RTD Thermowell Response Time


Thread Starter


Is it possible to ascertain the thermal response time of RTDs in thermowells installed in a particular application without conducting any testing? Are there any first-order or second order differential equations which can give the response time?

We've used ABB Temperature transmitters on SIL-2 and SIL-3 rated loops and due to the various factors influencing the actual response time such as thermowell dimensions, weight, material used, spacing between the RTD insert and thermowell, heat transfer coefficient and process parameters like velocity, heat capacity etc., all research I've done so far on the subject suggests that it's hard to determine the actual response time.

The problem is that ABB has provided some of the dimensions in their GA drawings and for the response time, they've conducted tests of immersing the sensor in flowing water at 0.4 m/sec in line with IEC 60751, however this only gives a rough guide of the thermal response time and the actual response time can vary significantly depending on the application.

ABB provides response times in the range of 60+ secs for the sensor-thermowell combination we've used but in practice the sensors could take longer to respond and hence can become critical where the process safety time is low.

I was wondering if mathematical analysis can be done to estimate the actual response time for a particular application, as this may obviate the need for in-situ testing of temperature sensors for response times?

Any inputs will be appreciated.

James Fountas

I did some playing around with this about two decades back. Forgive me, my memory is not good enough to give you the actual numbers. The test was to see the response time of various jackets from simple thermocouples in 1/16" SS probes to 1/4" with and without thermowells and the thermowells were with and without conductive material (paste, oil, etc. in the well).

I believe we defined the response time as the time it took for the sensor to detect 90% of the change in temperature. I remember the 1/4" sheathed thermocouple took close to a minute to respond without a thermowell. Anything in a thermowell took easily a minute or longer to respond. We only used SS.

We came to five general conclusions. One was that if the media you were trying to measure had good heat transfer properties you could just do a calculation for heat transfer through a plate to estimate what the response time was through a thermowell. Two was that larger thermcouples seemed to have disproportional longer response times. We believed this to due with the packing material inside of the sheaths. The internal material would be ceramic and void space which would have a lower heat transfer then the SS sheath. The thermocouple wire tips were not grounded to the thermocouple sheaths and thus the distance of the bead to the sheath probably varied and using smaller diameter sheaths was best option. Three was that if you did not use a conductive past in the wells, all bets were off and just plan on horrible and unpredictable response times. Four was that you can use a rigid sensor if you knew what you were doing and why, but just order spring loaded sensors so you did not have to double check anyone else work for critical temperatures. Five was that you should use brass, copper or even no thermwell at all were you could get away with it if response time or accuracy was an issue.

The thermwells themselves could be predicted somewhat well by a heat transfer through a flat plate. This used data only from small diameter sheathed thermcouples. The larger diameters ones (1/4") were more unpredictable.

James Fountas
james [at] fountas.net


Dave Bartran

Certainly possible, and done routinely. The only issue are the details.

The simulation that I've run agree with flow tested measurements within fractions of a degree, given that you know the details of the installation and the fluid.

Therein in the difficulty: most thermowell installations are relegated to that of a pipe fitting and the details are hard to come by.

Based on experience 60+ secs response times (as a first order time constant) is about right for most applications. Some safety critical processes actually perform in place testing of the sensors every few months. If you are seeing time lags of several minutes then you need to engineer the installation.

present some of your details and you'll enable the responders to better answer you inquiry.

Malvern Jones

Interesting tests. I did some similar tests back in the mid 80's and compared the results to manufacturer's data sheets. My results concurred with yours.

An interesting variable that came to light was that manufacturers can't be relied upon to tell you at what percentage of a step change at the input they measure the response. Some like you consider 90% of the input, others 99.9%.

In these days of hyped marketing, most manufacturers are measuring response times at 63% of the step change at the input.

A variable to watch!
topic FYI:

The home page for Correlsense response time management tool (above) appears to measure response times for ethernet communications over networks, not temperature measurement thermal lag through thermowells.

Dave Bartran

The quick answer is that the thermal mass of the tip goes as sqr (Tip_new/Tip_old), so the time constant increases by a factor of 2.6 or so.

Watch out this is a first order lag estimate, while the thermowell response is actually governed by conduction, diffusion, and radiation if you ignore the dynamics and the errors associated with the sensor. Equilibrium is achieved in 4-5 time constants more or less.

Frank Johnson

This is a great question. First of all, you may request response time information in any format you choose. However, the standard defined by ASTM and others all over the world is a step change from room temperature to boiling water moving 3 meters per second. The response time is the time it takes to reach 63.2% of the total temperature change. This number as in 30 some years of doing this stuff I have never encountered an industrial that was boiling water moving at 3 meters per second with instant step changes occurring. Therefore, all the info on response times must be treated as relative information.
The best way to derive a real number it to go empirical. You could stick a very small (say, 1/16 inch diameter) sheathed thermocouple with no well in the same media and time the results.

The medium that the well is installed in is significant as is the material of the well, the sensor, the construction of the sensor, the insulation etc..

You mentioned you did not want to do an in-situ test, but that is a very easy test to run and is fun.

Replace the present sensor with a dual element sensor. A thermocouple or RTD will work. An RTD is better. Run electrical current through one of the sensors while measuring the temperature of the other. When the temperature gets as high as you want it to get, remove the current source and time how long it takes to get back to normal. You can use 63.2%, 90% whatever you want to.

The heat flux going one way through the sensor and thermowell is exactly the same as it is going the other way.
I am curious to know if anyone was indeed able to obtain a mathematical model for thermocouple time response determination.
Thermowells have first order time constants. The Omega Temperature catalog lists values in the technical section
the published literature has several theoretical models, but are heavily dependent on the details of the sensor design, how it is installed, the material parameters of the sensor and thermowell (if used), the thermal environment (radiation, conduction, & convection) of the measurement, the amount of insulation used, the physical installation, the location of the connection head, ambient conditions, amount of isolation, cold fronts, rain showers, you name it.

Takes about a week per measurement to gather up all the details, but then only a fraction of a second to run the calculation (FEA). Yes, models and calculations exist and fun to use.

Sharad Devre


I'm preparing a thermal response test setup. In which different temperature sensors i.e RTD, Thermocouple will be exposed to different temperature profile. And then i'm supposed to observe response time of the particular sensor used. Now the question is can you suggest me what kind of different heat sources I can use? for example bucket containing boiled water. How different variations of sensors (direct immersion, thermowell immersion) can be used? Have you came across any such kind of setup? Any info/article/weblink for more awareness on this topic is appreciated.

Special calibration equipment is typically used for did testing.

For 0-100Deg C, temperature controlled water vat, with lid can be used for dip testing.

for higher temperatures, various non-flamable liquids can be used and in some for more extreme cases ceramic grains are used.

Call your local supplier, some of them have maintenance testing for on-site testing calibration in some of the heavy industrial areas. They would be happy to demonstrate it to you.

If you need flow testing, then you need to contact one of the flow testing labs. It won't be cheap.