Ungrounded thermocouple construction advice

[MDM3D]

So I am running into problems with my current thermocouple setup because the thermocouples are welded to a metal block and the heaters in the block are not isolated from the block electrically so my thermocouple amplifier gets upset when it tries to take a measurement while power is applied. I have been looking into ungrounded thermocouples that are isolated electrically. Does anyone have any insight into what material is used as the insulating material?  I need the insulating material to be stable up to 550 C but also a good thermal conductor. My first thought was firebrick cement but that appears to have a  high thermal resistance and I don't want a huge delay added to my measurement. My next thought was beryllium oxide but this seems a bit rare.  Thanks in advance for your help.

[C]

The short answer is to change what the thermocouple amplifier is sensing.

You add a signal to each side of thermocouple amplifier input based on current power applied to cancel effect of power applied.

You will have inputs of = +power + -power + thermocouple.
The + power & - power will cancel leaving thermocouple value.

Need an example to make it clearer.
A battery has an internal resistance. You can not directly measure that resistance with a meter. If you measure the current power a battery is producing( the voltage & Current) and quickly change the current, the battery's internal voltage will be slow to respond.
The two voltage & current readings lets you compute the internal resistance.
Your thermocouple acts like a low voltage battery.
By changing power to heater quickly you can expose values.

Think of your setup
  You have a heater resistor in series with thermocouple in some fashion. You have a voltage divider.
You can create a very low power matching copy of heater by scaling the resistors.

To measure power to heater block you need to add a current sense resistor.
Heater = current sense resistor, heater resistor, thermocouple.
Matching copy = 100x current sense resistor, 100x heater resistor, thermocouple.

Note that with the copy having 100 x the resistance, the heating change will be less.

When both chains are driven from same power source, you can compute the temp change of resistors if you change power level.

You have one thermocouple that changes based on heater block.
You have one thermocouple that reads room temp.

The copy  thermocouple could be replaced with a resistor.
 
Note that if you connect your thermocouple amplifier to have one input being the heater block chain and other input the copy chain you can get the power cancel.

You have added two current sense resistors, one or two power sense inputs and a pot.

To check this out you might want to replace the second current sense resistor with a pot.
You want the voltage across both current sense resistors to match.
As power changes you adjust the original pot for 0 Change.
If power source to heater is AC this is easy.

[MDM3D]

Thank you for your detailed response.

The heaters are powered by DC and with varying power levels via PMW. The heater and block have been design with very low thermal mass to improve response rate. I do understand what you are saying but my heaters are powered by 24V and the digital thermcouple amplifiers are powered by 3.3V to interface with a micro controller.  The amplifiers are MAX31855.  Ungrounded thermocouple will work and i have order some but I have to wait a couple of weeks till they arrive and I was looking to make my own until the proper ones come in. The rest of the system that they are going in doesn't have enough excess analog pins to account for the 4 thermocouples that I need to interface with.

[jbb]

I need the insulating material to be stable up to 550 C but also a good thermal conductor. ... My next thought was beryllium oxide but this seems a bit rare.

Don't use beryllium oxide; it's toxic! That's why it's rare these days. A solid block is sort of OK, but any rubbing, grinding, chipping, filing or (gasp) sanding will release toxic dust that you might inhale.

Aluminium oxide might be OK, but I don't know if it's toxic in powder form.  If you want to try it you should do some research.

Some other insulating opportunities:

• an RTV (Room Temperature Vulcanization) silicone, but it could be hard to find something that goes that hot
• a drop of mineral oil but 550 deg C would boil that (and possibly set it on fire)
• a thin walled glass tube, which will have considerable thermal resistance

As an intermediate step, I suggest you isolate the MAX31855.  An ADuM5411 might be a good choice for low volumes (it's highly integrated but a bit expensive).

[mzzj]

Metal sheat mineral insulated thermocouples are bog standard and cheap, no need to invent your own.
I have been using tcdirect as they have very competitive prices:
https://www.tcdirect.co.uk/Default.aspx?level=2&department_id=190/1

Insulation inside the metal sheath is magnesium oxide powder. Sheath material is usually the limiting factor, Inconel 600 or Nicrobell-D is good up to 1000-1300c depending on desired lifetime.

You can also get ceramic tubes to protect and insulate your thermocouple(thermowells) but the thermal coupling can be less than ideal at lower temperatures.

[MDM3D]

Metal sheat mineral insulated thermocouples are bog standard and cheap, no need to invent your own.
I have been using tcdirect as they have very competitive prices:
https://www.tcdirect.co.uk/Default.aspx?level=2&department_id=190/1

Insulation inside the metal sheath is magnesium oxide powder. Sheath material is usually the limiting factor, Inconel 600 or Nicrobell-D is good up to 1000-1300c depending on desired lifetime.

You can also get ceramic tubes to protect and insulate your thermocouple(thermowells) but the thermal coupling can be less than ideal at lower temperatures.

Thanks! Mg0 appears to be very widespread in thermocouple use this sounds promising. For the application I am looking to use this for keeping the thermal mass low so that I can change temperatures very quickly. My target is roughly 50 C a second. Heating is easy, add power but forced cooling via air or liquid cooling would be a massive pain. I am relying strongly on radiation and convection which means I can't have much stored energy or measurement delay if I want a stable temperature of +-1.5 C.

[Kleinstein]

50 K / s is pretty fast.

Shielded metal clad thermocouples are available as standard parts even with quite small diameter.  AFAIR I use 0.5 mm diameter ones up to 600 C with no problems (though only to about 0.3 K / second).  Response can be reasonable fast, but about 1 second delay may be there. The grounded version is usually a bit faster in response.

[MDM3D]

50 K / s is pretty fast.

Shielded metal clad thermocouples are available as standard parts even with quite small diameter.  AFAIR I use 0.5 mm diameter ones up to 600 C with no problems (though only to about 0.3 K / second).  Response can be reasonable fast, but about 1 second delay may be there. The grounded version is usually a bit faster in response.

You are absolutely correct a lot of work has been done to make the block have a small thermal mass and a high heat transfer rate. If the ungrounded thermocouple has to much thermal resistance then isolated heaters will be my next step followed by a feed-foward temperature controller. 

[ikrase]

There's always aluminum nitride. It doesn't kill you.

At my job we have aluminum nitride heating blocks on some fixtures. Kinda creepy how cold it feels when you pick it up compared to more normal silica/porcelain/alumina ceramics.

[MDM3D]

There's always aluminum nitride. It doesn't kill you.

At my job we have aluminum nitride heating blocks on some fixtures. Kinda creepy how cold it feels when you pick it up compared to more normal silica/porcelain/alumina ceramics.

Definitely some neat material, Thank you for the recommendation!

[GerryBags]

Graphite? It's available as a very fine powder which you can mix up into a slurry, pour into a mold and bake dry or vac-chamber it to form whatever shape you need. It's stable at far higher temps and is an excellent thermal conductor.

[Kleinstein]

Using a differential input (and thus not a special thermocouple input chip) could be an option. So it would be more like instrumentation amplifier and ADC with a separate temperature sensor (could be just a diode or a digital one) for the cold junction compensation. Even direct interface to an differential ADC (e.g. MCP355x) might be an option. Doing the math for the cold junction is usually not a big deal for the µC and regulation could be done based on the voltage as well. Especially for a system with more channels a separate cold junction might be a good option.

Feed forward can be quite useful when you are at the limits of the thermal setup. It helps especially with things like changing ramps. I did that on a slower scale and it really helped to get something like a nice triangular temperature profile. For a software implemented PID regulator feed forward is not that complicated.

For thermal response it is not only the thermal conductivity that counts, but also the size. So a thin layer of mica might still be a good option, as it can be very thin and low mass.