Comparing an IR camera to a thermocouple

@jed,

The camera resolution is pretty much the problem. This can be compensated for by getting closer to the object, within a safe temperature distance for the camera. However, this creates problems with the many cells required to acurately calculate radiant output if there are many different temperatures in the camera view simultaneously. Backing off and getting the camera to average over a larger area is probably simpler and probably nearly as effective as attempting to manually calculate a composite average over many small areas, such as individual valleys and ribs (and what about the rib profiles? At what point should one cease dividing the measurement areas?)

This topic came up in a discussion of the Lugano experiment. I do not recall where. Anyway --

In the Lugano experiment, I suggested that it would have been a good idea to confirm the IR camera temperature readings by comparing them to a thermocouple (TC) held to the surface of the cell. This was what Levi et al. did in the previous experiment. In the previous experiments, the IR camera agreed with the TC to within 2°C.

There are some possible issues with using a thermocouple to measure the temperature along the ridges in an Lugano like experiment.

First of all at the bottom of the fins on the tube there is more thermal mass then at the top of the fins.

Since the metal thermocoule wires are good thermal conductors, going to an environment with a much lower temperature, they will drain heat away at the tip of the thermocouple.

This will lower the temperature of the material measured and that of the tip of the thermocouple, especially if due to a lower thermal conductivity and thermall mass of the material measured, the lost heat can not be supllied fast enough.

Since there is less thermall mass at the top of the fin then at the bottom, the temperature due to the thermal drain of the thermal couple will be less at the bottom then at the top.

The effect will cause lower temperatures to be measured then the real temperatures and lower at the top then at the bottom.

Note that in the previous experiment you referred to, the thermall mass is much higher with no ridges and thus that situation is much more suited for calibrating with a thermocouple.

Secondly there is the effect of a changed emissivity due to the ridges.

As known, a small deep hole can be considered as a black body, independent of the material surrounding te hole. While the ridges are not holes, they create valeys which makes the material more gray then the flat material. This causes the emissivity to increase resulting in lower measured temperatures by the thermal camera. I have seen in the past one manufacturer of thermal camera's which in a document warned for this potentional measurement error.

The amount of increase of the emissivity can be calulated with the infinite reflection method if the view factors are known.

The change in measured temperature requires for an accurate temperature measurement calibration but as stated it is difficult to measure accurately with thermocouples due to the thermal drain.

Third the thermocuple measurement itself. Many people have no idea of the possible measurement errors when using thermocuples.

First of all, the temperature versus voltage curves for a type of thermocouple changes from thermocouple to thermocouple, the error often being a few percent. You can buy calibrated thermocouples which are supplied with a callibration curve to allivate this problem.

Then the thermocuple wires are connected at a certain point to normal wires, most of the time copper. That point is called the "cold junction" because it is normally at a place with a much lower temperature then the thermocuple tip.

The thermocuple measurement is dependent on the temperature of the cold junction, so you have to compensate for this effect and you can only compensate if the temperature of the cold junction is accurately known which requires another calibrated temperature measurement.

Then besides the cold junction, due to material dissimilarities in the electronics and wiring , we can have contact potentials resulting in offset voltages which are causing additional measurement errors.

And then we have the amplifier part where the thermocuple signal is amplified. This amplifier has also offset voltage errors and gain errors resulting in measurement errors.

Then we have also possible electromagnetic interference on the thermocuple signal, most of the time picking up 50/60 Hz magnetic fields which causes an ac voltage superimposed on the thermocuple voltage, again resulting in possible measurment errors if they are not filtered out.

The above problems can be largely deminished if adequate electronics (more complex and thus more expensive) are used.

For temperature measurements as in Lugano, where you can live with some limited inaccuracy in the temperature measurement I agree with Para that the averaging function of the Optris camera will give adequate results. However since also the in band emissivity to be set on the camera changes with temperature you will need to calibrate for this and then we have also to compensate for the increase in emissivity due to the grayer area caused by the fins. To allivate some of these problems I would suggest instead of using a thermocuple to measure with a dual or multiband pyrometer which automatically compensates for the emissivity (But averages the temperature also over the area as the Optris does) . That was indeed what the MFMP did by using a dual band Williamson pyrometer in addition to their Optris during their investigation of the Lugano Hotcat reactor.

I doubt that the in band emissivity required for the IR camera changes significantly with temperature. The broad band (total) emissivity required to calculate radiant power may change a small amount between a valley and ridge, but probably not very significantly either, at any given simultaneous valley and rib temperature.

The camera can select a very small measurement area, and this can be matched to a thermocouple position, assuming that the thermocouple does not conduct a significant amount of heat from the attachment location, and the thermocouple (which will have a different emissivity) is not direct imaged.

In general, thermocouples can indicate whether the IR camera is reporting a temperature that is reasonable for a specific area, but exact matching of temperature requires considerable thermal homogeneity of the surfaces being compared. A thermocouple only measures a very tiny area, and is quite limited by that fact. Several thermocouples could be attached, and the IR camera is roughly equivalent, when calibrated effectively, to hundreds of thermocouples being operated at one time.

Comparing IR to a thermocouple is like comparing the taste of two types of apples to each other. Not apples to oranges, but more like how two independent tastes of the same bite of the same apple are also not possible.

Additionally, two band pyrometers are not as effective as one might think when dealing with objects that have complex selective emissivity profiles. The pyrometers are attempting to determine a single temperature compatible with two different and simultaneously measured IR band radiant powers. It estimates an emissivity slope from that information to calculate a greybody equivalent. If the broadband selective emissivity profile of an object is complex, then the slope may be estimated poorly. 3 (or more) bands will improve the estimate. A greybody, which in general includes most materials (to some level of effective detail) can be very accurately tested, however.

I doubt that the in band emissivity required for the IR camera changes significantly with temperature.

Using emissivity curves for alumina taken at different temperatures I estimated once that the in band emissivity changes between about .86 at 200 degree C to .95 at 1100 degree C.

If you had used the .95 throughout then the error at 200 degree C would have been about 3%

I don't know if you consider this being significantly or not.

Wonder if you ever tried to determine the in band emissivity for different temperatures with your casts in order to arrive at the temperatures measured with the thermocouple.

Maybe that gives a better estimate/indication then my calculation.

Put the tube inside a second larger tube with sand and put the thermocouples in the sand.

LDM,

Did you integrate the radiant power for the in band emissivity at various temperatures, and then determine a suitable in band emissivity, or just average the in band emissivity values?

From 0.86 to 0.95 over a very wide temperature range does not seem out of line, however.

Is that 3% error in temperature or in output power?

With the Durapot, the apparent in band emissivity moved about 0.02 - 0.03 from 20 C to (I think) about 1200 C. I would have to review my notes to confirm that. I believe that I posted the results a while ago, immediately after testing it. There was a fairly sudden change, rather than gradual, at around 750 - 800 C, detectable both when increasing and decreasing the temperature.

My point was more that the local temperature differences at any given time, for example between the ridges and valleys, would not have a large deviation in the in band (or even total) emissivity. It is not likely that immediately adjacent parts of a similar structure would have a large enough temperature gradient to require independent tuning of the local emissivity. Between the Caps and Main Tube, (obviously different structures), there could be a minor correction required.

By your estimate, how much deviation in the in band emissivity is there from 450 C to 900 C?

Paradigmnoia

Did you integrate the radiant power for the in band emissivity at various temperatures, and then determine a suitable in band emissivity, or just average the in band emissivity values?

I did it by numerical integrating over the in band frequency range the black body spectrum times the alumina emissivity devided by the black body spectrum.

From 0.86 to 0.95 over a very wide temperature range does not seem out of line, however.

Is that 3% error in temperature or in output power?

That is the change in temperature using the formula that the measured temperature changes by the power 1/3 for the Optris. (.86/.95)^1/3 = .967 or about 3% difference.

The same formula was used by the MFMP, however at lower temperatures it deviates.

With the Durapot, the apparent in band emissivity moved about 0.02 - 0.03 from 20 C to (I think) about 1200 C. I would have to review my notes to confirm that. I believe that I posted the results a while ago, immediately after testing it. There was a fairly sudden change, rather than gradual, at around 750 - 800 C, detectable both when increasing and decreasing the temperature.

From your post of nov 6th 2017:

I have only just done the emissivity test of Durapot 810, and it is only applicable to the 8 to 14 micron band. It was only one test, with one pyrometer, but over 10 temperature data points. It very slowly dropped from 0.9 to 0.87, in that band, from 300 C increasing to the maximum external temperature tested of 990 C. (0.88 at 500 C)

My point was more that the local temperature differences at any given time, for example between the ridges and valleys, would not have a large deviation in the in band (or even total) emissivity. It is not likely that immediately adjacent parts of a similar structure would have a large enough temperature gradient to require independent tuning of the local emissivity. Between the Caps and Main Tube, (obviously different structures), there could be a minor correction required.

I think that the correction would indeed only be minor.

I also still think that the measured difference between valley and top when using the Optris for the temperature measurement is influenced by the difference in view factor between the valey and top, resulting in different emissivity correction for the valey and top.

I am thinking about investigating these differences since with view3D I should be able to calulate the view factor for the valey and the top.

By your estimate, how much deviation in the in band emissivity is there from 450 C to 900 C?

.90 at 450 degree C

.94 at 900 degree C

LDM,

I don't think that the view factor has much effect for the camera, as long as the resolution is is not such that individual valleys and ribs can be clearly imaged. (And even then I wonder.) Otherwise the Lugano reactor images of the main tube area would appear hotter in the middle (normal to the camera lens, valleys visible to the lens) and cooler towards the caps (oblique to the lens, valleys obscured by ribs).

Paradigmnoia

I don't think that the view factor has much effect for the camera, as long as the resolution is is not such that individual valleys and ribs can be clearly imaged. (And even then I wonder.)

As I stated I have seen a document of one thermal camera manufacturer in the past explicitly warning for the problem that the change in emissivity by ridges causes an additional error on the measured camera temperature. So while the camera will average the reading over the ridges, this reading will most likely be somewhat off.

Otherwise the Lugano reactor images of the main tube area would appear hotter in the middle (normal to the camera lens, valleys visible to the lens) and cooler towards the caps (oblique to the lens, valleys obscured by ribs)

What the effect is of not seeing the valleys to the end depends also on how large the temperature difference is between the tops and the valeys.

What I want to try if I can find time for it is to make with view3D an estimate of the view factor at the bottom and the top and then calulate the emissivity corrections.

That will give us an indication how much of the temperature difference between top and valley is due to the differences in emissivity.

It might be interesting to know this since it was stated somewhere in the past that the temperature difference between the top and valley could be near 100 degree C at higher temperatures.

If that is true then towards the caps where as you say the valleys are (partly) obscured, the temperature seen near the caps by the thermal camera should drop significantly.

However if the thermal camera by your statement is not seeing this temperature drop, then it would mean that the temperature differences between tops and valleys are much lower then that 100 degree C.

So I want to explore if some of the larger reported temperature differences can be explained by the differences in emissivities between top and bottom.

At least it will maybe tell us if it is something we need to take into account or not.

How does one effectively measure the temperature of a fin at the scale of those in the Lugano device with a thermocouple? I have attempted it, but have never assured myself that the measurement was very accurate. The best that I have managed is the spring pressure of the thermocouple wire against a fin. At very high heat, that too does not last long.

Anything affixing the thermocouple to a fin tip externally affects the measurement. I may be able to cast a thermocouple into the near-fin tip, but the fin would be weakened considerably, the closer to the tip the thermocouple is cast. I have managed to cast a thermocouple into a flat surface, with the thermocouple wire a couple of cm from the junction dipping deeper into the Durapot a bit to keep it from pulling out easily.

Wat about casting a round tube with a heater coil embedded, say for example 10 cm long and then cast only one rib in the middle.

You can then measure easily with a thermocuple the temperature at the base of the fin by mounting a thermocouple on the tube next to the fin.

Since the thermal mass seen by the thermocuple is large, the deviation in temperature measurement will minor.

(You can also run the thermocuple wires for a short distance over the tube to reduce thermal drain from the tip)

The fin itself can be measured by a thermal camera pointing perpendicular to the surface of the fin if the thermal camera has enough resolution to resolve temperatures between bottom and top of the fin.

Since there are no opposing fins there will be almost no influence of view factors on the emissivity and thus the measured temperatures will be representative for the temperature distribution over the fin.

One rib on a cylinder is not at all similar to many parallel ribs. I suspect that at a minimum 5 ribs would be required, with measurements of the center one, to deal with the multiple absorption and emission effects, and insulating vs conducting effects of the base of the ribs. However after casting and testing with several ribs, the surrounding ribs could be ground off to in order test one by itself easily enough.

Fun discussion of course, but: Does it really matter how output power is measured in Rossi's tests? What I mean is that what matters is how the measuring method is verified.

Rossi's "reactors" all have electrical heaters. All that is required is to employ these heaters to calibrate and verify whatever method is used for power measurement. The really important parts are 1) to keep Rossi and his prestidigitating hands far away from the entire experiment. 2) To keep anyone closely allied with Rossi away from it. 3) To use accurate methods to determine the input power to the heater including completely isolating the power source and wires for the heater so Rossi has nothing to do with furnishing them. 4) Run that calibration over the full operating temperature range of the reactor. 5)Optional: use more than one method for measuring the thermal power and energy out.

This is basically what happened when Industrial Heat accidentally tested an unfueled reactor and got the same result as with a fueled one (nothing), suggesting that there is nothing more to Rossi's reactors than the electrical heater with the liberal addition of misdirection and bovine intestinal effluvia.

The fatal flaw in Rossi tests has always been, from the very beginning, a failure to isolate the reactor's input (heater) power, provide one's own measuring equipment and heater power source, and keeping Rossi and friends away from the experiment for the entire duration. Testing by Industrial Heat solved the problem though with a different result than what they expected!

My humble opinion of course. YMMV.

Also, Rossi has often claimed "self-running" reactors which do not require electrical heating to function after their initial start. Anyone ever seen one? Why did he not offer such a reactor for testing, ?

Perhaps we have been going about this quite backwards, although it is an informative and intellectually stimulating exercise to work out the thermodynamic minutiae of the Lugano device.

Regardless of the inexact replication of the MFMP Lugano simulacrum, it is indeed made of alumina, and therefore the total emissivity of alumina should apply grossly if not closely to the device. Using as closely as applicable the output calculation methods used in the Lugano report, whether these are entirely accurate or acceptable or not, and applying the (silly) re-iterated emissivity procedure as described in the Lugano report to the MFMP .rav files, using the equivalent measurement boxes for the Optris software, the COP can be calculated. If it is quite close to the Lugano COP at the same re-iterated emissivity-temperatures, then either the MFMP have demonstrated a > 3 COP device built without a secret recipe, open sourced to all, or the Lugano demonstration should be tossed into the dustbin of history. And attempts to scavenge a COP of 1.1 or similar from the Lugano report should be equally applied to the MFMP data. If that acts similarly, then the fraction above COP 1 is most likely an error. Or mundane objects do produce COP >1, counter to a few hundred years of experiments.

Quote from JedRothwell

Yes. My purpose in asking was not to rehash the Lugano debates. It was to ask how one can go about confirming an IR camera with a TC. That seems like a useful technique in some future experiment. Based on the answers here, I gather it might be a challenge because of IR camera resolution limitations, and because physically holding the TC against the cell at high temperatures might be difficult. This I did not know.

Like most things, from software and operating systems to choice of fuels to heat something, spectral thermography has advantages and disadvantages compared to other ways to do the same task.

A thermocouple may be more suitable for checking the temperature of a computer chip or a room, but spectral thermography can check the temperature of a star, or detect leaks in the insulation envelope of a building more easily than a thermocouple.

A certain level of accuracy and precision might have to be sacrificed depending on the situation and the level of sophistication required to do a given job. Sometimes the level of detail can exceed the requirement.

The important point to consider is, "Does this method tell us reliably what we need to know? Is the level of detail sufficient for the purpose? Are we using the method with best practices, and mindfully?" If the answers to any of these questions is "no", then the 'information' provided can range from anecdotal evidence at best, to garbage-in, garbage-out uselessness, and could extend to a source of misdirection (intentional or otherwise) at the worst end of the scale of credibility.

Influence of view factor on measuring the temperature of the top and bottom of a Lugano style fin.

In an earlier post we showed the effect of the view factor between fins on the thermal calculation. One of the factors is that the apparent emissivity changes by a factor

1/(1 - F_ff(1-ε))

The correction is dependent on both F_ff, the view factor between the fins and e the emissivity used, The recommended setting for e for the Optris being .95

The view factors can be calculated with the NIST program view3D by using an area around the circumfence of the fin with a small height compared to the height of the fin.

In our calculation we used a height of .05 mm for a circumfence at the top and the bottom.

The results of the calculation for the top and bottom of the fin are :

View factors (fin to fin)

----Top----------0.007

----Bottom-----0.747

Using the emissivity correction formula and the found view factors, the emissivity correction factors are :

Emissivity correction factors

----Top----------1.0035

----Bottom-----1.0388

And the corrected emissivities become :

----Top----------0.950

----Bottom-----0.987

Since the corrected emissivity at the top is the same as the used emissivity, the temperature shown by the Optris for the top will be correct.

For the bottom of the fin the higher emissivity means that the in band radiated power per area is a factor .987/.95 = 1.039 higher.

Since the temperature of the Optris is by approximation proportional to the power 1/3 of the power density it means that the Optris will show the temperature higher by a factor 1.039^(1/3) = 1.013. (temperature in degree Kelvin)

If we take as an example a temperature of 750 degree C or 1023.15 degree K, then the corrected temperature becomes 1023.15 * 1.013 = 1036.264 degree K or 763 degree C.

Thus the temperature shown by the Optris for the bottom of the fin will be 13 degree C higher then the real temperature (assuming in in band emissivity of .95 was set on the Optris). The total temperature difference between top and bottom of the fin as measure by the Optris now becomes the real difference plus the calculated correction.

One of the reasons I got on this forum was to ask a question about the use of the Optris for temperature measurement. I have another I'd like to ask. When one computes the power from a blackbody point source, the triple integtral 'over all space' reduces down to the extremely simple sigma*T^4. But when you use a real body made of alumina, the blackbody equation has problems since the alumina emissivity is non-Plankian. I believe one could express the emissivity curve as a sum of a blackbody at some temperature plus the deviations from the blackbody equation. Then the triple integral can be separated into the blackbody part and the deviation. The deviation term represents an 'error', or more correctly, an error correction term. Has anybody evaluated that term's magnitude for the Lugano experiment? Also, would the fact that the Optris is not looking at a point source matter, or is that taken into account by the software?