Sensor and I/O Calibrations

Calibration Procedures and Instrumental Accuracy Estimates of TAO Temperature, Relative Humidity and Radiation Measurements

H. Paul Freitag, Yue Feng, Linda J. Mangum, Michael J. McPhaden, LT Julia Neander, and Linda D. Stratton


INDIVIDUAL SENSOR AND I/O BOARD CALIBRATIONS

Air Temperature Sensor

Air temperature measurements were made by model MP-100 humidity-temperature probes manufactured by Rotronic Instrument Corporation of Huntington, New York. Specifications pertaining to sensor accuracy given by the manufacturer are listed in Table 1. Sensors were calibrated at PMEL by immersing them in a controlled water bath and measuring their output voltage at seven temperatures over a range of 14°C to 32°C. The resultant temperature-voltage pairs were fit to a least squares linear equation. A sample calibration is included in Appendix A. Statistics from the calibration of 31 PROTEUS and 154 ATLAS AT sensors were quite similar (Table 2). The RMS maximum residual for both groups of sensors was an order of magnitude smaller than the manufacturer's stated accuracy of 0.5°C and linearity of 0.2°C. The sensor gain (coefficient b) was essentially equal for both PROTEUS and ATLAS groups and was within 1.4% of the manufacturer's nominal gain of 100. Sensor to sensor differences were small with the standard deviation of the gain being only 1% of the mean. Sensor bias (coefficient a) of 0.2°C (PROTEUS) to 0.3°C (ATLAS) is near the manufacturer's nominal bias of 0. The standard deviation as a percentage of the mean for coefficient a is large (>100%) for both ATLAS and PROTEUS moorings because the coefficient itself is so small.

Table 2. Statistics for sensor calibrations. M is the number of sensors calibrated. Calibration equations are shown in Figure 3. Coefficient percent deviation is the standard deviation divided by the mean expressed as a percent.

Table 3. Statistics for board calibrations. M is the number of boards calibrated. Calibration equations are shown in Figure 3. Coefficient percent deviation is the standard deviation divided by the mean expressed as a percent. Residuals have been scaled to have same units as sensors.

Air Temperature I/O Board

Air temperature I/O boards have 10-bit A/D converters and were calibrated at PMEL by applying a known voltage and recording the digital output. These data were then fit to a least squares linear equation. A sample calibration is included in Appendix A. Statistics from the calibration of 34 PROTEUS and 186 ATLAS AT boards are shown in Table 3. The RMS maximum residual for the PROTEUS boards (expressed in temperature units) was 0.019°C or half the single-bit resolution of the boards and thus as small as possible for the given resolution. This value was half the RMS maximum residual for the AT sensor. The RMS maximum residual for the ATLAS boards was significantly higher at 0.136°C. The difference was due to the range of voltages over which the calibration was performed. Originally both groups of boards were calibrated down to 0°C (0 volts) and both groups had comparable maximum residuals. It was noticed that the maximum residual occurred almost entirely at the 0 v value. Since this was a value that would never be experienced in the tropics the calibrations were recomputed for the PROTEUS boards after omitting this calibration point and the maximum residuals decreased to the level shown in Table 3.

Air temperature data from both PROTEUS and ATLAS moorings archived before 1994 were computed using calibration coefficients which used the 0 v value, but were surely more accurate than what the 0.136°C residual implies since at tropical air temperatures (18 - 32°C) calibration residuals were typically much lower. Data archived in 1994 and after will use calibration coefficients based on typical tropical ocean values. AT board gain (coefficient b in Table 3) was essentially equal for both PROTEUS and ATLAS groups, despite the fact that the PROTEUS calibrations omitted the 0 v data. This implies that the true accuracy of the ATLAS board was comparable to the PROTEUS.

Sea Surface Temperature Sensor

Sea surface temperature measurements were made by model 46006 thermistors manufactured by Yellow Springs Instrument Co., Inc. (YSI) of Yellow Springs, Ohio. Specifications pertaining to sensor accuracy given by the manufacturer are listed in Table 1. Sensors were calibrated at PMEL by immersing them in a controlled water bath and measuring their output resistance at seven temperatures over a range of 14°C to 32°C. The resultant temperature-resistance pairs were fit to a non-linear equation (Fig. 3). A sample calibration is included in Appendix A. Statistics from the calibration of 38 PROTEUS and 269 ATLAS SST sensors were quite similar (Table 2). The RMS maximum residual for both groups of sensors was nearly two orders of magnitude smaller than the manufacturer's stated interchangeability of 0.2°C. Differences between the PROTEUS and ATLAS sensors appear to be insignificant.

Sea Surface Temperature I/O Board

Sea surface temperature I/O boards converted thermistor resistance to voltage, which in turn was converted to a frequency proportional to the voltage. This frequency was then counted for 4 seconds and recorded. The boards were calibrated at PMEL by placing precision Vishay resistors on the boards and recording the output counts. This procedure was repeated at seven levels which corresponded to a temperature range of 14°C to 32°C. These data were then fit to a least squares linear equation. A sample calibration is included in Appendix A. Statistics from the calibration of 34 PROTEUS and 196 ATLAS SST boards are shown in Table 3. The RMS maximum residuals (expressed in temperature units) were 0.002°C (PROTEUS) and 0.004°C (ATLAS) and were of the same magnitude as the SST sensor residuals. SST board gain (coefficient b in Table 3) was essentially equal for both PROTEUS and ATLAS groups.

Relative Humidity Sensors

Relative humidity was measured by the same sensor (Rotronic model MP-100) used to measure AT. Specifications pertaining to sensor accuracy given by the manufacturer are listed in Table 1. Sensors were calibrated at PMEL in a manner recommended by the manufacturer. The sensors were attached to a calibration chamber into which a humidity standard was introduced. Both the calibration chamber and humidity standards were supplied by the manufacturer. The humidity standards were precisely titrated saturated-salt solutions with stated accuracies of better than 0.5% RH. Sensors were calibrated between 20% RH and 95% RH (presently modified to 50% RH to 95% RH) at 15% RH increments. A sample calibration is included in Appendix A. Calibration chambers used on the PROTEUS and ATLAS sensors differed in that the PROTEUS sensors were calibrated with the filter cap removed while ATLAS sensors are calibrated with the filter cap installed. Rotronic cautions that while a clogged filter may not produce an erroneous reading, it may significantly increase the sensor response time.

Statistics from the calibration of 33 PROTEUS and 40 ATLAS RH sensors differed more than the AT calibrations from the same sensors (Table 2). The RMS maximum residual for PROTEUS sensors was 0.07% RH, or about half the nominal accuracy of 2% RH, but the residual for ATLAS sensors was 2.6% RH which exceeded the nominal accuracy by about 30%. ATLAS sensor gain (coefficient b) was 6% higher than the nominal value of 100 while the PROTEUS gain was 3% higher than the nominal value. One explanation for this difference would be that the ATLAS sensors did not come to equilibrium (due to increased response time caused by clogged filters) before the calibration values were recorded. In any case, both of the calibration residuals were large in a relative sense when compared to those for AT and SST which were smaller than manufacturers' specifications by at least an order of magnitude.

Relative Humidity I/O Board

Relative humidity I/O boards have 10-bit A/D converters and were calibrated at PMEL by applying a known voltage and recording the digital output. These data were then fit to a least squares linear equation. A sample calibration is included in Appendix A. Statistics from the calibration of 34 PROTEUS and 190 ATLAS RH boards are shown in Table 3. The RMS maximum residuals (expressed in humidity units) were 0.17% RH (PROTEUS) and 0.26% RH (ATLAS). Both were less than the single-bit resolution of the boards. As with the AT boards it was found that maximum residuals tended to be found at the 0% RH (0 volts) calibration point. This point was omitted from the PROTEUS calibration data which may account for it being smaller than the ATLAS value. For both ATLAS and PROTEUS the RH I/O board residuals were about an order of magnitude smaller than the sensor residuals. RH board gain (coefficient b in Table 3) was about equal for both PROTEUS and ATLAS groups.

Shortwave Radiation Sensors

Shortwave radiation measurements PROTEUS moorings only) were made with model PSP precision pyranometers manufactured by the Eppley Laboratory, Inc. of Newport, Rhode Island. Specifications pertaining to sensor accuracy given by the manufacturer are listed in Table 1. The sensors were calibrated by the manufacturer when new and were returned for recalibration after mooring recovery. The calibration procedure entailed comparing the sensor output at 700 W m -2 to a standard sensor. Quoting from Eppley's calibration report, these calibrations are "traceable to standard self-calibrating cavity pyrheliometers in terms of the Systems Internationale des Unites (SI units), which participated in the Seventh International Pyrheliometric Comparisons (IPCVII) at Davos, Switzerland in October 1990." Eppley specifies that the sensors are linear to 0.5% at radiation intensities up to 1400 W m -2. The mean and standard deviation of calibration coefficients from 23 calibrations performed by Eppley are given in Table 2.

Shortwave Radiation I/O Board

Shortwave radiation I/O boards have 10-bit A/D converters and were calibrated at PMEL by applying a known voltage and recording the digital output. These data were then fit to a least squares linear equation. A sample calibration is included in Appendix A. Statistics from the calibration of 23 PROTEUS SWR boards are shown in Table 3. The RMS maximum residuals (expressed in radiation units) was 1.8 W m -2, which is only slightly larger than the resolution of the board. By comparison, the SWR sensor's manufacturer specification of 0.5% linearity and 1% temperature dependance imply uncertainties of 5 W m -2 and 10 W m -2, respectively, for insolation values of 1000 W m-2 (typical of cloudless, mid-day values in the tropics.)

Subsurface Temperature Sensors

Subsurface temperature measurements were made by the same YSI 46006 thermistor that was used for SST. Calibration procedures differed from those for SST in that SBT was calibrated as a total system, i.e., the thermistor and V/F counter were calibrated as one at 10 temperatures over a range of 6°C to 32°C. The calibration data were fit to the same non-linear equation as the SST sensor after applying a nominal gain factor to the V/F output. A sample calibration is included in Appendix A. The RMS maximum residual from the calibration of 3666 ATLAS SBT sensors of 0.002 C was smaller than the RMS residual for SST sensors (Table 2). One reason for the lower SBT residual may be that the number of SBT sensors calibrated was an order of magnitude larger than that for SST sensors and therefore less weight was given to outliers.
Return to Table of Contents
Go to Next Section