Pro/Forums - sensor calibration

Pro/Forums (http://forums.procooling.com/vbb/index.php)

- Testing and Benchmarking (http://forums.procooling.com/vbb/forumdisplay.php?f=61)

- - sensor calibration (http://forums.procooling.com/vbb/showthread.php?t=12585)

sensor calibration

you read a temp, is it 'real' ?
what is the purpose of calibration ?
- many certs state "in spec", but this info while welcome is not useful
what is needed is a calibration curve indicating the sensor/instrument's deviation

the first graph is an air cal of 2 RTDs and 2 TCs, the second a liquid cal of 3 RTDs;
note that the yellow RTD is the same in both

(this was the delay Joe)

Re: sensor calibration

This begs the question; how does one run a calibration, which we've been over several times.

Then we have 3, 4, 5 and 6 point calibrations, and how it's applied.

You have a Quartz reference, most DIY'ers will not.

As a DIY tester, I'll have access to the following 'real' temps:
-freezing (0.01 deg C)
-boiling
-a mercury thermometer, ranged from 19 deg C to 27 deg C, in 0.1 deg C increments (NIST traceable, but I do not hold the cert, I assume good to 0.1 deg C, until it is abused).

(may be able to get various freezing/boiling of other substances)

I believe I'll be able to run a limited 3 point cal, but over a range that is far wider than interesting/relevant.

Or I could run a 3+ point cal using only the mercury thermometer, over its (narrow and interested) range, but I highly suspect it'll be useless (too low res).

Re: sensor calibration

when you have it caled ask for the points you want, say 5°C increments from 20 to 45
use The Meter Shop on Director's Row
w/o a good cal you are lost

of the many tools I've bought, no single one has had the impact on accuracy as the $900 spent on that thermometer (w/2 probes, all in cal)

EDIT
you after a sensor cal procedure ?

Re: sensor calibration

Sounds good.

I'm still pondering wether or not I should calibrate it myself, but I do plan, at least, to check temperature readings on a regular basis (and log all the results).

Since I've opted to build the 24 bit ADC, there is going to be a calibration procedure involved. I'm thinking more and more about having a starting point, i.e. one of those Fluke units, which I'd send out for calibration.

Not sure what you mean by "The Meter Shop on Director's Row" ?

With the wind tunnel, I've got more temp probes to keep up with than I care to count!

Re: sensor calibration

if you are building your own, you will need an in-house ref (not that mercury thing)
0.01 ref res = 0.1 cal capability
The Meter Shop is on Director's Row (street) in Houston (610 & 290)

Re: sensor calibration

Nice, thanks for the info.

Re: sensor calibration

Another nice one!

"0.07 deg F" accuracy; do you believe that to be possible? Am asking because I don't know. Would guess it would be achievable because of the EM shielding?!?

Re: sensor calibration

multiple sensors, sure
that relates to the sensors only, see how they are read ?

I would note that I can cal my setup a lot closer, but its ~6 hours of work + the necessary instruments

Re: sensor calibration

I wouldnt choose water boiling point as a reference, well i'm used to live not at sea level... Here water boils sensibly lower than the reference 100°C, and even at sea level it depends on atmospheric conditions. Or you make it boil in a pressure regulated chamber :P

Re: sensor calibration

about only you, pH, and TM could make me respond, most burnt out with the clue level here

neither, a bath or environmental chamber depending on water or air cal
the key is an HP 2804A w/dual sensors, all in cal of course (0.001°C resolution, but sloooow)
I cal in 5°C increments from 15 to 45, 70 if a heat die RTD; then use that curve in a spreadsheet

I'm in Houston so the Inst Shop cal (freezing, midpoint, boiling) is ok, just not sufficient for my desired accuracy
note that the cal cert does specify the barometric pressure, so it can be corrected for
(got to admit I've yet to see anyone actually do so, self included)

Re: sensor calibration

I have not investigated this well, however... I have seen (in scientific american) an article on producing an accurate temperature via triple point conditions. Doable by a hobbyist too. Let me plagarize below. BTW, I hope that with this, I can rouse your interest as well, Bill.

Re: sensor calibration

actually i've seen it, in the Portuguese Institute of Quality (they "supply" the standard cal).
I got the tour. Anyway, it's done in several tubes (3 if i recall, per container) in a aluminium pressure chamber. If that's "hobbyist" friendly, then hell, i really have to move to the US. :)

http://www.iop.org/EJ/S/0/M32185/pTF...2/9/me7209.pdf
Hmm. Brians : this? Sounds simple. But these things never are.

PS: had my laughts when they used a king size enviormental chamber to test a small pressure device for tyres. The thing is as big as a living room, yet the device is the size of a bread box.

Re: sensor calibration

This copied from Scientific American. This is a great magazine, guys, well worth reading and supporting with real money.

Quote:

Originally Posted by Scientific American

To build Schmermund's thermometer, contact a local glassblower to purchase a Pyrex tube 30 centimeters (12 inches) long and eight millimeters (0.3 inch) in diameter. At one end of the tube, have the glassblower form a receptacle that is five centimeters (two inches) long for the RTD sensor.

Next, attach lead wires to the sensor. If you solder the leads or use wires insulated with plastic, you'll be restricted to temperatures below the melting point of those substances. That's not a problem for many applications. To allow the maximum possible range of temperatures, however, Schmermund spot-welds the RTD to bare 10-mil nickel wires that he then insulates in thin Pyrex sleeves. He gets these sleeves in 46- centimeter lengths from a local glassblower, but capillary tubes, which are available from any scientific supply house, work equally well when strung on the wire like beads on a necklace.

For a thermometer that will be used with a four-wire ohm meter, Schmermund bundles four of his long tubes and delicately tapes them together at one end. He then bends two one-meter lengths of nickel wire in half and threads each half through a different tube from the untaped end. Finally, he spot-welds the RTD to the bends in the two wires. (Note: If you will not be making four-wire measurements, simply connect one wire to each of the RTD leads.)

To secure the insides of the device and to thwart convection currents from forming, Schmermund packs the instrument with tiny glass beads that are only about 25 microns in diameter. These are expensive and must be purchased from a scientific supply house. Fortunately, fine silica sand (grit 30 or greater) also does the job. You can purchase a 23-kilogram (50-pound) sack from a hardware store for just a few dollars.

Because any moisture that becomes trapped inside the thermometer will distort your readings, all water must be driven from both the filler and the glassware before assembly. Bake everything, including the entire sensor assembly, at 250 degrees F for approximately two hours.

Image: DANIELS & DANIELS
PLASTIC CANISTER protects a terminal strip that connects the thermometer to a cable leading to a digital multimeter.
You must complete the next steps while everything is hot, so be sure to exercise the proper care by wearing gloves, an eye shield and protective clothing. Secure the large tube in a vise. A clean rag wrapped around each jaw will allow you to hold the glass tube firmly without breaking it. Insert the RTD assembly into the tube and use a small glass funnel to pour in enough of the desiccated sand to cover the sensor completely. Lift the assembly just a bit to make sure the RTD is suspended about two millimeters above the bottom of the well, without it touching the glass wall. Remove the tape and slowly fill the tube with hot sand to within about half a centimeter from the top, stopping frequently to tap the glass with a pencil to consolidate the material

Hermetically seal the thermometer by topping off the sand with glue from a hot-glue gun. If you're using uninsulated wires, heat them with a hair dryer for a few seconds before the adhesive sets so that the wires will seat themselves into the glue.

There's more there, so please follow the link. I'm unable to find out if you can get all of Scientific American's Amateur Scientist column on CD-ROM though. The canonical collection is sold out, but a $29.95 version for "all" the projects might be available here at Amazon. Search for science fair and tinkers guild (no apostrophe in tinkers).

Now that solves an issue of a single point, if that is accurate enough. For multiple point calibration, you need to either use something that is known to be linear temperature through the interesting range (is a mercury thermometer linear enough?), or use other temperature set points like the melting point of mercury and watch the calibrated probe as the mercury does the phase change. I think that mercury is less sensitive to atmospheric pressure than water, although, you could put it into a sealed cell and try to measure the temp through the glass walls.

Anyways, hope this is interesting.

Re: sensor calibration

Quote:

Originally Posted by TerraMex

This article is much more hobby friendly, TM, and could cost less than $150 (article says $100, but inflation happens).

As for the size of the env chamber... real accuracy is expensive and space consuming.

Edit: btw, ordered a CD from amazon of those Amateur Scientists projects. Will see how good it is.

Re: sensor calibration

Also, found the prior thread that talked about this. It is by Ice Czar here, and I see Bill's objection to the issue. RTDs are NOT linear through the range we are interested in, so this calibration may be insufficient without other physical references.

Re: sensor calibration

Honestly i dont really like SA and PM (Scientific American and Popular Mechanic) because of the holes on the info they provide. It's like that Dilbert comment "Build your own helicopter from common household items, on PM!."

The first problem is the cell itself. I admit that small variations in wall thickness might not be a big issue, but when you increase the required accuracy, they could be. Just saying.
So having some guy from your local glass shop make one ... well. I'd go to a medical supply, or even scientific supply store and get one. I'd venture that it's not that expensive.

Another point is that glue thing. regular glue doesnt attach to glass very well, and neither for a long time. And using (as it stated) sand, which creates a very diffuse contact point, with glue, doesn't seem very stable, specially through temperature differences.

Third point is very simple. Any welding or soldering on sensor is something to avoid. Any weld made will change the ohmic value of the sensor, per wire. Add that to the ohmic value of the wire, which you should know, and size of each, which sould remain equal, and you get a mess to calibrate.

but hey, that's why labs exist :) .

PS1: for that particular range, you really have to use a platinum resistance thermometer (PRT).

PS2: for the ones that haven't figured it out what a "cell" is supposed to be:
(10m, paint! -> best app ever)

Re: sensor calibration

there is a guy selling these things pretty regularly over the last several years on eBay (search Schmermund)
for different temps different matls are used
the thermometer is only $350, but the cells . . . whew

it still gets back to the PRTD's non-linearity over our range of interest
no way to avoid creating a cal curve that I know of, at 1/3 DIN they are pretty much the same - but still have a curve

of course, on the censored forum the resident wizard has concluded that 0.1°C resolution is sufficient for WCing component testing, we do love to pick the pepper

Re: sensor calibration

lazy again?
http://cgi.ebay.com/ICE-POINT-CELL-S...QQcmdZViewItem
http://cgi.ebay.com/SCHMERMUND-DUAL-...QQcmdZViewItem
750 for a bong? :)

Eh. Anyway, they (cells) look decent.

Re: sensor calibration

there is an ad running now during the Olympics showing a patient receiving what the Dr. had removed in an operation
his 'lazybone'

take mine away and I'd be jello
no TM, the pride is in doing - not being shown
I'm not sure that guy has ever sold one, rather thin market I suspect

Re: sensor calibration

thin indeed. but i suspect he had it done for "pennies".
i checked his feedback, a few sold (that i could see):
http://cgi.ebay.com/ws/eBayISAPI.dll...tem=7578300064
example.

Re: sensor calibration

Quote:

Originally Posted by BillA

it still gets back to the PRTD's non-linearity over our range of interest
no way to avoid creating a cal curve that I know of, at 1/3 DIN they are pretty much the same - but still have a curve

What do the labs use, then? A very linear measurement device that is calibrated to triple point water and something else like melting point of <X> ? Do they look at emitted light frequency for dark body radiation for some known substance?

There's a limit to the number of reference points you can create until you start using another measurement device like boiling point of water at X mm Hg and then try to accurately measure air pressure.

Quote:

Originally Posted by BillA

of course, on the censored forum the resident wizard has concluded that 0.1°C resolution is sufficient for WCing component testing, we do love to pick the pepper

censored forum? need 0.01C to test some devices, esp rads. not sure how sensitive you need to be when testing tubing. :D

Re: sensor calibration

Quote:

Originally Posted by Brians256

What do the labs use, then? . . . . . .
censored forum? need 0.01C to test some devices, esp rads. not sure how sensitive you need to be when testing tubing. :D

I cannot speak for 'scientific' labs, though others here can/could if they wished to do so
industry (where I live) generally use ASTM or similar type equipment and procedures; good enough to produce repeatable results over time and between facilities
- the 'acceptability' of data is quite predicated on others being able to replicate it

individual companies will often develop in-house procedures for greater precision and accuracy;
in Intel these are called BKMs (Best Known Method) and the writers of such are held in very high esteem, and the docs are Secret
- one such that made it into the 'public domain' is the TC to IHS soldering procedure in the stepped groove (but I would note that a TC is still used) in the P4 docs
being who I am I attempted to use an RTD instead bit could not source a 4-wire to fit into the groove, lol

I have led many here down a merry path with my personal pursuit of ever greater 'accuracy' on a budget of pennies;
but this is the consequence of what I was doing - product design, where it was essential to be able to distinguish the very small differences resulting from very small changes

now all this is wonderful for comparative product testing also, but the minutiae so exposed are generally invisible to a typical user
yes, I started the WCing 'spec wars'; but I did also elevate the level of performance in available products

returning to the original question "What do the labs use, then?";
ideally the SAME equipment used by the company for whom the data is being generated; why Intel provides TTVs, eh ?
OR
equipment and procedures SO GOOD that a room full of double PhDs will record the data points and murmur asset (what I like, lol)
- this was done recently (by some friends) with a vacuum bell on the die backside to eliminate the heat flux question, but the temp was still taken with a (calibrated) TC

refer back to the cal curves in the initial post, these can only be done with a certain type of setup
- and are only necessary for those after greater accuracy than a thermometer with a cal sticker can provide
- the greater the resolution, the greater the possible differentiation
(do note the difference in both the cal and instrument 'quality', but as I caled them they are all good - understanding that the indicated temp is NOT the 'real' temp)

this is procooling, zeros are our friends

Re: sensor calibration

Quote:

Originally Posted by BillA

Sure. Was just interested in how it's done. Saw ASTM.ORG but didn't dig through their stuff. That's the real thing, of course. But, that costs money.

Quote:

Originally Posted by BillA

this is procooling, zeros are our friends

Sure. Didn't know if there was a shortcut to getting more zeroes. I'm not going to comment on overall test setup though. I've already put my one cent in and it didn't seem to matter. I was more interested in knowing how lab instruments get calibrated to such exacting standards. Perhaps it's cheaper to get an ultra accurate and high resolution instrument calibrated and then use it to calibrate lesser instruments than try to use physical "set points" to calibrate each instrument for the labs. I say cheaper for the calibration businesses, not us. For us, it'd be cheaper to get our instruments calibrated by one of those businesses if we want trustable results at great resolution.