FIGURE 9. Map of air-temperatures in Fahrenheit degrees of a cross-sectional area during a typical test when ▲t= 40 deg F. Data obtained from an array of butt-welded chromel constantan thermocouples fabricated from No. 36 wire (AWG). Array was immediately upstream from the probe.

velocity of 500 fpm. It was for the array of thermocouples immediately upstream from the probe. The cross-sectional temperatures, it will be noted, were essentially uniform in the central area for the seven thermocouples nearest to the center of the duct. Analyses of comparable data for the other two arrays along the length of the duct during the same test showed a similar and satisfactory pattern. This is illustrated in table 2, showing temperatures for selected corresponding positions in the arrays.

The mixing was satisfactory for all tests. Comparison made of the pattern in figure 9 with two other determinations, made at the same station at 40 deg F At and at 500 fpm, showed that in each case the maximum difference between any of the seven center readings was less than 0.30 deg F. The difference between the average of these 7 readings and the center reading was always within 0.10 deg F. For the tests at 50 deg F at at the same velocity, the maximum difference between any of the seven readings was about 1.0 deg F. The difference between the average and the center was 0.7 deg F. For the tests performed at smaller values of At, the differences were considerably smaller.

The temperature of the inlet air coming into the duct was monitored using a copper-constantan

1234567

78.34

80.59

83.97

78.46

78.64

79. 14

78.37

81.36

83.99

thermocouple in conjunction with the digital voltmeter. There were unavoidable small cycles in the temperature of this inlet air, and to minimize any effect of these cycles, determination of the air pattern was made only at instances in time when that temperature was constant within 0.1 deg F. The mass in the metallic air mixers tended to dampen the cycles and help even more. The readings for the thermocouples in the arrays were taken in rapid succession.

After the tests with the 10-in shields were performed, tests without shielding around the probe but with small shields around the reference sensor were made at 300, 500 and 1300 fpm. The small reference sensor, in its special set of shields (fig. 6), was placed immediately upstream from the probe. The same test conditions as used for the tests with the 10-in shields were used. For the test at 1300 fpm it was not possible to obtain readings for At=50 deg F because of shortage of electrical power in the test facility.

The temperature of the moving air was steady near the axis of the duct; therefore it was possible to determine the readings of the fine-wire, buttwelded, thermocouples in the reference sensors with the manually operated potentiometer.

The results of the studies are shown in figures 10 and 11. Figure 10 includes curves showing the effect of radiation when the three sensors were measuring the temperature of the moving air at 500 and 1300 fpm, and figure 11, curves showing the effect when the sensors were measuring this temperature at 300 fpm. For the unshielded sensors under the conditions of 1300 fpm (6.6 m/s)

air velocity and a ▲t of 50 deg F (28 K), the observed error was about 1 deg F (0.6 K). For 500 fpm (2.5 m/s) and 300 fpm (1.5 m/s), under the same test conditions, the error was about 3.0 deg F (1.65 K) and 3.8 deg F (2.1 K), respectively. When the sensors were shielded, at both 300 and 500 fpm and under the same test conditions, the error was less than 0.2 deg F (0.1 K).

The indicated radiation error was greater for the thermistors than for the other two types of sensors for all but one test condition. However, the indicated radiation errors for the thermocouples and the resistance thermometer change in magnitude relative to each other from one year to the next. The cause of the change is not understood,

but it is probable that from the spacing on figures 10 and 11 between the thermistor curves and the resistance thermometer curves, and from the constancy of their relative positions, that the thermocouples had changed, not the thermistors or the resistance thermometer.

Uncertainties occurring in the experimental work resulted in small errors in the observed results. For the tests where the probe was shielded, the possible error in experimental results was relatively large when compared to the radiation error reported. This uncertainty was principally due to the drift in observed values between calibrations for the thermocouple wire in the reference probes. Table 1, giving the change in performance of the probe sensors and the reference sensors, shows that the performance of the reference thermocouples sometimes changed 0.06 deg F between calibrations, and for two measurement temperatures changed -0.10 and -0.12 deg F, respectively. Another uncertainty was involved in using the aspiration technique on the reference sensors where the measured radiation error was 0.04 deg F at At=0.

It is possible that some of the experimental errors for the shielded probe could have been cumulative and could have totaled as much as 50 percent of the reported radiation error of 0.2 deg F at At =50 deg F. However, in view of the several possible sources of error it is more probable that there were compensating effects and that they

For the tests where the probe was not shielded, the possible error in experimental results was relatively small when compared to the radiation error reported. The same causes for experimental error occurred for these studies as for the studies when the probe was shielded, but the effect of cycling of the temperature of the air stream was more evident. Another contributing factor seemed to be the scatter of readings in the thermocouples on the duct which made it difficult to know the difference (At) between air temperature and duct surface temperature for a given test with an accuracy of better than 1 deg F. Because of the difficulty, a choice was made after testing to plot the curves of figures 10 and 11 using the preselected nominal values of At. These factors contributed to the dispersion of the plotted error values in figures 10 and 11.

Results showed that the three types of sensors in the probe were materially affected by radiation. At 50 deg F (28 K) difference in temperature between the duct wall and the air at the center of the duct, the error for all three unshielded sensors due to radiation was about 3.8 deg F (2.1 K) at 300 fpm (1.5 m/s) 3.0 deg F (1.7 K) at 500 fpm (2.5 m/s). At 1300 fpm (6.6 m/s) the expected error with the sensors unshielded would be about 1 deg F (0.6 K). Radiation error for all three shielded sensors was less than 0.2 deg F (0.1 K) even at as low a velocity as 300 fpm (1.5 m/s) and for a At of 50 deg F (28 K).

There appeared to be little change in calibration of the thermistors during the period of two years in which the studies were made. The change in the thermocouples, including the reference thermocouple ranged from 0.0 to about 0.10 deg F (0.06 K). The change in calibration of the resistance thermometer during the period was not significant. The use of Chromel P-constantan thermocouples has the advantage of a high and of low dT

dE

thermal conductivity. They were also found to be easy to use during fabrication. They were rugged despite their small diameter. The electrical resistance of Chromel P-constantan wires is high, however, and when used with a digital voltmeter, without a preamplifier of high input impedance, such resistance may cause significant error. The Chromel P and the constantan wires should be tested for inhomogeneities [11].

It is recommended that a shielded probe as described be used for measuring temperature of moving air with either thermocouples, thermistors, or a resistance thermometer when the temperature of the duct is 5 deg F (2.8 K) or more above or below the temperature of the moving air in the center of the duct. Any of the three sensors could be used with greater accuracy at low air velocities when aspirated in a shield. The investigator can determine the approximate degree of error involved from the use of the curves of figures 10 and 11. It is immaterial whether the duct wall is colder or hotter than the moving air. If colder, the same

magnitude of error determined from the curves will apply but will have a negative sign.

Under conditions often occurring in a psychrometric calorimeter for measuring the capacity of air-conditioning equipment when the air velocity is 1300 fpm and At=5 deg F, the expected radiation error for unshielded sensors would be about 0.1 deg F. For 500 and 300 fpm, the error would be 0.2 deg F and 0.25 deg F, respectively. For all of these conditions the expected radiation error for shielded sensors is less than 0.1 deg F (0.06 K). Under the test conditions that prevail in the testing of air-conditioners and heat pumps in laboratories, it should be possible to reduce the total error in temperature measurement of the moving air to about 0.2 deg F (0.1 K) by a combination of suitable air mixers, duct insulation, radiation shields and calibration techniques. This value is in agreement with the value recommended by Davis, Faison, and Achenbach [1] for Temperature Measuring Standards.

If all other conditions remained the same, increasing the diameter of the test duct beyond the 10 in size used for this study would decrease the magnitude of the radiation errors of a shielded probe somewhat because of the smaller radiation effects through the open ends of the shields, but would have very little effect on the errors of an unshielded probe. The curves for the shielded condition are applicable only if shields of the same

material and low temperature-emissivity characteristics are used. A probe fabricated with material having a lower emissivity than methyl methacrylate, which has an emissivity of about 0.7 to 0.9, or one which is wrapped with a covering of reflective foil, would have a smaller radiation error when used as an unshielded probe. In an actual application, the temperature sensor would probably consist of a single element of considerably smaller diameter than the entire probe used for the study. Such an element would probably have a somewhat more favorable relationship between radiation heat gain and convection heat loss than the test probe. Presumably individual sensors could be designed which have no more error than the reference sensors. It is important that the design of a temperature-sensing probe incorporate principles for minimizing conduction errors similar to those used in this investigation.

Appreciation is expressed to Messrs. Thomas K. Faison, Walter M. Ellis, and Jesse Dungan of the NBS staff for their assistance in performing the tests and in difficult fabrication of some of the pieces of equipment.

As determined from materials such as paints which have similar radiation characteristics at the temperatures under consideration.

[1] Davis, J. C., Faison, T. K., and Achenbach, P. R., Errors in temperature measurement of moving air under isothermal conditions using thermocouples, thermistors, and thermometers. ASHRAE Trans., vol. 73, Part I pp. VII 1.1–1.10 (1967).

[2] Caldwell, Frank R., Temperatures of thermocouple reference junctions in an ice bath, J. Research NBS c, Eng. and Instr. 69C, No. 2 (April-June 1965). [3] Friedberg, S. A., Semiconductors as thermometers, Temperature, Its Measurement and Control in Science and Industry, 2, Reinhold Publ. Corp. (New York, N. Y., 1955).

[4] Drums, C. R., Thermistors for temperature measurements, Temperature, Its Measurement and Control in Science and Industry, 3, Pt. 2, Reinhold Publ. Corp. (New York, N. Y., 1962).

[5] Faison, T. K., Davis, J. C., and Achenbach, P. R., Performance of louvered devices as air mixers. NBS Building Science Series Report (in preparation).

[6] Roeser, W. F., Thermoelectric thermometry, Temperature Its Measurement and Control in Science and Industry, 1, Reinhold Publ. Corp. (New York, N.Y., 1939).

[7] Riddle, John L., Notes to supplement resistance thermometer reports. Obtainable from Temperature Physics Section, Heat Division, Institute of Basic Standards, National Bureau of Standards, Washington, D.C. 20234.

[8]

[9]

Eckert and Drake, Mass and Heat Transfer, pp. 426-428, McGraw Hill Book Co., Inc. (1959). Fiock, E. F., Olsen, Lief O., and Freeze, Paul D., The use of thermocouples in streaming exhaust gas, Volume of the Third Symposium on Combustion, Flame and Explosion Phenomena at University of Wisconsin, pp. 655–658 (1948).

[10] Kreith, Frank, Principles of Heat Transfer, pp. 376 and 377, International Book Co. (1958). [11] Powell, Robert, Caywood, Lindsay P., Jr., and Bunch, M. D., Low-temperature thermocouples, Temperature and Its Measurement and Control in Science and Industry 3, Pt 2, Reinhold Publ. Corp. (New York, N. Y. 1962).