J-type Thermocouple

[jnnewton]

I'm going to be using a J-Type thermocouple placed in temperatures in the range of -4 to 200 C and will need to monitor temperatures from 0 to 200 C. I only included the -4 in case it might matter (i think my circuit will just output voltage corresponding to 0 C). I plan to use the AD8494 from analog devices. The output of this (10mV / C) will be fed to a microchip microcontroller.

My first question relates to the connection of the TC to the PCB. I would like to be able to remove and re-connect it, so I don't want to solder it down if it can be avoided. We have a standard 2 lead one, with interchangeable tips (spring loaded caps). I expect it to wear out and / or get broken during the lifetime of the project. I would like to add a pair of phoenix contact terminal blocks to the PCB and connect the TC to those. How severely will this affect measurments? Are the TC-tin and tin-solder/trace interfaces an issue that I need to be concerned with?

My second question is related to temperature range and compensation and isn't really a question at all. The ambient temperature will be 4-40 C. The temperature range of the AD8494 shows 0-50C, but i believe this is only to stay within the 1% accuracy range, and it appears the accuracy drifts linearly, so a bit more temperature (may be hotter in enclosure than ambient) probably won't push me much over 1 C in error. That conclusion is based on pretty thin info, and assumptions and I would simply like for someone to verify my thinking, or tell me I'm wrong, and why.

 

[JonSea]

There are special thermocouple connectors for PCB mounting but these have been difficult to source in small quantities. You can use a terminal block or other connector as long as you are aware that this connector then becomes the cold junction point. If there isn't a big thermal gradient, this won't impact your accuracy much. Use large areas of copper for each terminal on the PCB and keep your cold junction measurement close.

You might consider other options to the T/C chip. If you use a chip with a digital output (I2C or SPI) you avoid having to read a small signal with the PIC's ADC. Analog and Maxim both have options to do this, although I don't know what's available for a type J.

I take a slightly different approach to read T/Cs. I use a TI TMP512 or TMP513 to make the measurements. These chips are power monitors with a built-in temperate measurement (or remote using standard transistors) and a high resolution adjustable gain ADC for measuring voltage across a current shunt that's ideal for reading small thermocouple signals. I use a lookup table in the PIC for linearization so you can customize for any type of T/C, temperature range and needed accuracy. More details are shown here:

A Thermocouple Measurement Circuit with Swordfish Software

 

[KeepItSimpleStupid]

A T-type thermocouple would be a better choice for that temperature range if compatable with your process.

Very few people ever did cold junction compensation right. A USB device from Measuremet Computing really sucked at it. Most manufacturers put a thermister nearthe terminals and it expect it to work. A little breeze and it's useless.

One Fluke Thermometer did it right. They used a large unk of metal and effectively mouted a nice TO-5 (I think) sensor with a thermal insulator between the large pieces of metal.

Use the standard panel mount connectors for thermocouples (mineature or standard). Inside your instrument, use extension wire to the actual 1/2 junctions.