EEVblog #419 - Thermocouple Tutorial

[EEVblog]

Everything you need to know about how Thermocouples work.
K type thermocouples, the Seebeck effect, the Seebeck coefficient, and cold junction compensation.
Along with some practical measurements with a multimeter to demonstrate the effect.



Dave.

[nessatse]

Dave,


Your explanation of the NIST tables is unfortunately incorrect.  The horizontal axis represents the units of the junction temperature (0-10°C), to  be added to the vertical temperature which increments in 10's.  It has nothing to do with the cold junction temperature  which is assumed to be 0°C.

[jucole]

doh you beat me to it! - yes the tables provide all the voltages for each temperature from 0 to -270 using a cold junction of 0c

[EEVblog]

Yes, oops, have already corrected that in annotation.

Dave.

[Salas]

F87V reads just OPEn without the their brown Tcouple inserted. Any idea if it has an internal sensor to compare or it works otherwise?

[Mark Bolton]

I only just stumbled on this blog a few days ago and it is fortuitous that the subject of thermocouples comes up now. I am building an airplane and have a bunch of thermo couples and old gauges that i am hoping to put into service.  Cylinder head / Oil temp  - 400 / 200 deg F. 

[EEVblog]

F87V reads just OPEn without the their brown Tcouple inserted. Any idea if it has an internal sensor to compare or it works otherwise?

Certainly would have an internal temp sensor, it's just not displaying it and has an open detection mechanism. It could have simply always displayed the compensated temp all the time, but then if the probe is broken and you get only the ambient temp displayed, that could be dangerous. i.e not detecting high temps. So better to display OPEN.

Dave.

[Jon Chandler]

I've come up with an inexpensive thermocouple circuit based on the TI TMP512/TMP513 Temperature and Power Supply SystemMonitor.  The chip has a local temperature sensor for cold junction compensation, and the 40 mV range of the ADC (for measuring shunt voltage) provides good resolution of the thermocouple output.  It's a lower cost alternative to the dedicated chips from Maxim and AD.

The article describes the software written in Swordfish Basic fro PIC18F series micros.  I use a lookup table for linear results.

A Thermocouple Measurement Circuit With Swordfish Software

[Ketturi]

This was very interesting, keep up Dave! Although table was bit goof-up

Is there way to fix those connections in thermocouples? I broke mine but was able to fix wires together by arc welding them, but now it's showing temperature couple degrees off.

[Psi]

I thought it would be cool to show a bit more energy being produced from the effect.
So i soldered a lamp to a peltier block and took a blowtorch to it

Here's the result.


I didn't think to measure the voltage and current.
If anyone really wants to know i can do it again.
Maybe put the steel block into the freezer overnight, that should max it out.

[HKJ]

F87V reads just OPEn without the their brown Tcouple inserted. Any idea if it has an internal sensor to compare or it works otherwise?

It does, just short the input terminals with a wire and you can read the internal sensor.

[Salas]

F87V reads just OPEn without the their brown Tcouple inserted. Any idea if it has an internal sensor to compare or it works otherwise?

It does, just short the input terminals with a wire and you can read the internal sensor.

Indeed. Just a link and we got ourselves a quick room temp indicator. 

[SLJ]

I found this a few months ago.  You don't need a temp option on your DMM or analog meter.  Just convert the ohms reading on your meter to the temp (F or C) listed on the conversion chart on the side of the probe.  It does work and it's pretty close.  Cost in 1973 was $12.50 U.S..

[LaurenceW]

SLJ, I think what you have there is "just" a Thermistor - a temperature dependant resistor element. It produces a change in resistance, not a voltage difference, when heated. the relationship between the resistance and temperature is either non-linear, or certainly not "convenient" (i.e 1 ohm per volt or similar), so it needs a scale to cross-refer the resistance to a temperature.

[SLJ]

Yes,  as I stated there is a conversion chart on the side of the probe (ohms to F or C) and a larger paper chart in the case.  Not as accurate as my 87V and it does not have the temp range but I found it to be an interesting item.

Here's the chart:  http://www.stevenjohnson.com/data/thermy.pdf

[Jeff1946]

Couple of comments, Type K is often used because the metals are oxidation resistant.  A simple way to make a junction is twist the wires together than "fuse" them by striking with a hammer against a hard surface.   In the lab we melted them together using a hydrogen oxygen torch with a slightly reducing flame.

[bitwelder]

Question: let's say that the tip of a thermocouple probe got spoiled/damaged. Would it be possible just to cut away a few cm of the wires and join them together on the tip, to have the probe working again?

[SeanB]

Yes, either you spot weld them or arc or gas weld them together, or just quick and dirty wind them together tightly. Some were stir welded together.

A common use for them is in gas valves, where they keep gas flowing to the pilot light, but switch it off if the pilot goes out for any reason. May only be millivolts, but a big one can supply a lot of current, enough to keep a valve open.

[nicknails]

Analog Devices also has a nice whiteboard series about thermocouples.

[AlfBaz]

I remember looking into thermocouples years and years ago, and I wrote some C code to calculate the voltage produced by the thermocouple of choice given the temperature.

It can also create a table of all the nist values for the temperature range specified and saves it to a file. Rename the file to csv, txt or what ever and it will look pretty much the same as the NIST tables Dave showed.

In the attached zip you have an exe which requires no dll's other than what comes with a bog standard install of windows and just runs without any install bs. It worked in windows 98 and It still works on windows7. The source code is attached also and all the math is done in a function at the end of the file

Thought it may be of interest

[Skimask]

Ahhh...memories....
Great video...

Short story made long...

I used to work B-52's, specifically instruments/autopilot.
Way back when I was a young technician with not a lot of credibility, we had a problem concerning an engine on a particular spot that was overheating.  The jet would land, we'd inspect it the engine, find no evidence of overheating, but ended up changing out a few engines (at great cost, a few million $$$ each) because that's what the book said to do.

I didn't come into the picture until the 3rd engine had been changed.  Previously the wiring had been checked (numerous times or so I was told), gauges cal'd, engines ran up...  Everything checked good....On the ground.

The problem only seemed to happen while it was flying at altitude, with an ambient OAT of...well...we'll say about -50C.  AND the engines were overheating by just about double that amount.  Normal max temp was about 485C and the engines would get up to about 600C before they shut them down.

In this particular case, the book told us to hook up a Fluke 87 in resistance mode, take a reading in "one direction" (e.g. + to +, - to -), then take another reading in "the other direction" (e.g. + to -, - to +), and take the average of those 2 readings to get an overall resistance value for the wiring.  That would tell us if any of the terminals needed cleaning, repair, etc.  I knew why this particular method worked (e.g. the way it was connected, the Fluke was either fighting or going along with the current being generated in the thermocouples).

Well, short story made long (again), I ended up sitting down one night with a chart of wire resistances for the various wire types (which I found in the back of my brand new just bought a week before book "The Art of Electronics"), along with a wiring diagram for the aircraft.
I calculated a fair estimate of each leg's resistance for each wire's type, wrote it down, graphed it, etc.  A lot of math extending way out to the right of the decimal point and a lot of math that nobody else could begin to understand.

At the end of the night, I came to the conclusion that the wiring was swapped (e.g. + for -) where the engine wiring meets the wing.  Problem was to get to that wiring meant a lot of work for a lot of people to pull panels from the aircraft, technicians turning screwdrivers, metal technicians drilling out stripped out screws, and so on.

At the morning meeting (with a lot of heavy hitters and much higher supervision than I cared to deal with since I was basically a young punk), I showed my supervision all of my math (not that one person in the room understood any of it), and tried to explain what I thought the problem was.
At the end of this hours long (seemed like, actually maybe 20 minutes) explanation of what was going on, the big boss (the Wing Commander) directed the flightline expeditor (my boss) to assembled a team of screwdriver turners and screw-taker-outers to assist me....All the while I was getting these looks like I'm a big time dumbass wasting other peoples time.  At this point, I'd been on shift for about 13 hours and was getting anxious.

We got a half dozen folks out there, started turning screws, removing panels, etc.  I was up there pulling screws out with my paperwork showing how I thought wire #1 and wire #2 were misplaced on terminal #2 and terminal #1 respectively.
After about an hour, we got all the panels off and I was the first to stick my head in there.
and...
Came back out with a grin on my face.
I motioned for the flightline expeditor (my boss) to come over and stick his head in there take a look see.
He came out with a look of "W...T...F..." on his face.
30 seconds later, the wires were swapped.

The worst part about this whole thing....
3 days later was this aircraft's next and final flight to the "Boneyard" at Davis-Monthan A.F.B. in Tucson, Az.
You can see that exact aircraft, 61-027 sitting in the "Boneyard", across the street, right at the junction of Quijota St. and S. Wilmot Rd, circled, with an arrow pointing to the troublesome engine #5 and the panel is under the wing right at the pod root.

Like I said, short story made long.


EDIT:  Forgot one small detail...
We figured out later on that when that aircraft had been in depot maintenance, that particular dual engine pod had been removed entirely to facilitate corrosion work on the leading edge and those particular wires that were swapped were 2 of 12 wires NOT in some sort of cannon plug or multi-pin connector...terminal lugs all the way for that bunch.

[Wytnucls]

Great story Skimask.
It wouldn't be the first time that electrical wiring was connected the wrong way on an aircraft, nor the last.
I remember an incident where a Lufthansa A-320 got airborne with the Captain's flight controls connections reversed. A sharp First Officer took over, after take-off, saving the day.

[Jon Chandler]

I used to work in a naval shipyard and remember a couple weapons systems technicians checking out some wiring on a submarine one day.  One guy had the meter, while the second guy was reading the test procedure and recording the results.

"Measure from terminal 1 to terminal 7.  Five volts, right?"

"Yep."

"Measure from terminal 3 to terminal 12. Zero volts, right?"

"Yep."

Amazingly, all of the measurements I heard were exactly right.  Imagine that?  Perhaps these guys helped the Air Force sometime?

[Skimask]

Amazingly, all of the measurements I heard were exactly right.  Imagine that?  Perhaps these guys helped the Air Force sometime?

Normally, I'd say, yep, should've been an obvious fix, except that the problem didn't present itself until there was a marked ambient temperature change.  I watched the guys do the calibrations, did one myself, watched the engine run ops checks, we did everything by the book, just like we're supposed to do to keep from getting somebody killed.  This is one of those cases where you expected the guys weeks/months before you to do the right work as well.  And I was the only one with enough balls (e.g. brains) to think outside the box.
And what did I get for a great fix?  An "Atta-boy" from my shop chief at the time.  Didn't hear squat from the higher ups...as usual.  Such is the life of a maintainer...

[SeanB]

Similar, i got a batch of computers all from the same plane, all marked as fail system test. Politely phoned and asked what the problem was, then had them do the very complex swap out of the big air data computer ( needs autopilot removed, fuel system drained so you can get the tank on top of it out then remove a whole lot of panels held in by rivets, and not blind ones but solid ones, and then take it out , replace and put the big pile back together then test fly it) that was out of tolerance on the one output transformer. They did not like me for that.

Best still was the computer that came in from the training flight of the OC, written on the sheet "First bomb fell on target". I had to phone him ( big guy, but nice enough) and congratulate him, then ask for more detail. He selected a spread of 5 practise bombs, but they were released early, with no1 instead of no3 landing on target. yet another call to my old trainer to run another self check and change something else, but this was much easier, just 4 screws and 3 plugs inside the cockpit on the right.

[xDR1TeK]

Hey Skimask, cool story man. Just to prove, pen and paper can stretch a long way.
Very moving, you didn't need your peers' approval, it is gratifying alone to know that you can do it. It was all you. I don't normally read long posts, but yours was very well written.

Dave, you are the man, thanks for the video. Love it. Many thumbs up.

Oh, just remembered, the book you mentioned, The Art of Electronics, fantastic book.
My instructor advised me to get it. Old book and awfully nice.

[Paul Moir]

Yes, either you spot weld them or arc or gas weld them together, or just quick and dirty wind them together tightly. Some were stir welded together.

You can also silver hard solder them if you need a reliable & small connection but don't have the welding gear.  A propane torch will do.  I used to do them that way until I got a TIG welder.

One thing Dave touched on but didn't really go much into is that thermocouple wire isn't all that accurate.  Special grade K wire for example is +-1.1degC out of the box.  You can compensate for that with calibration, but that can get tricky.

[Wytnucls]

This is how temperature measurement is implemented on the Fluke 17B:
I don't pretend to understand it, but someone might.

[AlfBaz]

I wonder what the voltages for VDD, VDDA and VSS are?
Also whats the res of the cct hanging off pin 3 on U6?

My guess is U6 is a diff amp that is subtracting the temp measured by TMP35 at the junction from the thermocouple temperature

[Wytnucls]

Yes . I think so too. Voltage differential.
Here is TMP35 datasheet:
http://www.analog.com/static/imported-files/data_sheets/TMP35_36_37.pdf
And here is the whole circuit:

[EEVblog]

I used to work in a naval shipyard and remember a couple weapons systems technicians checking out some wiring on a submarine one day.  One guy had the meter, while the second guy was reading the test procedure and recording the results.
"Measure from terminal 1 to terminal 7.  Five volts, right?"
"Yep."
"Measure from terminal 3 to terminal 12. Zero volts, right?"
"Yep."
Amazingly, all of the measurements I heard were exactly right.  Imagine that?  Perhaps these guys helped the Air Force sometime?

Yep, that's the military, they have reams of procedures and check lists for everything, and they never change them.
That's why companies like Fluke have to offer for example, an average responding version of the Fluke 28-II, instead of the True RMS one. Because all the procedures and values are based around old average responding meters.

Dave.

[Jon Chandler]

I think their test procedure might have been more valid if the guy making the readings was calling them out, rather than simply agreeing with every number called out...but that's just me.

[Skimask]

Yep, that's the military, they have reams of procedures and check lists for everything, and they never change them.
That's why companies like Fluke have to offer for example, an average responding version of the Fluke 28-II, instead of the True RMS one. Because all the procedures and values are based around old average responding meters.

Dave.

That's no joke...
20+ years of working aircraft directly, plenty of times troubleshooting a problem when I knew something was wrong, but the "book" said it was correct.  And not a thing I could do about it at that particular moment because the paperwork to make the changes had to go up the chain and get back down again, and if I was to go with my gut instinct and something went wrong, it was literally my ass in a wringer.  On the upside, a person could call the problem worse than it actually was and ground the aircraft, if anything just to buy some time to get the phone calls in and get authorization to do whatever it was that needed to be done correctly.
It's a bit better these days in the U.S.A.F.  A few years ago, they put all of the tech orders on ruggedized laptops, each one with the relevant tech info needed to do the job.  Updates come daily thru the network when the laptops are put in "the rack".  Changes can take literally hours vs. weeks/months like it used to.

I digress...A tech rep for Panasonic came to the shop one day to show us their design and how good it was.  He had three of them laid out on the desk for us to play with and ask questions.  I sat down in front of one, played with it, and pushed it off the table, ~3ft drop to the carpeted floor.  Screen cracked, hard drive crashed (heard the click of death when we picked it up), and a hinge cracked.  Got "the look" from the tech rep and from some of my supervision.  I look, laughed, and walked away as did a few other fellow maintainers.
So much for ruggedized...

On the Fluke thing.  Back in the day, the books only called for the PSM-37 meters to be used for this particular job of checking the E.G.T. wiring.  Old dinosaur of a meter, but at the same time, it didn't care if it was cold, hot, wet, whatever.  When the PSM-37 went out of style, they switched over to the 8025A, which didn't like to work below 20F ambient temp, which is a problem half the year up here in North Dakota.  Had to rewrite the books 3 times for that, once for the meter switch, again for the cautions about temperature, and again when we got the 8025B, which worked when it was cold.
Then we switched over to the Fluke 87 series...and all was well...

On another side-side note, before the PSM-37, we had to use an actual Wheatstone Bridge circuit to take readings on the E.G.T. wiring...bah!  Who came up with that bright idea?

[SeanB]

When I was an appy I had to go and read through all the STI's on the base...... found a gem there though, a STI about us instrument mechs and our shoes, and that we were prohibited from polishing them. drove the RSM nuts for about 3 months coming onto parade with dirty unpolished shoes, and pointing him to the relevant order, which he had actually signed for. My solution for when they got too scruffy was the one item we had plenty of, matt black spray cans......... Ever resprayed an entire aircraft using cans of olive drab and desert brown? Did that after going over the outside with a box of scotchbrite pads. paint shop was busy with another, and we needed the chopper fast with a fresh paint job.

[jesuscf]

Not long ago I wrote this little program to generate the NIST table for the k type thermocouple.  It may be useful for somebody...

Code: [Select]

/* This program generates a table similar to the one at   http://srdata.nist.gov/its90/download/type_k.tab */
 
#include <stdio.h>
#include <stdlib.h>
#include <math.h>

// if( t < 0ºC)
//     E = coeff[0][0]+coeff[1][0]*t+coeff[2][0]*t^2+...+coeff[10][0]*t^10
// else
//     E = coeff[0][0]+coeff[1][0]*t+coeff[2][0]*t^2+...+coeff[9][0]*t^9+a0*exp(a1*(t - a2)*(t - a2)).
//
// coefficients for E=f(t), E is in mV and t is in oC.

const float coeff[11][2] =
{
{  0.000000000000E+00, -0.176004136860E-01 },
{  0.394501280250E-01,  0.389212049750E-01 },
{  0.236223735980E-04,  0.185587700320E-04 },
{ -0.328589067840E-06, -0.994575928740E-07 },
{ -0.499048287770E-08,  0.318409457190E-09 },
{ -0.675090591730E-10, -0.560728448890E-12 },
{ -0.574103274280E-12,  0.560750590590E-15 },
{ -0.310888728940E-14, -0.320207200030E-18 },
{ -0.104516093650E-16,  0.971511471520E-22 },
{ -0.198892668780E-19, -0.121047212750E-25 },
{ -0.163226974860E-22,  0.000000000000E+00 }
};

// Coefficients for exponential portion of equation above
const float a[] = { 0.1185976E+00, -0.1183432E-03, 0.1269686E+03 };

// Ranges for coefficients above
const float range[] = { -270.000 ,  0.000, 1372.00 };

float C_to_mv (float t)
{
int k;
float mv=0.0;

if(t<range[1])
{
for(k=10; k>0; k--) mv=(mv+coeff[k][0])*t;
// mv+=coeff[0][0]; // Not needed because coeff[0][0]=0.0
}
else
{
for(k=9; k>0; k--) mv=(mv+coeff[k][1])*t;
mv+=coeff[0][1]+a[0]*exp(a[1]*(t-a[2])*(t-a[2]));
}
return mv;
}

void Create_Table (void)
{
float t;
char i;

printf("\n oC      0     -1     -2     -3     -4     -5     -6     -7     -8     -9    -10\n");
printf("\n%4.0f %6.3f", range[0], C_to_mv(range[0]));

for(t=range[0]+10.0; t<=range[1]; t+=10.0)
{
printf("\n%4.0f", t);
for(i=0; i>=-10; i--)
{
printf(" %6.3f", C_to_mv(t+i));
}
}

printf("\n");
printf("\n oC      0      1      2      3      4      5      6      7      8      9     10\n");
for(t=range[1]; t<=range[2]; t+=10.0)
{
printf("\n%4.0f", t);
for(i=0; i<=10; i++)
{
if((t+i)<(range[2]+0.1))
printf(" %6.3f", C_to_mv(t+i));
else
return;
}
}
}

void main (void)
{
Create_Table();
}

[xDR1TeK]

Holy Moly!!

[Orpheus]

Dave, I first saw this tutorial when just after you announced it on Twitter, but you already seemed annoyed with the YouTYube critics, so I didn't post anything here or there, knowing you'd already heard from others, and clearly knew enough to see their logic.

But I just checked, and you still have that copper-only loop demo of the Seebeck effect. Now since I KNOW you would never want to do anything that could be used to support the audiophile "directional copper" fallacy, and you did reemphasize a bit later that a thermocouple could be made from a single copper, I feel obligated to raise my voice.

When you apply the soldering iron to the midpoint of the copper loop, it does experience a Seebeck potential, but that is absolutely NOT what you are measuring on the meter. The isothermal plate assures that the two Seebeck reference points (closely linked to the sockets A and B) are at the same temperature. The meter circuit assures that these point are also at the same potential, absent a real external emf. However when you heat the midloop point L, the Seebeck potentials LA and LB are identical, and A =B.

Unless the wire has some preferred direction (for whatever reason), the meter will see no potential difference at the sockets *except* as some effect other than the symmetric Seebeck potential.

You know that there isn't a preferred direction in the wire and the potential drop is exactly the same on the + and - legs, but many Young Players are excessively swayed by the polarized labels on the meter (or in any schematic) and instinctively assume a "right" direction flow. Surely you remember some problem set in Basic Circuit Theory where a classmate made this assumption or railed against the "impossibility" of the problem due to assuming that a labeled polarity dictated a direction of current flow. (e.g. I've often had to explain that the ideal voltage source shouldn't be interpreted as a battery or such, but as a fixed [guaranteed] voltage drop or potential, and further explained that they'd see real examples of fixed [non-proportional, non-ohmic] voltage drops soon enough.)

I hope you don't feel I'm being a pain. I'm just trying to help knock out a minor ding. Maybe you just meant a few of those sentences to mean something other than what I and others interpreted them to mean, and you could do a quick audio patch.

[reagle]

Here are a few pictures from a cheap Chinese Thermocouple thermometer.
They do read local temp for cold junction compensation, but I don't see much thermal mass near the connectors.

[AlfBaz]

I was under the impression that you are compensating with the temperature of the "cold" junction not the ambient temperature. The mass of the pcb, solder connection and terminal block may create a temperature difference to the placement of that thermistor if you take a reading after moving somewhere where the ambient temperature hasn't had time to "soak" the above mentioned elements.
If this is indeed is the case I'd be tempted to put a blob of heatsink compound to help

[Orpheus]

Both are essential: knowing the temperature at the proximal (near) end of the thermocouple AND having both "near" terminals at the same temperature. Well, perhaps not technically "essential" but it's a cheaper and more accurate set-up, so why not

In order to interpret the thermocouple voltage, you NEED to know the "base" voltage at the ner end of the thermocouple, because the thermocouple voltage only tells you the *difference* between the near and far ends, not the absolute temperature. You also need both near ends to be at the same temperature, or the two legs of the thermocouple will be reading "different differences". Sure, you could measure the temperature at the proximal end of each leg separately and temperature compensate each leg separately, based on its metal, but the parts cost and computation/look-up would be dissuasive -- and you wouldn't necessarily be as accurate, because the standard NIST tables weren't generated with that assumption.

Actually, now that I think about it, if the near ends are at different temperatures, it won't even always be clear what the "exact correct" reading should be: the two "near" temperatures would be measurements, and we all know that measurements almost inevitably disagree a few digits out. Since the thermocouple legs generate Seebeck voltages in the microvolt or millivolt range, the independent variation/uncertainty in the two temperature sensors *plus* the various contact/thermocouple uncertainties might create errors that can't be calibrated away anywhere as readily as a "one temperature" system. A man with two watches, and all that...

[jaqkar]

Hi, I am new here

Just watched the video and some things are not clear to me yet. Hopefully someone can help with clarification.

1. The nist tables work on assumed reference junction temp of 0, if so how do you compensate for ambient temp. Substract temp from final reading or do you take the diff in mv from ambient and measured?

2. If you use for example type J cable which I understand is Iron/Constantan to a point and from there you use a copper connector and copper wires to a temperature controller it will mess up the actual readings because of the 3rd different metal you introduce?


 

[retrolefty]

Hi, I am new here

Just watched the video and some things are not clear to me yet. Hopefully someone can help with clarification.

1. The nist tables work on assumed reference junction temp of 0, if so how do you compensate for ambient temp. Substract temp from final reading or do you take the diff in mv from ambient and measured?

2. If you use for example type J cable which I understand is Iron/Constantan to a point and from there you use a copper connector and copper wires to a temperature controller it will mess up the actual readings because of the 3rd different metal you introduce?

 Yes, it will mess the reading up. Every transition from TC wires to copper wires forms 'new' junctions that add or subtract from the actual measurement. That is why TCs require TC lead wiring (including special screw terminals made from the same metal material as the TC wires use) from the TC sensor tip to the first TC to copper transition which is where the TC compensation sensor measurement must be taken.

 I worked for 30 years with TCs at an oil refinery, we had thousands of them. There main advantage is the very large temperature range that an TC can handle without 'burning open'. They are not particularly accurate outside of lab conditions using expensive 'certified' TC material.

 We started converting over to 3 or 4 wire platinum 100 ohm RTD sensors the last couple of decades for new installations, as they had come down in price over the years relative to TCs and are much more accurate in real word installations. However they could only be used up to about 500 F without using extra rugged and expensive ones that could push up to about 1,000 F. RTD is a much better choice these days in my opinion expect if you have to go up beyond 1,000 F.