And the real temperature is...

[bureau13]

OK, so I currently have three temperature-indicating devices in my tank, all of which show something different. I have a Little Time and Temp digital thermometer in my sump, showing 80.8 degrees F. I have the temp readout on my in-line chiller showing 79.7. I have my AQJr temp probe in the display reading 82.3 (all taken at the same time this afternoon). The AQJr will allow me to set an offset, if I feel its reading incorrectly, but who am I to say what is correct and what is not. Now, the AQJr probe is nearer the lights, but I have pretty decent turnover...I don't think the water has much time to heat up there relative to anywhere else in the system. At any rate, I bought a cheap glass thermometer at Petco and measured the temp at roughly the same location as the AQJr, and it showed around 80-81. Yay, so that agrees with the Little Time and Temp reading, so I'll set the AQJr to match.

Of course, its never that easy. I also just happened to buy a little Coralife digital thermometer for my QT tank, and its measuring about 2 degrees hotter than the glass thermometer, i.e. roughly equal to the AQJr!

Generally speaking, what do you all think about the accuracy of these devices? I would tend to say the simple glass thermometers should be consistently accurate...they're just not precise (begs the question how valuable is precision if its not accurate???) but I hesitate to throw all my trust in anything that cost $2.99.

jds

[mg426]

Good question. They are telling you something differant but all are in the accepatable range for a reef system. Mabe average them all and set the AQjr to the average.

[bureau13]

Hmmm, not a bad idea, however one of my problems is that I've set up my QT tank in the garage, and it gets pretty hot in the summer. The Coralife digital thermometer currently says its 85.5. If that's really 83.5 then I'm not as concerned, but closing on 86 degrees...that's making me a little nervous.

jds

[jcraft]

you can do what we do to set all our thermometers at work (restaurant- thermo's for temping out meats)

Fill up a glass full of ice, making sure it's packed full. Then add a lil bit of water- you don't want the ice to be floating around, but the water will fill in the gaps between the ice and give you a better reading.

Put the probe into the ice, make sure its fully covered and leave it. If the thermometer is accurate, the temp reading will be right around 32F- the temp that water freezes. I calibrated my Toms thermo this way- it took about 10 minutes, but the reading eventually bottomed out at 32.2F - pretty darn accurate if you ask me

HTH

[dots]

jcraft has the right idea, If he does not mind, I would like to add on why for good measure.

All mechanical devices have error and precision, and accuracy based on that. Being that precision is 1 degree or .01 degees off, how much error it is self explanatory, and accuracy is the ability to repeat that error with precision every time.

Obviously the higher quality the more expensive and hopefully the more precise, and accurate to the true measurement.

To figure this out for each instrument they need to be calibrated against a standard as mentioned. You can use standards such as ice water baths that will be close to 32 degrees F or boiling water. That is why they are 0 and 100 degrees respectivly in the Celsius scale.

Or use a lab grade mecury thermometer(mercury is still more accurate), to calibrate them.

By the way, I have that Coralife too, and don't trust it too much....

For what it is worth, I am a Mechanical Engineering major and last semester I took a measuments class and this was the first lab and the way he descibes it is how we did it......

[essexboy73]

Just wanted to add my two sense...i was relying on a 40$ difital job that i bought...the temp kept dropping and i was thinking that the heater accuracy was off, so i kept bumping it up...once i got to 88 degrees, i said to myself...what the ****...so i went out and bought a 5$ submersible thermometer and realized that i was about to have broiled fish for dinner....

Sometimes technology kills...

[ChemE]

I had a little trouble following sacramentodots's explanation so please allow me to rehash.

All instruments have instrumental error which is how far they deviate from a true value. In the case of that cheap glass thermometer, it probably has an instrumental error of a few percent, so if the true temperature is 100F it may read 98F or 102F.

Now there is also an error associated with reading an instrument. The glass thermometer probably has a little tiny mark for each degree. So if the mercury (or methanol w/ red dye if it is really cheap) is between the 80F mark and 81F mark you can't tell if it is 80.7F or 80.2F. There is a reading error of 0.5F. Reading error is always half the finest scale of the instrument.

Now we come to the oft confused difference between accuracy and precision. Accuracy is how close we get to the true value. Precision is how repeatably we can make a measurement.

This sharpshooter is accurate because the bullet holes are all very close to the bullseye but not very precise because each shot is far from the others.

This sharpshooter is very precise because the shots are clustered tightly but not very accurate because all the shots are somewhat far from the bullseye.


OK, all that said, what you really want in a reef tank is a precise thermometer not an accurate one. We don't actually care if the tank is 78F or 82F but if the temperature doesn't change, I want to read the same slightly wrong (inaccurate) temperature over and over (precise). It should be noted that an instruement can't be accurate without first being precise (just think about it).

Now, you could calibrate your thermometer by sticking it in an ice bath and boiling water (212F) but this is pretty much worthless for us because you'd be calibrating over a much much wider range than where you want to use it. Ideally, you'd do a 3-point calibration with standard at 75F, 80F, and 85F. This would give you great confidence in your measurements between 75F and 85F but wouldn't be much help if you wanted to measure temperatures in a freezer. See what we're doing here.

[blenny100]

I have been faced with this problem myself. The best (and simplest) way that I found, was to (1) Make sure I was "feeling OK", (2) Use the devices to take my own (body) temperature (stick the thermometer under my arm!)

Since normal body temp if 98.6, but readings taken in the armpit are about 0.5 - 1 degree lower, you can use this to figure out how accurate your devices are.

It is not foolproof. Body temp does vary - but I found it to be useful. Plus you are calibrating somewhat closer to the temp of a reef tank, as compared to calibrating at boiling point or freezing point.

[bureau13]

If you want a more accurate body temp there are other places you can stick that thermometer

Ahem...enough of that. I was into the ice water thing, but 'tis true...that's not in the range we care about, so I'm not sure that will tell me how accurate it is for our uses.

As for the precision vs. accuracy thing...I agree that we generally want precision, however in my example above...how close to cooking my QT tank am I coming...I would like to see some reasonable accuracy.

It appears we are not closer to an answer as to how best to calibrate temp readings for the range we care about, short of acquiring expensive test equipment or known calibrated thermometers. Damn...

jds

[bad351stang]

Advertising removed...

Ryan
RC Staff

[bad351stang]

The ice water bath is not scientifically accurate but it's as close to 32°F as we can get with the equipment available to us. With a thermoelectric chip, a variable dc power supply and a little insulation and I can hit exactly 0°C but that doesn't help you. Boiling water can't go beyond the transition state in liquid form, excessive heat is released with the vapor, this is why steam is hotter and carries energy that can move locomotives. When you calibrate a digital thermometer it doesn't matter what range you expect to use it in. A thermistor that is 3% off at 100°C will be the same 3% off with the same + or - offset at 0°C, or 23°C, or -75°C but you get the picture. The % is not a percent of the actual figure in solid state electronics, it's a percent of accuracy and involves manufacturing process. They shoot for a resistance figure and hit it within an industry acceptable level of accuracy but once its made... it's made, like a coin die. The same accuracy lends to analog thermometers and the inaccuracy derives from the manufacturing quality of the expantion tube and/or graduation on the scale.

The other obviously knowledgeable replies you received were good advice. It was good reading and typical chem. 101 stuff but didn't answer your question. I hope I have helped. Calibrate an digital thermometer in your choice of method and test all of your other thermometers against it. At least all of your thermometers will be reading from the same page.

Good Luck,
Steve

Originally posted by bureau13
If you want a more accurate body temp there are other places you can stick that thermometer

Ahem...enough of that. I was into the ice water thing, but 'tis true...that's not in the range we care about, so I'm not sure that will tell me how accurate it is for our uses.

As for the precision vs. accuracy thing...I agree that we generally want precision, however in my example above...how close to cooking my QT tank am I coming...I would like to see some reasonable accuracy.

It appears we are not closer to an answer as to how best to calibrate temp readings for the range we care about, short of acquiring expensive test equipment or known calibrated thermometers. Damn...

jds

[MJAnderson]

I just went through the same thing. Bought an electronic one that said my tank was 81.5 when my glass one said 80. Returned it and got a different one...said it was 79 when my glass one (still) said 80. Went to my LFS and asked them to measure a glass of water with their Tunze temp controller. Figured those Germans would have the best inistruments. My glass thermometer and the Tunze matched. Not sure why an electronic one has such a hard time.

[bureau13]

Good stuff Steve! I didn't know the cheapo Walmart digital thermometers were adjustable. You'd think the one sold for aquarium use would be as well, but they don't appear to be.

jds

[bad351stang]

They all should be adjustable, It's a simple tiny 5 cent part and minor circuitry on an easy to edit printed circuit board and some very basic EEPROM programming. I have numerous digital thermometers. Some cost me $50. You cant adjust them and they are all over the map. Of 3 of the same $50 digitals one is right on (+/- .5°F), another is -5° and the other is -2°. The one that reads 5° low, -5 is 1.13% error in the operational range and is 2.77% in the range of freezing to boiling. Knowing the error is the key though.

If you open it up you may find a very little box with a tiny slotted screwdriver hole in it or it may not even be boxed. If you find one you can adjust it.

Steve

Originally posted by bureau13
Good stuff Steve! I didn't know the cheapo Walmart digital thermometers were adjustable. You'd think the one sold for aquarium use would be as well, but they don't appear to be.

jds

[dots]

ChemE, Thank you for restating what I was saying.....I remember the textbooks using that same visual of the dartboard and was thinking of that. I think there could be another variable called resolution is something we are talking about also, from that the error as how you described it comes into play. I was trying to convey the difference between measument of 1 and 1.000 and how and why the 1.000 is more accurate and more costly.

You are correct about the range of operation and those need to be double checked, before dunking it in the two ranges.

[ChemE]

No problem. What sacramentodots is referring to is the precision of the instrument i.e. how finely it is capable of measuring. In science when something is 1 that means that the instrument can either measure 0, 1, or 2 and the reading was 1. This also comes with the built in implication that that number is 1±0.5. When something is 1.000 that means that the instrument can either measure 1.999, 1.000, or 1.001 and the measurement came out to be 1.000. This implies that the measurement is 1.000±0.005.

All of this discussion feeds into something called error propagation it turns out to be incredibly useful for conducting careful scientific measurements. Anyone who neglects this is not to be taken seriously.