mr.crow - 25-4-2011 at 15:38
Here is something you may be interested in:
Take a glass tube and heat it with a torch. Pull them apart and seal off one of the ends. When its cool drop a small piece of hot glue in and heat it
gently, when it melts push the end of a thermocouple in. The glue makes for better thermal contact with the glass.
As a bonus: wrap your plastic bottle caps with metal tape. Wrap the seam between the cap and the bottle with electrical tape or more metal tape.
Nothing is going to leak in or out of that! Never had any luck with teflon tape.
smaerd - 25-4-2011 at 16:09
Any chance you'd be able to take a picture 
?
m1tanker78 - 25-4-2011 at 16:39
One of my upcoming projects is going to be a high temp thermistor-based digital thermometer with a microcontroller. Actually it will be 2
thermometers, one ambient and the other will calculate the junction temperature based on ambient temp and junction resistance. I believe this is the
only way I can reliably measure the temp of molten metals and molten salts.
I'm not sure how you went from glass tubes to plastic bottle caps?? Can you post a diagram or pic like smeard said?
Tom
mr.crow - 25-4-2011 at 16:52
Oh yeah the bottle caps are a totally separate idea, but it was on my mind today. Good for all the solvent bottles or things that are very
hygroscopic.
Here is a picture. (may go offline after a while with no views)
Thermocouples can go over 1000 degrees if they have the right insulation. Mine will just melt
smaerd - 25-4-2011 at 21:27
Interesting! I have a thermometer with a broken bottom to it, that's just been laying around, looks like it might have some use now!
smuv - 26-4-2011 at 08:42
I have done the same thing, just using molten solder where you are using hot glue. In my experience it works pretty well.
Also m1 tanker you don't need to do cold junction compensation for a thermistor. Do you mean thermocouple? there are many IC's that do auto cold
junction compensation which makes thermocouple integration very easy.
[Edited on 4-26-2011 by smuv]
m1tanker78 - 26-4-2011 at 09:08
smuv: Sorry, I meant thermocouple. Although a thermistor design isn't off the table (yet). I might be able to compensate for a thermistor's
non-linearity with a good lookup table. There's also the probe design to keep in consideration. I've also used a solder bead in an aluminum tube
before but my current t'couple can't withstand the high temps of molten metals and salts. I've even fathomed using sodium metal for heat transfer.
It's been a while since I've done this type of project. Time to dust off my gear and have a go.
Tom
smuv - 26-4-2011 at 12:02
I recommend the MAX6675 chip. It does everything, all you need to do is read it via spi. Should only take you a few hours to go from perf board to
working circuit. All the info you need is in the data sheet.
It has a 0.25C resoltuion with realistically a 0.75C accuracy (once calibrated). Can read 0 to 1024C.
IrC - 26-4-2011 at 12:46
I have a hacker thought on this but no pic so will try to describe. Take a good IR target disc, say ~1/4" diameter. Adhere at bottom of say an 8"
glass tube with flat bottom. Tube first prepared by mirroring the inside, say by rolling Gallium around in it. Take the IR sensor from a cheap baby
thermometer and mount it near top of tube carefully aimed at disk. Wire to display/electronics portion mounted at top in some convenient way to make a
nice IR thermometer. Possibly you could find a cheap IR thermometer with a better range to make the project more useful. Just a thought but it sounds
like a great hacker weekend project? Just push a button and get a reading, with peak temperature memory.
Oh well just thinking out loud.