Technical

Getting Grounded: A Thermocouple Rite of Passage

In today’s consumer product design world, thermal design is often the dark sheep of engineering.  The spider webbed phone screen – or even the waterlogged laptop on critical life support in a bag of rice – is a familiar sight in our tech-acclimated society.  But the blank screen with an error message, “iPhone needs to cool down before you can use it”?  Sacrebleu!

Testing thermal limits just doesn’t sound as glamorous as pushing processor speed or stretching screen size. And yet it’s perhaps the greatest limit to faster computers, quick-charging electric cars, and affordable LED lighting. I’ll save my defense for another post, but I urge you to become a better-informed thermal citizen by paying a little more attention to what’s hot (or not) in your surroundings. Nurture your inner engineer and grab a thermometer, or even a thermocouple.

Ah, the humble, misunderstood thermocouple: two dissimilar metals joined at a tip with which you probe the thermal world. It’s the magic behind digital food thermometers popularized by shows like America’s Test Kitchen and Alton Brown’s Good Eats. A quick read on Wikipedia will treat you to a 100-level piece of thermocouple trivia:

The “sensor” of a thermocouple isn’t the little solder bead at the end of a bare wire thermocouple, which allegedly produces a voltage difference at that point. Rather, thermocouples produce their signal based on the temperature gradient along the entire length from the tip to “the other end,” or reference junction.

 

What you won’t find on Wikipedia, at least as of this writing, is a 500-level trick question: when you need to measure multiple temperatures in a liquid, simply place a thermocouple at each point of interest. Not so fast, if your liquid (or metallic surface) has any significant conductivity! Much to our chagrin, ground loops form through the conductive device-under-test (DUT) that add oft-catastrophic noise to the millivolt-level thermocouple signals.

Testing thermal limits just doesn’t sound as glamorous as pushing processor speed or stretching screen size. And yet it’s perhaps the greatest limit to faster computers, quick-charging electric cars, and affordable LED lighting. I’ll save my defense for another post, but I urge you to become a better-informed thermal citizen by paying a little more attention to what’s hot (or not) in your surroundings. Nurture your inner engineer and grab a thermometer, or even a thermocouple.

Ah, the humble, misunderstood thermocouple: two dissimilar metals joined at a tip with which you probe the thermal world. It’s the magic behind digital food thermometers popularized by shows like America’s Test Kitchen and Alton Brown’s Good Eats. A quick read on Wikipedia will treat you to a 100-level piece of thermocouple trivia:

The “sensor” of a thermocouple isn’t the little solder bead at the end of a bare wire thermocouple, which allegedly produces a voltage difference at that point. Rather, thermocouples produce their signal based on the temperature gradient along the entire length from the tip to “the other end,” or reference junction.

What you won’t find on Wikipedia, at least as of this writing, is a 500-level trick question: when you need to measure multiple temperatures in a liquid, simply place a thermocouple at each point of interest. Not so fast, if your liquid (or metallic surface) has any significant conductivity! Much to our chagrin, ground loops form through the conductive device-under-test (DUT) that add oft-catastrophic noise to the millivolt-level thermocouple signals.

Skeptical? Here’s a comparison from a recent project that caught me by surprise:

2

Thermocouple ground-loop solutions fall into two categories:

  1. Isolate the probe from the conductive DUT.
  2. Isolate the thermocouple amplifiers from the world.

The first is the simplest approach used in industrial controls. Any seasoned industrial engineer would be quick to educate you on the value of buying ungrounded immersion probes. As shown to the right, their encapsulated construction isolates the thermocouple wires from the environment.

Unfortunately, response time suffers substantially. In my application, the response time in still water rose from roughly 10 to 60 seconds. Even inserting a thin layer of nonconductive adhesive tape between the DUT and bare thermocouple could do the trick. Applied to a metal surface in a recent project, I found a 90% response time of 15 seconds with this MacGyver approach, compared to 3 seconds without electrical isolation. The response curves are plotted below.

4

The second approach places a much larger burden on the signal conditioning electronics. If you can spare the cost and complexity, you’ll earn the privilege of continuing to use cheap, fast, bare thermocouples. A quick Internet search for isolated thermocouple amplifiers provides several reference designs from Analog Devices and TI, along with numerous ready-to-use commercial products. I did not have the chance to test isolated amplifiers with this recent project.

 

Approach Advantages Disadvantages
Ungrounded probe, by construction or by external electrical insulation Drop-in replacement for typical, non-isolated systems Slow response time, less flexibility with probe form factors
Isolated amplifier Fast response time with cheap, grounded probes Expensive signal conditioning, especially for multiple channels

 

Lastly, don’t forget that the most popular K-type thermocouples have substantial manufacturing variations (often quoted up to ±6°C!). Calibrating for consistency across sensors is key when you’re interested in degree-level resolution. Need to do much better than that? Consider meticulous, non-linear calibration, or alternative sensors such as platinum resistance temperature devices (RTDs) or thermistors. But that’s another blog post…