Temperature - subjectiveness and miscalibration of thermometers

[kaon]

Hi all,
Not sure where I'm going with this, just want to start a discussion on the related frustrations of temperature (and humidity) measurement.

It seems to me that temperature is a quantity that is more difficult to measure with sufficient certainty for every day purposes than say length, mass, force, voltage, time.

Gather a bunch of thermometers, and it is not unusual to find disagreements of 1 degree C or worse, which is significant for HVAC purposes. Thermometers that were once good, can drift over time.

Having a thermometer in each cooling water pipe, each room, each part of a chiller etc, isn't as helpful as it might be for troubleshooting, because there is too much uncertainty in whether they different sensors are calibrated the same. i.e. If deltas of 0.5 degree C are considered important, then we need thermometers that are known to agree within at worst half that.

Another scenario:
You have building occupants that complain about the temperature, "too-cold"/"too-warm", so you try to place more thermometers around the place, to try to get more objective feedback. But the thermometers don't even agree with one another...

What's the solution?
Better thermometers? (which ones are good?)
Using a single thermometer / probe to stick into each well?
Any other thoughts?

[gmcd]

Don't forget the mean radiant temperature from building surfaces, especially around the crappy perimeters on most buildings. In a cold climate, the inside surface of a double glazed window can be very cold, creating a radiant cooling panel. Conversely, in a sunny climate, that heat absorbing dark tinted solar control glass can warm up to have inside surface temperatures over 95F, creating a large radiant heating panel.

The problem in North America is that all of the commercial HVAC thermostats and sensors measure "air" temperature and they don't measure mean radiant temperature, or relative humidity (unless you use a humidistat mounted beside the thermostat).

ASHRAE-55 Standard will tell you what you need to know about the true comfort temperature - the "operative" or "resultant" temperature. If you live in a hot and humid climate, you often hear the weather report mention the "humidex" reading - that's the temperature you really feel when the air is warm and is very humid- the air dry bulb temperature might be 90F, but if it's 90% relative humidity it will feel like 97F or more - that's just another way of expressing the "operative" or "resultant" temperature. Take a nice sunny spring day - the outdoor air temperature might be 60F, but sitting in the sun will feel warm because of that radiant heater in the sky that makes you feel like it's more like 70F.

[xarralu]

We always calibrate thermostats with one know calibrated hot wire thermometer. Like you said different thermometers will eventually read different from one another. At least though there is some consistency in what you are reading, repeatability. Also, you will see guys calibrating thermostats with a laser thermometer. That is not the correct way. This is reading surface temperature.

[osiyo]

It seems to me that temperature is a quantity that is more difficult to measure with sufficient certainty for every day purposes than say length, mass, force, voltage, time.

Probably not ... yah just THINK you're more certain about length, mass, force, voltage, and time. <G>


Gather a bunch of thermometers, and it is not unusual to find disagreements of 1 degree C or worse, which is significant for HVAC purposes. Thermometers that were once good, can drift over time.

True enough, common and to be expected.

A quick check of several manufacturers' web sites for instruments commonly used in HVAC systems shows that they usually quote either a 1% or 2% accuracy over full range. For a 0 to 220 'F thermometer this would indicate that one could expect a reading to be off by up to 2.2 'F. Above OR below the real temperature. So one might find that when comparing the current readings from 2 similar thermometers the spread between the 2 might be as much as 4.4 degrees. And yet the thermometers would be within the manufacturer's allowable accuracy limits.

If these are analog thermometers, dial or liquid filled column types, it also makes a difference ... and therefore increases possible inaccuracy of reading ... in exactly how you're looking at the thermometer at the time. Head and eyes to the left or right? Held lower or higher than last time you looked at the scale? Etc. In calibration labs and in high accuracy required process control (ie manufacturing), personnel are trained to be careful of exactly how their heads and eyes are positioned when taking a reading on analog instruments. And in fact often employ some sort of mechanical guide or reference point to ensure they're looking at the indicator exactly the same way and from the same angle each time.

And, indeed, thermometers do drift over time. This is a well known phenomenon, MUCH discussed and studied for well over a century. And the reason quality control of thermometer manufacturing was implemented and calibration standards and labs were created.


Having a thermometer in each cooling water pipe, each room, each part of a chiller etc, isn't as helpful as it might be for troubleshooting, because there is too much uncertainty in whether they different sensors are calibrated the same. i.e. If deltas of 0.5 degree C are considered important, then we need thermometers that are known to agree within at worst half that.

Hmmm ... while I have seen literature indicating that deltas of 0.5 degrees are considered important, fact is that in most (almost all) installations I've seen, and I've seen a lot, the instrumentation installed was not capable of such resolution or accuracy. Except in somebody's dreams. Exceptions were in cases of high accuracy requirements in process control for manufacturing environments and in some chiller controls (digital). Although, while digital chiller controls often come from the factory adjusted to high accuracy, the fact is that over time RTD or other similar electronic sensors do also tend to drift significantly and one rarely sees anyone doing calibration checks and adjustments on them.

In any event, for standard HVAC equipment installations instruments most commonly used are not all that accurate in the first place, are commonly installed with somewhat less than ideal care and placement or adherence to instrument manufacturers' directions, selection of which thermometer by proper range ability to use is often ignored, and who knows if the suckers have EVER been checked for calibration or accuracy since they've been installed; be that one year ago or 20 years ago.

Typically operators/repairmen just tweak this adjustment or that one, higher or lower (as compared to equipment manufacturer's directions) until they achieve satisfactory or at least acceptable final results (output).

BTW, all this also applies to pressure measuring instruments. They're often horridly inaccurate. But operator/repairman tweaks system adjustments this way or that until he gets acceptable output. I can't remember the number of times I've watched some fellow "calibrating" his pneumatic control system and boasting about how accurate his results are, while he's using a $5 pressure gage that looks to be as old as he is. Fact is, he didn't "calibrate" the system, he just tweaked it until he got an outcome that made some office occupant stop complaining. Hey, it works ! I'm not faulting a guy for that. But results are often less than energy efficient.


Another scenario:
You have building occupants that complain about the temperature, "too-cold"/"too-warm", so you try to place more thermometers around the place, to try to get more objective feedback. But the thermometers don't even agree with one another...

What's the solution?
Better thermometers? (which ones are good?)
Using a single thermometer / probe to stick into each well?
Any other thoughts?

First off, you're always going to have complainers if the space is occupied by any significant number of people. There is no getting around that one. A "comfort zone" temperature is a subjective thing in large part. 70 'F at some specific, fixed value of relative humidity, is perfect for some; others will like 68 'F; still others will want 76 'F. Etc. Even when dealing with one person, what that person perceives as "comfortable" is going to change from time to time. Depending on how the person is dressed, level of physical activity, emotional mood that day, or general physical health (ie perhaps person has a cold, didn't sleep well previous night, etc). Also, the perceived air flow velocity upon the skin of an individual will make a difference in their perception of what is a "comfortable" temperature. A person who is otherwise comfortable at 72 degrees, might not be if moved over to where he/she is sitting under a diffuser.

But to address your question about thermometers.

I'd expect that in any room that I'd find that the temperature does indeed vary between this spot an that. If it didn't, I'd say you were using some really awful instruments as far as their accuracy goes. In fact, if yah stand in one place, in the average room, and measure temperature of air at knee level, then chest level, and then with instrument held over your head you should see temperature variations. Usually of at least a couple degrees.

So measuring at different places in a room SHOULD produce varying results even if yah use the same instrument. That's normal.

If you have a space with a LOT of temperature differential between this location and that, you might consider adjusting diffusers as to where they're located. Eliminating, if possible, any partition walls or high shelving units that might be blocking good air circulation to this area or that. Shading any windows. Etc. But you're always gonna have differences in measured temp around the room if it's any bigger than a walk-in closet.

A technique I've used for "problem areas". I'd survey the room with a reasonably accurate instrument. Look for "hot spots" and "cold spots". Record measurements. Try to eliminate causes of the worst of the hot and cold spots. Retake temps around room as before. Average them, then adjust room thermostat so that it's current sensed temperature matched that average. <Shrug> Sometimes that's the best yah can do.

If working with digital controls (which is what I now do for a living), I've used 4 sensors in a series-parallel arrangement to achieve an "averaging" of the room temp for problem spaces.

As concerns your thermometers and their questionable accuracy and what to do about it. If it really bothers yah, obtain an accurate instrument to use as a comparison instrument. ie An ASTM certified precision mercury thermometer (or equivalent). The difference between a thermometer of the type I mention and any other similar device is that the required precision in manufacturing the little capillary tubes inside is much stricter, mercury has a much more consistent and linear expansion and contraction rate than most other fluids, and the glass tube itself is a special glass that has less tendency to "warp and flow" than regular glass, plus has been specially "aged" so that most changes in its crystalline structure has already been done and settled out before it's calibrated, certified, and sold. In short, it hold it's accuracy a whole lot longer and better than is true with other instruments.

Use it to compare your other instruments to. Then, if the other instruments are not adjustable, affix a tag to each indicating it is off, plus or minus, by such and such an amount. Figure in that offset when taking your readings.

Another method would be to use the ice bath check. Use a small bucket or other container. Preferably bigger than a regular drinking glass as yah want to have some room (insulating factor) between middle of container and outside walls. Fill with ice chips, then add water to a level just below top level of ice. Stir well and let sit for 15 or 20 minutes. To allow temps to equalize. Insert your thermometer sensing base in the middle. Kind of stir it around a little, but not a lot. Keep in mind that bi-metal thermometers are usually manufactured so that the temp sensing element fills about the first 3 inches of the stem, measured from the tip. So make sure at least 3 inches is in your ice bath. Glass capillary tube type temperature sensors come in either a partial immersion or total immersion type. Partial immersion types should have a marking on glass that indicates how far it should be in the fluid to be measured. Total immersion type is immersed until water level is equal to top of the current fluid level inside the capillary, then pull out maybe half inch, just enough so yah can read the scale.

Yah should read 32'F (0'C) on your instrument ... after about 8 minutes. I say 8 minutes because if yah check manufacturers recommendations that's the average they give for a "settling time" to allow structural parts of the instrument (brass/copper tube surrounding bi-metal temp sensing element, or glass of thermometer, etc) to settle out and assume same temperature as whatever is being measured. If thermometer is off from that, adjust if applicable, or attach a tag to remind yah how far off it is to the plus or minus. Of course, instrument is still only gonna be accurate to whatever manufacturer's spec says. ie 1% over full range. But at least now yah can figure in any offset needed.

BTW, don't use ice blocks or ice cubes. In actual tests it has been shown that water with ice cubes, as versus ice chips, will usually assume a final temperature of around 34 or 35 degrees.

If yah want to do a two-point check (I wouldn't unless there is an actual need for high precision and accuracy), you can also use the boiling water immersion test. But adjust for altitude AND current barometer readings. Both can make a significant difference. OTOH, use of distilled water as versus ordinary tap water usually doesn't produce significant differences unless you're really picking-the-nit (or have some really hard water).

You could also go for a single, high precision and accuracy electronic test instrument. But any of them that's worth a darn are pretty pricey. Expect to pay at least a few hundred dollars for one. One we keep around the shop cost us over $1000. Add the cost of our sending it out for regular testing and recalibration by a properly rated testing facility. It does see use on actual customer sites, time to time. ie We do some work for pharmaceutical labs, etc were we must certify sensors we install to a certain "confidence level". So we use the expensive hummer to check sensors and add/subtract any offsets, etc. But more commonly we use the precision and NIST certified device to check our tool bag instruments. ie I can tell yah that my favored contact temp probe in my tool bag is off by 2.3 degrees (high). Okay for it to be off, as long as I know how far and in which direction. But that high precision, certified instrument is simply too expensive for every tech to have one and use every day. Such precision and accuracy us not really needed for most occasions anyway.


General notes:

1. Pay attention to my comment about "settling time" of 8 minutes. When you're moving around a live operating system that's constantly undergoing dynamic changes in response to changing conditions, your instruments are not gonna give "right now", instant results to that change.

2. I'm in the DDC automation side of things, and do a number of retro-fits where we change over from older controls to newer. I often observe thermometers (or other temp sensors) removed from wells that have never had any thermal conductive compound used on em when they were inserted into the wells. Use of thermal conductive compound can make a difference. USE it.

3. Dirt is your enemy. On an older system yah can have issues with accurate temp sensing simply because over time crud has built up over sensor element. When doing other maintenance, take time to check sensor and clean off the crud where possible.

4. Averaging sensors are often improperly/inadequately installed. I've seen endless systems, conventional controls or digital, where MA temp sensor was giving bogus numbers ... that were assumed to be reasonably correct. Air mixing is inadequate and air highly stratified, or just the placement of the MA temp sensor was poor. One way to check this is to wait for a particularly hot or cold day. A day when outside ambient air is significantly hotter or colder than return air (ie at least 10 degrees, 20 is better). Secure heating and cooling coils/elements (turn em off). Put economizer dampers in some position where you're gonna drive MA at least 10 degrees higher or lower than return. As measured by MA temp sensor. Then compare indicated temp sensor with DA temp sensor. By the time the air has hit the DA sensor, it is surely well mixed. Yah should expect about a 2 degree temp rise from MA to DA, heat added by fan. But if the two readings are way off, yah should consider moving around the long cap tube of the MA sensor so that at least some of it is getting hit by that hot or cold air stream that's obviously bypassing it. You'll never get it perfect. Well, almost never if it's any sizeable box at all. But yah can probably improve things and get a better operating unit.