Page images
PDF
EPUB
[merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][ocr errors][merged small][merged small][merged small][merged small]

a heating cable of its own. Copper-constantan thermocouples were placed in two rings around the inside surface of the duct, one upstream and one downstream from the probe, as well as on the inner surface of the duct along the bottom of the entire length of its measuring section, and on the bottom of the access door. These thermocouples were used to monitor the temperature of the duct wall.

The temperature of the air in the room and surrounding the duct was within 2 deg F of the temperature of the air at the intake to the duct. The air intake was from a large reservoir of air that had been held constant to within 0.15 °F for a period of at least 1 hr before each test. The air was exhausted from the duct to a remote part of the large air reservoir. The instrument room, remote from the room housing the duct, was maintained at about 75 °F throughout each test.

A concentric-louvered, metallic air-mixing device developed at the National Bureau of Standards [5] was used to insure thorough mixing of the inlet air. This resulted in a uniformity of air temperature at the entrance to the heated section. To observe the degree of temperature uniformity throughout a test, constant monitoring of air mixing took place using three arrays of 17 buttwelded Chromel P-constantan thermocouples in cross sectional areas in the duct. The butt-welded thermocouples were approximately 2 in from each other, and each one on the outer periphery of the array was approximately 1 in from the inside surface of the duct. One thermocouple was in the center. One array was upstream in the section of the duct that was unheated, another immediately upstream from the sensors in the heated section, and a third downstream, also in the heated section. No. 36 B and S gage wire (0.005 in diam) was used for the thermocouples. The use of the fine-gage, butt-welded thermocouple wires reduced conduction and radiation error. The flow was turbulent in the region close to the duct wall even at 300 fpm

and the low time constant of the thermocouples caused difficulty in making the thermocouple readings in this region. This difficulty was overcome by the use of an integrating type digital voltmeter by which an accurate average microvolt value for each thermocouple in this region was obtained. Due to the multiplicity of thermocouples, a zone box was used along with a single, common reference junction in an ice bath. The circuitry for using a zone box wih Chromel P-constantan thermocouple is described by Roeser [6].

There were radiation and conduction errors within each array, but these were sufficiently small so that the arrays served as satisfactory indicators of air mixing and radial temperature gradients. Throughout the testing program, when there was a measurable difference in temperature between the moving air and the duct wall, there were low radial temperature gradients and almost no fluctuation in temperature in the region immediately surrounding the probe. Throughout the rest of the volume of the stream of moving air, the air exhibited temperature fluctuations in increasing magnitude and frequency in locations near the duct wall.

Reference sensors were used as a means for obtaining temperature values almost completely unaffected by radiation and conduction and for making comparisons with the values obtained with the probe. These consisted of three thermocouple junctions fabricated from No. 36 Chromel P-constantan wire supported by a plastic cage-like structure. The six lead wires were bare and were strung back and forth in the cage so that about 5 in of the wire was exposed to the moving air. This technique, together with use of fine wires, minimized conduction error. The three junctions were butt welded to minimize radiation error. The average of the three thermocouple readings was used for the reference value.

Details of the experimental program required that two different types of reference sensors be used. The first, used for the tests where the probes

were shielded, fitted snugly into the inner of two 10-in-long concentric shields just upstream from the probe. This reference sensor and the concentric shields are shown in figure 5. The two shields were made of polished aluminum. The diameter of the inner shield was 2 in and that of the outer shield was 3 in. The second reference sensor, used for the tests where the probe was not shielded, fitted snugly into the inner of two 3-in-long concentric shields fabricated especially for the reference sensor. The sensor and its shields were placed immediately upstream from the probe. These are

[merged small][graphic]
[graphic]

FIGURE 5. Large shield and the reference sensor used in the tests where the test probe was shielded.

FIGURE 6. Small shield and the reference sensor used in tests where the test probe was unshielded. Removable legs for the shield are shown in lower left.

3. Calibration

The errors in uncalibrated temperature-sensing devices are usually determined by comparison of their indications with those of a secondary or primary standard whose errors have previously been analyzed. These comparisons are typically made in an apparatus in which there is minimum opportunity for heat transfer to or from the devices by radiation, conduction, or convection and in which the change of temperature with time is also minimized. An example is calibration of thermistors in a water bath whose temperature is close to the temperature of the room. In the present study the errors of several types of sensors were to be evaluated under conditions in which there was opportunity for heat transfer to or from the devices by radiation, convection, and conduction. In such a situation, a secondary or primary standard would also have unknown errors, so a simple comparison was no longer adequate.

Therefore, the following plan of attack was followed:

a. The platinum resistance thermometer used in the test probe was calibrated by the NBS Thermometry Laboratory, using the normal procedures described by Riddle [7].

b. Because of the well known stability of platinum resistance thermometers, the platinum resistance thermometer used in the test probe was used to calibrate the thermocouples and thermistors and also the two three-junction reference sensors in still air under isothermal conditions.

c. The magnitudes of small errors of the two three-junction reference sensors under nonisothermal conditions in the test duct were determined by comparing temperature indications when the junctions were aspirated at a velocity of 1400 fpm and when they were not aspirated.

d. The magnitudes of the larger errors of the thermocouples, thermistors, and the resistance thermometer in the probe were determined under nonisothermal conditions in the test duct for a range of air velocities by comparison with the calibrated reference sensors.

e. At the close of the program, the observed readings of the resistance thermometer in the probe were compared in still air with observed readings of another resistance thermometer calibrated by the NBS Thermometry

laboratory. At the four temperatures within the range of temperatures of 75 to 80°F at which the comparison between the two resistance thermometers was made, there was less than 0.02 deg F (0.01 K) difference between the observed readings of the two thermometers.

The thermocouples, thermistors, and the large reference sensor were calibrated four times within a year, and the small reference sensor twice, one year apart. These sensors were calibrated in air, shielded from ambient temperatures by a large insulated Dewar flask. This flask is shown in figure 7. An insulated cover was over the flask during calibration.

FIGURE 7. Calibration chamber for calibrating the thermistors, thermocouples, and the two reference sensors in still air.

The changes in performance for the thermocouples, thermistors and the reference sensors between calibrations are shown in table 1. The changes listed are consistent with those shown in isothermal tests reported by Davis, Faison, and Achenbach [1].

Before each calibration in still air, the Dewar flask housing a probe or the reference sensor remained for at least 24 hr at an ambient temperature constant within ±0.06 deg F and usually within ±0.04 deg F. After the calibration started there was usually a small rise in temperature in the room due to the heat from the lights and from the observers. Full consideration was therefore given to errors which might have been caused

by the order in which the thermistors or the thermocouples were read. Analysis showed that the contribution of self-heating of the resistance thermometer or the thermistors due to current flow was negligible.

When a reference sensor was calibrated, the cage-like structure was maintained at a center position in the flask so that the three No. 36 buttwelded Chromel-constantan thermocouple junctions were about 1 in from the end of the resistance thermometer. Figure 7 shows the reference sensor for the 10-in set of shields as it was placed inside the Dewar flask ready for insertion of the probe which housed the resistance thermometer.

The test probe with its three types of sensors, and the two reference sensors, were each calibrated as an assembly to avoid the problems of disassembly. These assemblies were calibrated in still air rather than water to avoid the need for insulating the sensors electrically.

When determining the magnitudes of the small errors in each reference sensor by the aspiration technique, the velocity of the air moving across the reference junctions within the inner shield was increased to 1400 fpm using a vacuum pump. Special adaptors were used for connecting the air-line from the inner shield to the vacuum pump. A comparison was made between aspirated and non-aspirated values at selected levels of temperature difference (At) between the duct wall and the moving air at the probe as measured by the resistance thermometer. Based on well-established theory [8, 9], calculations showed there was no appreciable error in the reference sensor due to radiation or conduction at a velocity of 1400 fpm, even for a At of 50 deg F.

This comparison showed that the error of the large reference sensor in the long 10-in set of shields reached a value of 0.08 °F when the difference in temperature between the duct wall and the moving air was 50 deg F and the air velocity was either 300 or 500 fpm. The errors of the small reference sensor in the small set of shields were considerably greater than for the large reference sensor under comparable conditions. Curves showing the errors determined in this manner, and used for corrections for both reference sensors, are shown in figure 8. A small uncertainty is indicated at the bottom of the curves for each reference sensor by a horizontal line across the figures at a distance about 0.04 deg F above the zero error line at the 0 °F point. For both reference sensors, the indicated temperatures when aspirated were 0.04 deg F lower than when unaspirated with no temperature difference between the duct wall and the moving air. This disparity is the reverse of what would be expected as a result of impact effects. While the cause is not known, it is probable that there was a change in the air mixing pattern within the shield when the air velocity was changed. Since the error involved was not due to radiation it was not considered in the corrections.

[graphic]
[graphic]

TABLE 1. Change in performance of the sensors between calibrations (deg F.)

Aug., 1965

[blocks in formation]

Tests were performed at 300 and 500 fpm with and without shields around the probe. Tests at 1300 fpm were performed only without shields, because the theory and results from the tests at 300 and 500 fpm showed that the radiation error at 1300 fpm with shields would be almost nonexistent even for a ▲t of 50 deg F. Tests for each condition were performed with none of the sensors aspirated, and during a continuous 12-hr period to reduce the effect of differences in ambient conditions, thereby facilitating better comparisons.

Before the design of the probe was fixed, calculations had been made to determine if conduction of heat from the duct wall along the wires was significant [9,10]. Before the tests, experimental verification was made. Hot metal, about 50 deg F hotter than the temperature of the duct wall, was applied suddenly in the hollow phenolic tube (designated as C on figure 1) to the plastic tube housing the lead wires. There was no indication of temperature change on any of the indicating instruments for the thermocouples, the thermistors, or the resistance thermometer. The same technique was used for the reference sensors. In each case there was no indication of temperature change.

All air velocity measurements were made with a pitot tube and a self-calibrating manometer which could be read to the nearest 0.01 in water gage.

Tests with the long metal shields around the probe were performed first. The sensors of the probe were not aspirated. The reference sensor assembly was placed in the end of the inner shield. Each test was performed at a selected level of velocity of 300 or 500 fpm. For these tests the difference in temperature (At) of the moving air and the duct wall was held at 0, 10, 20, 30, 40, or 50 deg F for a period long enough to assure steady-state conditions. The value of At was considered to be the difference between the reading of the resistance thermometer and the average of the readings of all the thermocouples in the rings around the duct, at the bottom, and on the access door. Steady-state conditions were obtained for each test, but no attempt was made to obtain values of At any closer than within 1 deg F of the preselected values. At each condition of velocity and temperature difference, 15 readings were taken for the thermocouples, the thermistors, and the resistance thermometers. The readings were 2 min apart.

Before taking the readings, readings of the thermocouples of the three arrays along the duct were taken to determine if the air mixing pattern was satisfactory. A satisfactory and randomly chosen cross-sectional temperature pattern from a test is shown in figure 9. This pattern was obtained during a test with At-40 deg F and with an air

« PreviousContinue »