yet - we had planning troubles which meant we only have partial mixley machines - none of us have both the electronics and mechanics exactly as setup here.

The electronics are getting a second use in project Tosqa - http://tosqa.nl/

Om te kunnen reageren moet je ingelogd zijn. Via deze link kun je inloggen als je al geregistreerd bent. Indien je nog geen account hebt kun je er hier één aanmaken.