Science, Technology and Society - An authentic exploration on IR thermometers application in schools

During the SARS crisis in 2003, we joined the onrushing crowd purchasing a few sets of infrared thermometer for measurement of students' body temperature before they enter the school. It was thought that such "hi-tech" device could provide us fast and convenient way of screening students from those having fever, a symptom of SARS. However, after a fairly long period of practice, it was found that readings on display of such devices varied in a range wide enough to make us unsure whether a student is having fever or not. Despite the calibration, the IR non-contact devices seemed not offering out consistent readings about a person's forehead temperature, sometimes ranging from 27°C to 33°C, which was so confusing. Therefore we decided to take a deep look into the matter to see what was happening. Below are our findings that we'd like to share with all those encountering similar situations.

At first, we suspected that the device was not well calibrated or mal-functioning. Hence we shipped them back to the factory requesting a thorough check as well as certified calibrations. We were confirmed that the devices were functioning as designed and calibrated as specified. Mal-functioning is not a reason to the fluctuations of readings. We could reject malfunctioning as a reason if we could prove that temperature displayed on the IR thermometer followed a straight line with rise and fall in temperature range between 22 to 55 degree Celcius. The same applied to our Data-Logger. Then we could countercheck temperature readings reciprocally using both devices. Hence we started our explorations for the whys and wherefores.

First of all, adopting the temperature sensors of the Data-Logger set, we decided to repeat the calibration on our own. The purpose of doing this is to make sure the device is accurate to an extent enough for our aim or, if not, how far it was from behaving as what was designed to feature. We used a Data-Logger temperature sensor to countercheck the linear response of the IR thermometer. To facilitate the calibration process, a wooden frame was constructed to hold the device under test firm in position and mounted on a dynamic trolley available in school Physics Laboratory as shown on the left. Markers were also adhered to the trolley as pointers indicating distances from the object being measured. The wooden frame was so mounted that the laser pointer from the device was pointing approximately the middle region of the object being measured. Since shinny surfaces exhibit low emmissivity of infrared wave, an iron mass of about 3 Kg with dark and rough surface was used as the object to be measured. To avoid fast heat loss while taking readings, such a heavy object was believed to possess adequate heat capacity.

Holes were drilled on the 3 Kg iron mass. The first one was drilled in the middle of the top face deep enough to reach the center of the mass inside for the first sensor to measure the inner temperature of the object. The second one was drilled just beneath the front surface for the second sensor to monitor the surface temperature which should most possibly match with that on the infrared thermometer display as expected.

In the first part, the mass was then heated up with a Bunsen flame to approximately 90°C at the beginning. It was then allowed to cool down gradually to room temperature while taking readings both from data logger and infrared thermometer simultaneously. See Fig. 1 below.

In the second part, the mass was then placed inside the freezer of a fridge to cool down to about -15°C at the beginning. It was then allowed to warm up gradually to room temperature while taking readings both from Data-Logger and infrared thermometer simultaneously. See Fig. 2 below.

Then 2 curves of the infrared readings against the inner temperature readings and the surface temperature readings were plotted respectively as indicated below in Fig. 3. It can be seen that the 2 curves were very close to straight lines which implied linearity characteristic of the infrared thermometer. In both Fig. 1 and Fig. 2, the blue line stands for infrared thermometer reading. The red line stands for the surface temperature detected by Data Logger. The green line stands for the inner temperature detected by the second sensor of Data Logger. Judging from Fig. 1 and Fig. 2 the infrared detector exhibited a linear responding characteristic as that of the temperature sensor monitoring the surface of the iron mass except with a constant difference that seemed to be a calibration problem. Looking from the slopes of the curves in Fig. 3, we found that the Data-Logger sensor and IR thermometer were both acceptable for the purpose of our investigations despite exhibiting different linear thermal properties.

After confirmation of the response characteristic of the infrared thermometer, we started exploring suspected possible causes of the fluctuations.

(1) Battery voltage drop after current consumption has reached a certain level
(2) Low quality and/or poor stability of the device
(3) Ambient temperature of the environment where the device is being used
(4) Skill of the operators holding the device
(5) Specifications requirement of the device not matched when being used

Since the device is powered by 009P type 9V battery with a current capacity of about 400mAH only, the first and most suspicious reason was voltage drop due to long time usage. To begin with, we had prepared two pieces of 009P batteries beforehand, one discharged with a current of about 200mA by a light bulb for approximately one and a half hour to about 6V and the other one brand new. Then the infrared thermometer was first fitted with a brand new battery and mounted on the wooden frame. Pointing at a 3Kg iron mass with dark rough surface preheated to about 40°C at a distance of about 30cm, the temperature reading was recorded. Then we changed the new battery with the exhausted one. Temperature reading was taken again at the same distance. In order to be sure, the process was repeated 5 times with data recorded as below.

From the table above, it was concluded that battery voltage had no significant influence on the readings of temperature detected. Therefore, battery voltage drop was NOT a cause of the fluctuation.

Quality and/or stability of the device could affect the output result quite dramatically. However, it was not difficult to check this out. The 3Kg iron mass was preheated to about 40°C and allowed to cool gradually. The infrared thermometer was then pointed at the center of the iron mass at a distance of 30 cm and reading taken. The process was repeated 5 times at intervals of 15 seconds. Result was recorded as below.

From the table above, it was clearly shown that readings were in consistency after 5 trials within a minute of time. Therefore, stability of the device was quite satisfactory.

Another suspicious cause of inaccuracy or fluctuations was ambient temperature. Since the environment where we take body temperatures of students were at the main gate of the school and it was occasionally windy. Thus, some of our staff suspected that cold air blowing over the forehead could possibly affect very much the temperature detected. Hence we set up the experiment below to take a closer look into this.

As consistent standard, the 3 Kg mass was preheated to about 40°C and allowed to cool gradually. But this time, we mounted a computer type mini cooling fan in front of the heated mass as shown such that cool air could be blown over the surface of the heated mass when necessary. Then readings were taken under 4 conditions as follows :

The time intervals between each successive reading taken were 30 seconds approximately. It was noticed that cool air either blowing from left to right or right to left over the heated surface of the iron mass did not bring much difference to the infra-red detection reading. The temperature drop of about 1 degree after 2 minutes is more likely being caused by natural cooling of the mass itself.

It became obvious that cool air blowing over the surface of the object being detected did not cause significant effect in its emission of infrared energy. Therefore, ambient temperature variation was NOT a cause of the fluctuation either.

For further confirmation that ambient temperature was not a factor to cause inaccuracy, a Bunsen flame instead of the computer cooling fan was placed in three different positions nearby the 3 Kg mass as that indicated below. As expected, the Bunsen flame did not cause any significant fluctuations in the IR readings. Temperatures recorded in these 3 cases were approximately 36°C only.

A series of tests had been used to investigate the fluctuation of the device, but still could not find out what the main reason was. It was quite disturbing to us. We began to investigate human factors, to simulate how users may use the device to measure the body temperatures of students by considering their attitude (focused vs. casual) in griping the device. Since there was a built-in laser pointer on the device, three kinds of aiming method were tried to see whether there were any new findings or not.

(1) Measurement with assistance of Laser Beam Aiming
(2) Measurement with assistance of Naked Eye Aiming in concentrated way
(3) Measurement WITHOUT any aiming assistance in approximate way

This time, the IR thermometer was held in hand as usual and measurements were taken from 5 different directions.

Distance from the mass to the device was not fixed during measurements so as to simulate the actual situations. Results were recorded as below.

From the table above, variations of readings were found in the test. It seemed that we're on the track of discovering something related. Hence, we carried on to the up-tilt/down-tilt test to see what else further could be found. In this session, the aiming laser beam was directed from the same origin to 5 different spots (Center, Upper edge, Lower edge, Left edge and Right edge) as indicated below.

The angle up- or down-tilt measured was as small as 9° only. Such a small deviation of angle in holding gesture did really cause a significant variation in the readout of the device.

From the table above, it was found that holding gestures of the users was the main reason in causing erratic fluctuations of readings to the device. We therefore considered that maybe it was the specification requirements of the device were not matched when being used. After perusal of the operation manual of the IR thermometer, several points were noted.

According to the operation manual, the device being used should follow the “Distance to Spot Ratio” requirement. The detection area must be larger than the required spot size to get correct readings. See ”Field of View” below.

As the reading shown on display was the AVERAGE temperature of the detected region, errors can easily occur if the object to be measured cannot fill up the Field of View of the detection spot. For example, if the diameter of the spot with consistent body temperature on human forehead was only about 40mm, the device should thus be placed at about 320mm away from the detected region all the time.

The device displaying higher values of readings means the more capability of radiation emittance the materials have. Emissivity means the energy-emitting characteristics of materials. This was also a factor of causing inaccuracy.

The information below shows different devices with different characteristics. Fig. 10 shows the specification of other brand's infra red thermometer. Laser sighting e.g. single spot, 8-point circle, 17-point circle with or without focus. Distance to Spot Ratio e.g. 6:1, 8:1, 12:1, 50:1 etc. Emissivity can be adjustable or fixed. Maximum temperature they can reach are different. All these criteria influence the accuracy of the readings we take. Fig. 11 shows the distribution of emissivity of a human body. The white lines on the human forehead clearly separate the temperatures into regions. The regions change from time to time according to the blood flow through the human forehead. Fig. 12 is a Demonstration of characteristics of average temperature taken within a spot. Fig. 12A~12E shows a clear picture of differences when the spot fall on different areas of the detected region. The white circle is the detected area formed according to the distance to spot ratio.

From the series of pictures 12A to 12E above, we clearly see that if the spot doesn't fall on and cover up the right region to be measured, erratic readings are to be expected due to detection of IR energy from elsewhere other than the region we want.

Since the acceptable range for deciding whether a person is having fever or not is only within 1°C to 2°C, and the requirement of operators to maintain a steady holding gesture is so difficult in every measurement. Either judging from the up-tilt/down-tilt test or the test in Fig. 12A to 12E, the errors were so huge that the device is NOT quite a suitable machine either in defining whether a person is having fever or to screen out students from having fever or not.

We thought all the tests were done with infrared thermometer of Brand A, again we met new challenge. We were told, Brand B (RayTek) must be more accurate because the readings taken were around 33°C ~ 35°C. a range connected to the known temperature measurement of human forehead; and it is three times more expensive. Brand A was said to be less accurate because it gave readings in the range 28°C ~ 30°C, away from the known range. Sounds reasonable, but now we know we can be cheated by our own pre-conception. Since we knew that erratic readouts from these types of IR thermometers can be so vast, we learned that judging the accuracy of the instrument simply by quoting that the readings are near to normal body temperature of human beings is not scientific at all. Judging the accuracy of the instrument could be counter-intuitive.

The next step was to conduct some more tests to see which one will be more accurate. Here are our results below.

Two different brands of devices were used to detect a high temperature (the 3 Kg iron mass heated to about 40-50°C ) and a low temperature ( a black painted beaker of melting ice) for comparison. In both cases, a common laboratory type of alcohol thermometer was used as reference. Fig.13 and 14 represents the device said to be inaccurate while Fig.15 and 16 represents the device claimed to be more accurate when measuring the body temperature of human forehead.

Strange !! Taking a closer look into the photos, device in Fig.13 and 14 was found more accurate than that of Fig.15 and 16.

It was really an interesting finding. Here, we may say that simple conclusions originated from erratic data could possibly lead to quite an opposite finding about a phenomenon of science. Our minds were bogged with established norm values. Many "concepts" have been planted inside our brain so firmly since we were born and brought up. These "concepts" sometimes reduce our curiosity in many aspects, mitigate our sensitivity of nuances, and make us neglect search for solid evidence. We knew from the calibration curve we presented earlier that the IR non-contact thermometer in our tests carried a calibration problem. However, it is still an accurate device on condition that we follow its requirement of distance to spot ratio as well as confirm the spot covers the region to be measured.

However, for measurement of human body temperatures, it seems not an appropriate one. Even with one offering a closer range of reading at the desired zone, the advertisement of local dealers could be very different from the opinion of the technical experts from the manufacturer.

Let us share the official reply from one of the infrared thermometer manufacturer.

	Trial #1	Trial #2	Trial #3	Trial #4	Trial #5
New Battery	39.5°C	39.0°C	38.5°C	38.0°C	39.5°C
Exhausted Battery	39.5°C	40.0°C	39.5°C	39.5°C	39.0°C

	Trial #1	Trial #2	Trial #3	Trial #4	Trial #5
Time	0 second	15 seconds	30 seconds	45 seconds	60 seconds
Temp Reading	38.5°C	38.0°C	38.5°C	38.5°C	38.0°C

	Situations		Temp
(1)	The infrared thermometer was pointed at the pre-heated mass with the fan turned OFF	T1	37.5°C
(2)	The infrared thermometer was pointed at the pre-heated mass with fan turned ON and cool air blown from left to right over the surface	T2	37.5°C
(3)	The infrared thermometer was pointed at the pre-heated mass with fan turned OFF again	T3	37.0°C
(4)	The infrared thermometer was pointed at the pre-heated mass with fan turned ON and cool air blown from right to left over the surface	T4	36.5°C

Aiming Assist	A	B	C	D	E
Laser Beam	35.5°C	34.5°C	34.0°C	34.5°C	34.0°C
Naked Eye Aim	35.0°C	31.0°C	33.5°C	32.0°C	28.5°C
Approx. Aim	34.5°C	24.0°C	33.0°C	33.0°C	19.0°C

	Situations		Temp
(1)	The infrared thermometer was pointed at the Centre of the pre-heated mass with laser beam	T1	32.0°C
(2)	The infrared thermometer was *Up Tilt* with the laser beam located just at the upper edge of the pre-heated mass	T2	34.5°C
(3)	The infrared thermometer was *Down Tilt* with the laser beam located just at the lower edge of the pre-heated mass	T3	16.5°C
(4)	The infrared thermometer was pointed at the *Left* hand side of the pre-heated mass with the laser beam just falling outside the mass	T4	30.0°C
(5)	The infrared thermometer was pointed at the *Right* hand side of the pre-heated mass with the laser beam just falling outside the mass	T5	20.5°C

	Brand A	Brand B
Field of View (Distance to Spot ratio)	8:1	12:1
Emissivity	0.98	0.95
Temperature Range	-20 ~ 500°C	-32 ~ 400°C
Accuracy	±2°C	±1°C(Above 23°C) ±2°C (-18 ~ -23°C) ±2.5°C (-26 ~ -18°C) ±3°C (-32 ~ -26°C
Type of Sighting	Single spot
Operation condition