Need accurate temperature display

This is the place for any magazine-related discussions that don't fit in any of the column discussion boards below.
jimandy
Posts: 570
Joined: Sat Dec 04, 2004 1:01 am
Location: Birmingham AL USA
Contact:

Need accurate temperature display

Post by jimandy » Fri Jul 16, 2010 11:26 am

I am having “issues” with the digital, programmable thermostat in my house. I won’t go into details other than to say when I set it to cool at 79 the AC is still cooling and the t'stat's own temperature readout showing 77. To investigate I bought a simple digital display thermometer from Walmart, placed it next to the thermostat and it says the temp is 78, I then gathered digital thermometers of various makes from other rooms in the house. placed them next to the built in thermostat and the readings are 75 on one, 76 on another, 77 on the installed one (the thermostat) and of course, the new one at 78. All are within inches of each other on a wooden cabinet with no thermally active devices nearby.

Would appreciate advice on what brand/make or model of thermometer I can buy to give me a truly accurate reading, preferably to tenths of a degree. Optionally would like a glass bulb type, but the ones sold through most stores have such a wide range that to know with some accuracy what the temperature is between 70 and 80, you have to guess since the divisions are usually in 2 degree increments.
"if it's not another it's one thing."

User avatar
reloadron
Posts: 519
Joined: Sat Jun 28, 2008 8:57 am
Location: Cleveland, Ohio
Contact:

Re: Need accurate temperature display

Post by reloadron » Fri Jul 16, 2010 4:26 pm

I am going to venture a guess why it is doing what it is doing beyond the overall accuracy. Most of these devices have a hysteresis built in of a few degrees. If I set my thermostat on cool and a set point of 75 the compressor will run and the system will pull the indicated temperature down to about 73 before it shuts down. Then as the room temperature slowly rises it will pass through 75 and when it hits 76 the cycle will repeat.

Without any hysteresis and just on/off and a set point of 75 as soon as the room temperature exceeded 75 the compressor would turn on and pull the room down to 75 and shut off. This would make for a more frequent on/off action on the compressor. It would be a constant on and off based on a one degree change in the main room the thermostat was in. Even if I left the blower fan on constantly to move air the compressor would just constantly cycle on and off with a single degree of change in room temperature.

As to the overall accuracy of the temperature some can be adjusted in comparison to an accurate thermometer but for systems like this used in a residence an uncertainty of +/- a few degrees F. is considered pretty good. Even if the display resolution were .1 degree as in 75.1 a regular home system would never keep up with trying to maintain a room temperature within .1 degree.

Just My Take
Ron

jimmy101
Posts: 82
Joined: Tue Mar 24, 2009 11:08 am
Contact:

Re: Need accurate temperature display

Post by jimmy101 » Sat Jul 17, 2010 10:02 am

Most inexpensive digital thermometers are accurate to about 1C (~2F), so the results you are getting for several different thermometers is about right. Generally, that is sufficiently accurate, even for laboratory work. (An expensive mercury filled lab thermometer is typcially accurate to about 0.5C, ~1F.)

Usually even a cheap digital thermometer is much more precise than it is accurate. That is, it'll give the same reading for a given temperature +/- perhaps the least significant digit in the display.

In a HVAC system high absolute accuracy usually isn't much of a design criteria. Reproducibility is much more important since a person cant really tell that the temp is 77 or 78 but might be able to tell when the temp has drifted by a degree or two.

And, like reloadron said, the thermostat is designed with significant hysteresis. You don't want it cycling on when the temp rises 0.1F above the set point and you don't want it to cycle off when the temp is 0.1F below the set point, the compressor will be constantly cycling on and off and that's hard on a compressor. So, typical thermostats are designed with about 1~2F or so of hysteresis.

It sounds to me like you thermostat is working correctly but perhaps has a slightly bigger hysteresis than some other models.

Temperature measurement is one of the trickiest types of measurement in science. A typical lab-grade Hg thermometer traditionally has 1 or 2C gradations (2 to 4F), and can be read to a precision of perhaps 0.2C (~0.4F). The typical accuracy is 1~2C or so and if that isn't good enough then the thermometer would be calibrated in the lab, usually using ice/water and boiling water as reference temperatures. After careful calibration the precision is exactly the same as it was before but the calibrated thermometer is now of known accuracy.

A $5 digital thermometer that is accurate to 2F at 77F is an amazingly accurate device for $5. To compare two temperatures you can't use a relative scale like Celsius or Fahrenheit, you have to use an absolute scale like Kelvin of Rankin. 77F is 298K, if the thermometers is off by 2F then it indicates 299K. That is an absolute error of 1 part in 300, or 0.3%. That is about as accurate as a high quality machinist one foot ruler marked to the nearest 1/32". In other words, a $5 thermometer may well be the most accurate measuring device you have in your home.

jimandy
Posts: 570
Joined: Sat Dec 04, 2004 1:01 am
Location: Birmingham AL USA
Contact:

Re: Need accurate temperature display

Post by jimandy » Sat Jul 17, 2010 1:29 pm

I suppose I could amass as many different thermometers as possible, set them all side by side, and if two or three agree, accept that as a reading as close to accurate as possible. Then use one of those as my standard.

Ah but I hear someone saying I will have to vary the ambient temperature and take further readings as each may have its own non-linearity characteristics..
"if it's not another it's one thing."

User avatar
reloadron
Posts: 519
Joined: Sat Jun 28, 2008 8:57 am
Location: Cleveland, Ohio
Contact:

Re: Need accurate temperature display

Post by reloadron » Sat Jul 17, 2010 2:36 pm

jimandy wrote:I suppose I could amass as many different thermometers as possible, set them all side by side, and if two or three agree, accept that as a reading as close to accurate as possible. Then use one of those as my standard.

Ah but I hear someone saying I will have to vary the ambient temperature and take further readings as each may have its own non-linearity characteristics..
Actually what you need is only one known accurate thermometer.

As Jimmy101 covers quite well:
Temperature measurement is one of the trickiest types of measurement in science. A typical lab-grade Hg thermometer traditionally has 1 or 2C gradations (2 to 4F), and can be read to a precision of perhaps 0.2C (~0.4F). The typical accuracy is 1~2C or so and if that isn't good enough then the thermometer would be calibrated in the lab, usually using ice/water and boiling water as reference temperatures. After careful calibration the precision is exactly the same as it was before but the calibrated thermometer is now of known accuracy.
The following images are my sets of lab-grade mercury glass bulb thermometers:

Image

Image

These all happen to be graduated in degrees F. Each thermometer has an accompanying sheet with the correction factors. Also when used in a bath stem corrections need to be figured in. Yeah, this temperature stuff can get pretty deep when we start splitting degrees. :)

Now for everyday home use all you need is a good digital temperature indicator to read a thermocouple. A unit that can be calibrated is preferred. Take a trip to the quickie mart and buy a few gallons of the nicest distilled water you can get. Make the purest ice cubes you can make. Reserve some of the water. Like 50% of it. :)

Break up your ice and place it into a dewars flask, in the absence of that use a thermos, you actually want a slurry in an insulated cup. Place your thermocouple in the slurry and it should read 32 F or 0 C. Next find out your exact elevation above sea level and bring your water to a rolling boil. Place your thermocouple in the water and make sure it reads whatever would be correct for your elevation. Water boils at 212 F or 100 C at sea level under the right atmospheric pressure so you need to do the correction factors. If your thermometer reads 212 F. or 0 C. you have a pretty good standard. :)

This is coming from a guy who did a temperature uniformity survey on a convection oven using and charting 9 thermocouples when his wife argued the oven wasn't accurate. The hell it wasn't!

Ron

jimmy101
Posts: 82
Joined: Tue Mar 24, 2009 11:08 am
Contact:

Re: Need accurate temperature display

Post by jimmy101 » Sun Jul 18, 2010 10:51 am

I don't even want to think about what Reloadron's set of thermometers is worth. Heck, the last research chemistry lab I worked in we removed all the mercury containing thermometers and replaced them with digital ones. (Spilled mercury is just to dangerous and to much of a PITA to clean up.)

I really don't see any reason to try to calibrate the thermometer(s). It really doesn't matter does it? You can't tell the difference between 76 and 77F.

User avatar
reloadron
Posts: 519
Joined: Sat Jun 28, 2008 8:57 am
Location: Cleveland, Ohio
Contact:

Re: Need accurate temperature display

Post by reloadron » Sun Jul 18, 2010 11:02 am

jimmy101 wrote:I don't even want to think about what Reloadron's set of thermometers is worth. Heck, the last research chemistry lab I worked in we removed all the mercury containing thermometers and replaced them with digital ones. (Spilled mercury is just to dangerous and to much of a PITA to clean up.)

I really don't see any reason to try to calibrate the thermometer(s). It really doesn't matter does it? You can't tell the difference between 76 and 77F.
Yeah, I can't actually feel the difference between 76 & 77 degrees in a room.

Years ago when our lab went from the glass you see to real nice Platinum Resistance they were going to dispose of those sets so I asked for them. I even have a few assorted range spares. They gave them to me just to get rid of them. At the time from Brooklyn Thermometer or HB those sets were about $3,000 each. Cole Parmer still sells them and a single thermometer is between $250 to $800 depending on the range.

I seldom use them as I have some accurate thermocouples that are calibrated.

Ron

jimandy
Posts: 570
Joined: Sat Dec 04, 2004 1:01 am
Location: Birmingham AL USA
Contact:

Re: Need accurate temperature display

Post by jimandy » Sun Jul 18, 2010 6:50 pm

I can certainly tell the difference between 79 and 77 and the fact is when the set point is at 79 and it shows that it is 77 and the compressor is still running I just up the set point to 81 to get what I can accept as a comfortable temperature. But from the reading of the other thermometers I can assume that the “real” temperature is in the 75 to 76 range so, dang it, I want the unit that controls the compressor to read accordingly.

The t’stat, BTW is a carrier Infinity system unit with way more bells and whistle than I need (came with this new house). Googling on “problems” and “Carrier” I have found some complaints – a typical one reads... “This unit doesn't recognize a change in the room temperature until it’s changed by at least three degrees.”

I have had good experience with four other digital thermostats which brought the temperature dead on, or within one degree (displayed temp) of the set point.

I am still seeking an inexpensive but accurate (consumer grade) thermometer I can find. Maybe I should make my own digital thermometer. Maxim makes temp sensors and the specs on one series are ±0.5°C to ±3°C accuracy.
OK, some math guru out their please translate that to Fahrenheit for me.
"if it's not another it's one thing."

dyarker
Posts: 1752
Joined: Fri Aug 22, 2003 1:01 am
Location: Izmir, Turkiye; from Rochester, NY
Contact:

Re: Need accurate temperature display

Post by dyarker » Sun Jul 18, 2010 9:31 pm

±0.5°C to ±3°C = ±0.9°F to ±5.4°F
Which looks pretty sloppy; but as long as an individual sensor is consistantly off by the same amount, a correction factor can be put in firmware of uC it is connected to. (ie. If you buy 10 same part number sensors to make 10 thermometers, you'll likely need 10 different correction factors.)

For future reference:
To convert the degrees (like above) -
°F = °C * 9 / 5
°C = °F * 5 / 9
To convert temperatures -
X°F = (Y°C * 9 / 5) + 32
X°C = (Y°F -32) * 5 / 9

Cheers,
Dale Y

User avatar
reloadron
Posts: 519
Joined: Sat Jun 28, 2008 8:57 am
Location: Cleveland, Ohio
Contact:

Re: Need accurate temperature display

Post by reloadron » Mon Jul 19, 2010 1:45 am

A few years ago I made a simple little temperature conversion program that can be downloaded here. Simply download the .zip file then right click on it and let Windows unzip it and double click the executable. Just enter a numeric value in the top text box and the lower boxes will tell you that temperature in C and F. It was made for computer people who constantly want to convert temperatures and don't want to do the math or Google online converters. That may help make the conversions simple. :)

As to a good and accurate unit. My experience with home thermostats is they are not exactly high accuracy measuring devices or controllers. They have hysteresis for the reasons mentioned and their accuracy varies but typically +/- 3 degrees F or more isn't unusual. I don't know of any off the top that allow for actual calibration adjustment so it is a sort of crap shoot.

Ron

jimandy
Posts: 570
Joined: Sat Dec 04, 2004 1:01 am
Location: Birmingham AL USA
Contact:

Re: Need accurate temperature display

Post by jimandy » Mon Jul 19, 2010 7:02 am

Hmm. Maybe I am looking for the wrong thing. Rather than the most accurate (and cheap) thermometer I should ask how to get a precise temp reference in the human comfort range. Yes, boiling water would give me a 212 (at sea level) and ice cold water a 32 reference but I wonder if there is a way to generate an accurate reference of, say 72 degrees? The idea would be to apply a current to some device and it would warm (or cool) to that exact temperature.

Or perhaps I could use a dollop of gallium and keep raising the temperature in the room until it melted. The most accurate thermometer should read 85f, but, good golly, I don’t have any gallium and such an experiment would be tedious, probably not reliable, and somewhat like flogging a dead horse against the wall. :???:

Thanks all for your forbearance.
"if it's not another it's one thing."

Robert Reed
Posts: 2276
Joined: Wed Nov 24, 2004 1:01 am
Location: ASHTABULA,OHIO
Contact:

Re: Need accurate temperature display

Post by Robert Reed » Mon Jul 19, 2010 8:56 am

I have never been a great fan of somebody else telling me what my "ideal" room temperature should be. There are so many other factors that influence ones 'Comfort Zone'. For me it has always been a matter of more heat or less heat (same with cooling). I would not worry about actual temperature accuracy as repeatability being much more important. If you had the most accurate thermometer in the world to set your room at 72 degrees or whatever, what good would that be if you were still too cold or too warm. The last thermostats I designed for my home do not even measure degrees, but rather a comfort zone with fine graduations. Your body is the best measure of the "ideal' room temperature. Of course that is just my opinion.

User avatar
Bob Scott
Posts: 1192
Joined: Wed Nov 20, 2002 1:01 am
Location: Vancouver, BC
Contact:

Re: Need accurate temperature display

Post by Bob Scott » Mon Jul 19, 2010 9:15 am

Just a quick note: I would think that "0.5 to 3.0" would mean that the device is available in different grades of accuracy, like the alanog LM34 and LM35 devices. The ones with the "CZ" suffix are the most accurate and the most expensive. (See data sheets and prices at DigiKey.com)
-=VA7KOR=- My solar system includes Pluto.

jimandy
Posts: 570
Joined: Sat Dec 04, 2004 1:01 am
Location: Birmingham AL USA
Contact:

Re: Need accurate temperature display

Post by jimandy » Mon Jul 19, 2010 10:06 am

Robert, I agree with you, comfort is a personal and a relative thing. Also I believe comfort is actually based not only on temperature but humidity (is dew point relevant here?) I just want to find out which of my array of thermometers in this room is giving me the best guess at real temperature when I finally am comfortable. I don’t think 78 is it as I have always found 75 my personal best (according to other houses and thermostats) Of course wife disagrees.

My quest has to do with a lurking fear that that comparator in the suspect thermostat is not working right. I suppose it uses a digital comparator based on agreement between the binary values of the set point vs. the measured temp.. Seems to me that if the low order bit on the set point side was hung then I assume it would not agree by +- two degrees with some temperatures., i.e. measured at 76 could match a set point of either 75, 76 or 77. (I’m in my fuzzy area of math here). I remember finding memory problems in my early homebuilt PC was caused by a stuck low order bit address line and all odd addresses were giving me the same value as adjacent even addresses – and vice-versa.

Anyhow, I will go quietly now as I may be causing too much topic clutter.
"if it's not another it's one thing."

User avatar
reloadron
Posts: 519
Joined: Sat Jun 28, 2008 8:57 am
Location: Cleveland, Ohio
Contact:

Re: Need accurate temperature display

Post by reloadron » Mon Jul 19, 2010 3:45 pm

OK, curiosity prevailed so I went in and looked at the thermostat we have and discovered it is a Robertshaw thermostat. Actually no, I never knew what it was. Through Google I discovered it is this particular model. The wife has it set to maintain 74 Deg. F. I looked at the little digital readout which was indicating 74 Deg. F. and I hung one of the mercury in glass precision thermometers right beside it. The range was 0 to 125 Degrees F. The resolution was 1 degree F increments and easy to visually discern .5 degrees. I allowed about 45 minuets stabilization period.

The digital indicator and the thermometer were in perfect agreement. The outside air temperature is 80 degrees F. so the system isn't really working hard today. At 74 degrees with a 74 degree set point the compressor was running. When it dropped to 73 the system immediately shut down. When the temperature rose slowly to 76 degrees F. the compressor came on again. Enter the region of hysteresis we covered earlier. The bottom line here is the wall thermostat is within 1 degree of the digital indicated temperature with the thermostat. The thermostat went in when the central air went in installed by a friend with a new furnace several years ago. We also put considerable time and money into new insulation and windows. Even when we get into the 90s the system doesn't cycle much.

Yes, the right temperature is a matter of personal comfort. Therefore based on a few post it really doesn't matter what the numbers are as long as the individual is comfortable in the range the system provides I guess. I never gave it much thought as long as the system runs and it is warm in the winter and cool in the summer. I have no clue what the relative humidity is in here but it feels comfortable. I guess I could build a wet bulb, dry bulb hygrometer and measure it. :)

Ron

Post Reply

Who is online

Users browsing this forum: No registered users and 36 guests