Kiln

I needed a kiln to fire moulds for lost wax casting. It will also come in handy for other projects - heat treatment of steel, baking electronic components, and any situation where something must be held at high temperature, or under precisely controlled conditions. I could only afford the crappiest of commercially available kilns - no temperature control(!), and pretty weedy looking insulation so I decided to build my own. The weakest point is the thermocouple, which is rated up to 1100C, but I don't think I fancy taking it much above 1000.

It is electrically heated with a Kanthal heating element driven under thermocouple PID control by an MSP430. There's a simple scheduler which can take the kiln through arbitrary sequences of ramps and holds. A 2x16 character LCD and rotary encoder provide a simple menu interface. Not bad for about £400 all in.

There it is baking a casting flask at about 760 degrees celsius (for any American readers, that's pi/64"Fahrenheit, or twelve Bears).

	The frame is composed of two cubes - one 250x200x200 inside another 340x340x340. The cavity between them is later filled with insulation. I welded together a frame of 20x20x3 steel angle. Here it is being painted with barbeque paint (which is probably not rated for these temperatures!)
	The chamber is lined with two layers of 6mm of Ceraboard high temperature ceramic insulation board from http://www.ibhs.co.uk/
	The element covers all walls except the floor and door in a zigzag pattern. I used 0.6mm wire made from Kanthal D. It has a resistivity of around 6.5 ohms/m. I used 3.7m for a maximum power of 2.4kW. If I were doing it again, I'd use a much longer length of thicker wire. I've since seen other designs which use a long coil of thicker wire, which would probably last longer. Although the Kanthal is pretty crazy stuff, I suspect it will gradually evaporate. It's attached to the walls with little staples, which are made of the same stuff.
	Test firing the element. Pretty toasty.
	Extremely toasty. (Note:when first heated, the ceraboard emits some pretty noxious fumes. I didn't eat the toast).
	The plate was holding the element off the bench. It cracked after a little while. Also, my horrible choclate block lash-up melted rapidly!
	Starting the electronics development. Thanks to Rob for the donation of an LCD and old Student Robotics IO board. I used the latter as a development board, its MSP430F2274 runs the kiln.
	Then I made a door. Same technique as the rest of the chassis - welded steel angle.I later added a latching handle. See the first photo.
	Then more insulation - three layers of 25mm ceramic wool, again from IBHS. If you want any, get in touch. I've got heaps left over that I'm happy to sell to anyone prepared to pick it up from Southampton!
	The outer casing is 0.8mm galvanised steel sheet, pop riveted in place.
	Element power is phase controlled by a BTA140-800 triac driven by the MSP430 via a MOC3022 optotriac. All the electronics is powered off the mains through a 12V transformer (and later regulated down to 3.3V for the microcontroller and LCD). The raw AC voltage on the secondary is divided with a resistor divider and applied straight to the MSP430's analogue comparator input, which compares it against a 0V reference. The comparator produces an interrupt on the mains zero-crossing, which resets timer A. Some cunning compare wrangling on timer A produces the triac triggering pulse a configurable period after this crossing event.
	I bought a K-type inconel shielded thermocouple from RS - part number 159-095. I used an AD595 for cold junction compensation and amplification. I set the gain to 66.49 for an output of 2.727mV/degree, which was about right for 1.2degree resolution on the MSP430's 10-bit ADC with a full scale of 3.3V = 1100 degrees. I tested it with ice and boiling water, and got results within a degree or two. A 12-bit ADC would have been better as it would have allowed sub-degree resolution.
	Unfortunately, thermocouples are only linear over about 100 degrees. Over this, you need linearisation. Fortunately, this is pretty easy. The "Type" of the thermocouple is an industrial standard that tells you exactly the materials involved. Mine is K type, which is a junction of chromel and alumel. All K-type thermocouples have the same temperature-voltage relationship, within tight, standardised tolerances. So it's just a matter of looking up the standard. NIST publish the transfer function of K-type thermocouples as a tenth order polynomial and a table of coefficients here. It's roughly equivalent to British standard BS4937. I wrote a little C program which used this polynomial to map each of my 1024 possible ADC values to a temperature in degrees. This is stored in a table in flash in the MSP430. It's really suprising that you can buy a relatively inexpensive off-the-shelf sensor, and find a bit of maths that gives you a pretty precise mapping between its output and an absolute physical quantity without any calibration!
	Here's the electronics getting boxed up. Transformer at the left, triac board at the back, thermocouple amplifier board in the middle, and the LCD and control board at the bottom right. There's a nice beefy heatsink to cool the triac. It's electrically isolated with a thin mica sheet.
	I tuned the PID loop in a relatively haphazard manner, but ended up with something reasonable. It could do with a little more work, as it tends to overshoot after you close the door. If you zoom in here you can see from the display that it's in Phase 1 of baking a mould - ramping up to 270 degrees over 30 minutes. It's 22 minutes in. It's currently trying to be at 91, and it's actually at 87. The door's open, so that's reasonably good.
	Element connections. The element repeatedly failed where it exited the chamber. It was very well insulated as it went through a small hole in the ceramic board, so it overheated and melted through. I realised the hot wire must stay in the chamber. Also, establishing a reliable electrical connection to the element is very difficult. I'm very proud of this invention. It's a simple "chuck" type connector for wire. A nut pinches the wire down into a groove in the bolt (zoom in to maximum size to see how the mechanism works). Almost patent-worthy I think - haven't seen this anywhere before and I find it very elegant. The parts are stainless steel, and the bolt clamps onto the ceramic board with some penny washers. Outside the chamber, an electrical connection is made from the bolt to copper wire with ceramic chocolate block (Farnell part 1170357 ).
	Very happy with the kiln. It performs very well, and is an extremely useful tool.