Inline pics to be added later.

I have been printing on a ceramic floor tile with decent results as long as it was warmed first (heat gun). As I didn't want to destroy my only working printbed, I looked around for something similar locally. Nothing was available for polished, flat tiles. There was a special name for the tile I had... basically it means it was polished until 'optically' flat. Reflections are not distorted... really, actually, flat. Refractory? Something like that.

Instead, I found the stone tiles. They were also polished (ground really) flat and were ~1/2in thick. They didn't say 'refractory' or whatever like my tile (which I could have ordered, but had to get a box of), but the reflections from it were not distorted. Good enough. They were also dirt cheap on clearance, and I could screw up 3 times.

For the ceramic tile, I was going to glue 1/4in nuts to the back, wind on those, and then encase the windings in cement. The extra thickness I now had gave me the idea of cutting grooves in the back to make my windings embedded. The back of my tile could be flat then, this should increase heat transfer, and I thought that heating the tile from the 'middle' may help eliminate any tendency to warp.

I wonder if I can cut these? Stone is hard, requires special blades, you need water cooling right? Nope. It was easy.

I had a masonry blade I bought a while back for my circular saw. The arbor also happened to fit my table saw. On it went.

Figured out a pattern which would give me a gap in the middle and a little wider edges. It was start at 1in, then 3/4 apart all the way across, and ending with 1in. I gingerly started to make the first cut, at 1/2 my intended depth. The dust was overwhelming at a few inches in, but it actually cut easier than hardwood plywood minus the blade not tracking at all. I started to get excited that this might actually work, then stopped everything and 'suited up' to be blasted with dust for hours. Respirator, goggles, hoodie tied shut, and gloves held on with rubber bands were donned. I've heard of people getting seriously ill, long term, from stone dust. Silicosis I believe it's called.

Cut the first few grooves at half depth. Was worried about heat buildup causing the stone to crack. Then I got impatient and switched to full depth on the first pass (half way through the tile). I figured, hey... I got 3 more tries if it explodes. It didn't. It cut like butter.

The tile I was using hung off the edges of my base and looked funky. Since this stuff was cutting so nicely I decided to see if I could get a circle out of it. Drew a circle on the back with a pencil, and free-handed some relief cuts close to the line. Then I got a little cocky and tried to free-hand the circle itself. It worked pretty well. The edge of the blade is a good grinder so you don't have to be perfect the first time... just stay on the right side of your line. :)

I needed a groove around the edge to allow my traces to turn around inside the tile. It was cutting so nicely, I decided to just try to free-hand that as well. Turns out free-handing a cut isn't so easy when you can't see the blade. It also tended to 'grab' when I missed the previous cut. The cut I ended up with wasn't useable for winding as it had sloped edges.

I started over.

Did all the straight cuts again, and this time I pre-drilled a hole almost through the tile right in the center for popping in my thermistor later on, and cut the extra slot to run the thermistor wire almost to the hole. Took a bit of plywood, put a pin in it, and clamped it to the saw with the pin located directly across from the top of my blade. Cutting a curve with a straight blade will probably be difficult. This time I took my time. Started with the blade down, turned the saw on, and raised it until it contacted. Rotated the tile. Raised the blade, rotated the tile... repeat this about 15 times and I was through. Not so hard..

Moved the jig over a half inch, and cut the edge groove I needed half way through. As this was the last cut, I really took my time on this one. Whew... hard part done.

I was aiming for 250w/500w of output. I only have the one 12v supply, but it's rated at 1000w so I'm good there.

Initial attempt at winding it used stainless steel wire as the heating element. I tested about 3 feet of it it on a car battery and it got cherry red relatively quickly, and didn't immediately develop a hot spot and disintegrate like normal steel wire does.

I spooled out 10 feed of it and got out my ohm-meter. The resistance per foot much higher than I expected, at like .5 ohms. That's way too high for the wattage I wanted, so I decided I would just keep adding parallel strands until it was in the right range. Winding the stuff it into the slots I cut proved impossible though even for one strand. It was too springy for the small area it had to stick to at the ends. I tried creating a back cut with an angle grinder where it was needed, but the top would chip away making it worse. Back to the drawing board.

The only other stuff I had on-hand with a high melting point was copper, I only had a couple of gauges, and almost all of it was stranded.

I measured (guessed) the length of wire it would take to wind it. About 36 feet would do it, and eventually I ended up here.

Wow... copper is a good conductor. I didn't have anything small enough. Oh yeah I did... I unplugged the network cable from my printer. It was 18 feet long. Long enough if I center tapped my winding.. I tore it open, untwisted a pair using the drill, and stripped the insulation. The insulation did not come off easy. It was bonded to the copper, so I had to carefully drag a razor along the entire length to get it off. I removed some copper (evenly) in the process.

18 feet of solid 24 gauge copper would give me a resistance of 0.411 Ohms. Driving it with 12v (12x12/0.411) worked out to be 311w. There was to be two elements. 622w would be a bit much. I decided to just center tap the windings and figure out if I wanted 311w or 622w of output later.

Winding the copper was easy. It pretty much stayed put just by bending it tightly around the ends. I wound it before I crimped on some wires, which I was still able to do without it flying apart. I had a few feet left over. Uh oh. That's going to drop my resistance... I don't think I want a 1000w heated bed, so I didn't bother to hook up my center tap. My ohm-meter is not really accurate enough to measure something with this low of resistance, so I really didn't believe it when it read (compensating for the test leads) .3 ohms per winding. I was hopeful though. I didn't remember how to solve resistor networks off the top of my head to make something more accurate I could actually measure. You don't use it, you loose it.

Here comes the messy part.

Furnace cement works by being highly thermally conductive, which should be ideal for embedding the heating element and filling in the grooves I created. I also knew it got hard when dry, and could bond things together reliably at high temperatures. It has the same chemistry as (sodium silicate based), but is 10 times cheaper than, the 'muffler cement' I had used before for making hot-ends. Muffler cement got pretty hard, but I'd never used that much of it on anything before.

Sprayed my tile with water per the instructions, and spread on the cement. Ran a screwdriver down each groove to make sure they didn't have any voids. A void would mean my wire would not have a heatsink at that spot, and it would burn out the first time I turned it on. It went on a lot like toothpaste, with sand in it. At the last minute, I dug out the center tap I had made, but not hooked up.

It dried like drywall mud you put on too thick, but slower... outside shell, inside liquid. Many hours later not much had changed. It was still squishy. Put it in front of the garage heater and went to bed.

The next day.. it was still squishy. It said to allow 24 hours to dry before subjecting to high heat. Well, high heat for this stuff is 2000 degrees.. so... I stuck it in the oven at 195F. It produced zero fumes.

I left it in there an hour. It was drier, but still squishy. I was hoping I wasn't wrong about this stuff at this point.. what if it never sets up? I turned the oven up to 220F. My hookup wire said 105C on it, so that's as high as I should probably go..

I came back 3 hours later. It was rock hard. I broke off some flashing from where I taped the edges. It was like if you made cement with epoxy. I was previously worried that my wires moving about would break the stuff. Not a chance.

I let it cool.

I hooked it up to a car battery. Sparks... that's good.. I left it hooked up.

I came back 10 minutes later. It was barely warm. I turned my meter to 10A range and tested it... over scale... Well that means it's at least 120w... I hooked it up via a 15A automotive breaker. It popped it, but not immediately. So it's more than 180w... I hooked up a 25A automotive fuse. It survived. Well, it's less than 300w. Well, we are in the right range. I'm going to assume my meter was right then and it should be outputting 240W.

Shesh... that should get hotter than this... Well, it is pretty big, it's not exactly warm out here, and it has a lot of thermal mass.

At this point I'm really glad I dug out that center tap before the cement got really hard. I'm also glad I used 8 gauge hookup wire to start now. The 8 gauge was fine stranded and nice and bendy. Being able to move/remove it was important.

The next day I made the wiring changes needed to run it center tapped. 500W was close enough to my target that I should be good with everything else if I'm lucky.

I left it cure for a while, (8 hours or so) then hooked it up to my printer power supply. It was pretty obvious I was getting a different result. It sparked more on hookup, and my server power supply shortly turned into a jet engine. The build surface got up to 70C in a matter of minutes. That is a lot of mass to be heating that quickly. I'm guessing, since my 8 gauge leads are heating a little, that my test battery must have been 'old' and I'm really in the 700W range.

As I've already cut the leads and encased the changes in furnace cement, I decided to use the center tap permanently, at 700W. Since the supply can handle it with everything else powered, I just need to add a few more mosfets. I tried the existing mosfet by itself, briefly. It worked, and didn't pop... but it got really hot in about 15 seconds while well heat-sinked. I'm beyond its thermal capacity or I'm not driving it hard enough. I'm just going to add more mosfets in parallel versus trying to cool/drive the one.

Fast forward a week. Stole two more of these mosfets from another project and made a water block with my cooling system running through it. Tried it out. I didn't turn on the cooling pump and was greeted with the sound of boiling water in about 30 seconds. One mosfet then popped and I shut it down.

Fast forward about 2 weeks. I stumbled across my stash of low gate voltage power mosfets. They were purchased specifically for use with the beaglebone: minimal gate current, fully on at 1.8v with a max gate voltage of 8v. I'm driving them at 3.3v. I replaced the old mosfets with three of these in parallel. Result was minimal heating of the mosfets now. They stay cool to the touch without the coolant flowing through the block.

One issue has happened since, which prompted some serious thought into a physical fail-safe.. The code which reads the thermistor locked up while I was printing, and I dozed off in the room.

The bed ended up fully on (within the limits I thankfully set for the PRU based PWM).

I woke up to the entire room being heated to 80 degrees. It is a really good thing I had limited the PWM duty cycle of the bed to 60 percent, and that the PWM code runs on the PRU of the beaglebone. It was not affected by the lockup. The bed was still hot enough to boil water a couple minutes after shutting it down, but was radiating enough heat due to the large surface area to not start anything on fire. I was a little preoccupied with what had just happened, and failed to get a real temperature measurement.

It has been several weeks since, and the above has never re-occurred... The lack of any thermal fuses capable of handling 60amps has led to a different fail-safe design. The current plan for a physical fail-safe is some power mosfets right on the bed, wired across the input. Gates will be tied high with a resistor, but held low by a thermal fuse. If the thermal fuse blows, input will be shorted, blowing the supply fuse.

Alternately, I could move the existing power mosfets to the bed itself and sink the inputs on failure.

There were simpler solutions, but they all involved external wiring and another possible source of failure.