Alan J. Fougere
Falmouth Scientific, Inc.
N. Falmouth, MA 02556

Neil L. Brown
Woods Hole Oceanographic Institution
Woods Hole, MA 02543

Edward Hobart
EC Microsystems, Inc.
Falmouth, MA 02540

ABSTRACT

1.0 INTRODUCTION

In response to the Improved Meteorological Measurement Program (IMET), a part of the World Ocean Circulation Experiment (WOCE), the authors have developed an improved temperature measurement device. The overall goal of the IMET program is to develop accurate and reliable means of making meteorological measurements from ships and buoys during the WOCE Hydrographic Program (WOCHP). The program goals are: atmospheric temperature measurements with a range of -45 to 45 Celsius with resolution of 0.001 Celsius, and sea surface temperature measurements with a range of -5 to 30 resolution of 0.001. Accuracy goals for both measurements are +1- 0.005 degrees celsius¹. The sensor module must also interface directly with the IMET data collection and logging system. To allow ease of integration to the IMET system a serial digital output data format was chosen using a standard interface protocol (RS-232C or RS-485). The wide temperature range of operation coupled with the high initial accuracy and long term stability required the use of a platinum resistance thermometer.

2.0 TEMPERATURE SENSOR

Both versions of the temperature modules use a platinum resistance thermometer. Shown in Figure 1 is an outline drawing of the sensor. The design objective of the sensor was to maximize the absolute stability of the device. The sensing element uses a helical wound four coil platinum element secured within a ceramic sleeve with four holes along its principal axis. The element is mounted inside the stainless steel sheath and is held in uniform contact through dense packing with ceramic powder. This design insures time constant uniformity among sensors. The predicted time constant for the sensor is 200 milliseconds.

To ensure high stability we selected a slightly larger sheath to ensure pressure protection of the element. The platinum element is housed in a pressure protecting non-corrosive 316 stainless steel sheath which is .093" in diameter and 2.0" long. The sheath is mounted in a guard designed to maximize the flow around the sensing element area of the device, while preventing the entrapment of bubbles typical of other platinum thermometers with guards with small holes. The guard structure has overall dimensions of .250" diameter by 2.5" long.

3.0 ELECTRONICS

In order to maintain the smallest sensor while providing the user with the standard serial data, a new approach to high accuracy temperature measurements was required. Existing AC ratiometric techniques² as used in previous sensors are too large and complex to enable mounting in a small self contained instrument. Alternate "Wien" bridge oscillator techniques are inappropriate due to their inherent temperature instability and reliance on high output, low stability Thermistor sensors. However, the use of four terminal measurement techniques coupled with frequent internal calibration against fixed resistors, simplified and reduced the size of the electronics. This approach relies on internal logic to select calibration elements or select sensor output, measure the circuitry's response to them, and then compute the actual measured value from the sensor. The advent of the single chip microcontroller when used in conjunction with the auto-calibrating electronics allows for the direct conversion of the output into engineering units in ASCII format.

An illustration of the calibration circuit is shown in Figure 2. The basic temperature bridge is made up of precision resistors Rl, R2 and Rf and the thermometer Rt. The self calibration circuit consists of precision resistors Ra, Rb, Rc, and Rd. A secondary circuit consisting of temperature sensor Rt' is used to measure and correct for thermal shifts in both the reference and calibration resistors. This secondary temperature measurement removes the requirement to tie the reference and calibration resistors thermally to the temperature source being measured. The large physical structures associated with such thermal connection result in long thermal stabilization times between the reference resistor and the outside temperature. The overall procedure for the system calibration is as follows:

Firstly, a laboratory calibration of the electronics is performed. A fixed value is connected for the thermometer Rt and measurements of the four output signals Etl, Ed, Ec2, and Ec3 are made at two temperatures. The thermal shifts in the reference and calibration circuit is determined and correction constants generated. These constants allow the microcontroller to correct the thermal drift of the reference network numerically.

Secondly, the platinum thermometer is reconnected to the circuitry. A laboratory calibration is performed where several values of temperature (T) and output data from Et (X) are obtained and a polynomial fit can be established such that:

T = Ac + Bc * X + Cc * X^2

At the same time when T is at the low end of the temperature range Sl is set to "Elo" position and the value of X noted (Xlo). Similarly, Sl is set to the "Ehi" and "Eha" positions and the values of X (Xhi) and (Xha) are noted respectively. Using the values of Ac, Bc, and Cc from above the values of Tb, Tha, and Thi calculated where Tb, Tha, and Thi are the precise temperatures simulated by the reference network. Ideally, the reference network Ra, Rb, Rc, Rd and Rf should have near zero temperature coefficients and be absolutely stable. However, as we correct for the effects of temperature by the direct measurement of the temperature of the resistors, only the absolute stability of the resistors is required. To ensure adequate stability of the resistors high precision "Vishay" type devices are used. These resistors exhibit only 1 to 2 PPM drift/year after elevated temperature burn-in stabilization. The approach of using a precision resistor network to accurately simulate a sensor output at accurately known values of the measured parameter allow the use of simple electronics with negligible requirement for electronic stability. This results in electronic techniques to be employed which dramatically reduce the size, cost, and power consumption associated with traditional methods.

The above description has been predicated on the use of electronics and sensors with a simple transfer function. However, if the transfer function is more complex a 3rd or 4th order polynomial fit can be performed using the same techniques. A calibration resistor network would be required to simulate 4 or 5 different values of the measured parameter with the appropriate correction algorithms. The additional cost and complexity of accommodating a more complex transfer function is trivial. Overall accuracy is thus dependent only on the platinum thermometer and the simulation network of resistors. Since the standard platinum thermometer transfer function has been accurately described by a 2nd order polynomial, the simulation network consists only of four precision resistors.

The sequence of events for a single measurement is as follows:

1. Switch 51 is set to the Ed position (see Figure 2) which is nominal "zero calibrate" position and a reading if the output (Xlo) is taken.

2. Switch Sl is set to Ec2 position which is the nominal "half scale calibrate" position and a reading (Xha) is taken.

3. Switch Sl is set to Ec3 position which is the nominal "full scale calibrate" position and a reading (Xhi) is taken.

4. Switch Si is set to Et position, (the sensor output) and a reading (X) is taken.

5. From the outputs Xlo, Xha, and Xhi for accurately known simulated temperatures Tb, Tha, and Thi already found from laboratory calibrations, we calculate A, B and C such that:

Tnn = A + B * X + C * X^2

6. The final step is to calculate the true temperature from the coefficients A, B, and C and the system output X obtained in 4 above.

The remaining elements of the circuitry are the "Signal Generator" and the "Signal Conditioning" circuits. These networks allow the analog measurement circuits outputs to be digitally interfaced to the microcontroller for processing.

4.0 DIGITAL ELECTRONICS

The microcontroller is an Intel 87C51FA. The 87C51FA microcontroller is an enhanced version of the 87C51 part with additional RAM space and internal hardware features. The controller has internal EPROM, 256 bytes RAM, 32 programmable I/O lines, and 7 interrupt sources. The controller has four main peripheral devices for the input and output of data. A 2K X 8 serial access EEPROM provides storage of configuration parameters and calibration constants. A 10 bit 2 channel AID converter makes the auxiliary temperature measurements of the reference resistors. All analog signals obtained above are converted to digital equivalents compatible for input the microcontroller. Engineering units are output in ASCII format compatible with either RS-232C or RS-485 protocols. The circuitry also contains DC/DC power converters to generate analog voltage levels and a precision 2.5 volt reference for the auxiliary temperature channel.

5.0 POWER

The unit operates from either a single 12.0 volt or 5.0 volt dc source. The unit can place the analog front end in standby mode, thereby reducing the overall power consumption to 5 milliwatts. When the unit is fully operational, it consumes 100 milliwatts of power.

6.0 FIRMWARE/SOFTWARE

The firmware requirements of the temperature sensor fall into two categories. The first is the real time requirements for continuous measurement of the sensor and calibration network signals. These measurements must then be converted into a singular temperature value in engineering units. Secondly, there are issues related to the user interface and control of the unit. Both tasks are limited by the size of the microcontroller's memory and classic "speed versus power" relationships. The priority one requirement for the firmware is the acquisition of analog signals and combining these into a single temperature value. In order to maximize the amount of data acquired while providing the highest level of resolution it is desirable to make continuous measurements; this requires the microcontroller to be prepared for the next task prior to the completion of the last. The firmware is structured such that it avoids gaps in the time series record of data and maximizes the performance of the measurement circuitry. Thus, the measurement circuit is utilized at a 100% duty cycle, thereby, effectively narrowing the bandwidth of the system and improving overall noise performance.

The firmware monitors the status of the conversion and ensures that the next cycle starts immediately at the conclusion of the last. Parameter selection and timing of the measurement cycle is determined by an upper level supervisory routine. Communications are interleaved into the measurement process. Initialization of the controller and its measurement cycle require special sections of code invoked by the supervisor based on status and control flags.

The user interface has two levels of operation. The first is providing the measured data₁ either on request or automatically when new data are available. Neither sequence interferes with measurement operations. The second level covers calibration, setting calibration constants, and operational parameters.

These tasks however do have an influence on measurement operations. Size of the RAM in the microcontroller restricts the length of commands and their responses. Commands are limited to four character tokens plus a number. Responses are equally short, the longest being a 14 character number (sign, 7 digits, decimal, 4 digits).

During calibration, commands to select specific parameters are accepted. While in this mode, data are transmitted for only the currently selected parameter. Changing the selected parameter causes re-synchronization of the internal measurement cycle. When installing calibration or operational parameters all analog measurements are suspended. Parameters may be written into temporary storage, reviewed and tested. If acceptable these may then be written into EEPROM for future operations. To assist the user in communicating and obtaining data from the instrument a complete set of IBM-PC compatible Quick-Basic routines have been developed. These routines allow for high level operation of the instrument without detailed knowledge of the command set by the user. This software is intended as a building block to the user's overall data collection and storage system.

7.0 PHYSICAL

All the electronic circuits are contained on two printed circuit cards which mount within a 1.5" inside diameter tube. The oceanographic sensor is constructed from 6AL-4V titanium permitting instrument operation to 7000 meters. An outline drawing of the device is shown in Figure 3. The unit's outside dimensions are 1.9" in diameter by 9.0" long.

The atmospheric sensor uses a waterproof Celcon HC grade plastic housing and is mounted in conjunction with a solar radiation shield. The outline drawing of the device is shown in Figure 4. Dimensions are 4.72" in diameter by 15.9" long.

Both instruments connect power and communication via a 4 pin waterproof connector.

8.0 TEST RESULTS

Testing was conducted to determine the thermal stability of the electronics. The test was performed by substituting a fixed precision resistor for the platinum thermometer. The unit was then thermally cycled over the operating range (the fixed resistor held at a constant temperature). Errors of less than +1- 5 PPM were observed. At the time of this writing, the temperature calibration of the electronics was in progress and initial results matched the theoretical performance objectives. Rigorous long term testing of the stability has not been completed. At the time of presentation calibration data for both sensors will be provided along with the latest data for absolute accuracy and stability. Presently the predicted system accuracy's are +1-0.003 and +/- 0.006 Celsius for the oceanographic and atmospheric sensors respectively. Data supplied by the thermometer manufacturer show its long term temperature drift to be less than 0.001 Celsius/year RMS. Based upon the initial experimental and calibration data, the instrument will exceed the IMET WOCHP measurement requirements.

9.0 CONCLUSION

A high accuracy digital output sensor has been developed which meets the requirements for the WOCE IMET air and sea surface temperature measurements. Additional applications include use as a secondary thermometer on EG&G MKIIIB CTD's. The small size, low power, digital output and high accuracy of the modules make them suitable for a wide range of oceanographic and atmospheric measurements.

ACKNOWLEDGEMENTS

The design team would like to thank Dr. Craig Dorman, Director of Woods Hole Oceanographic Institution, for his active support. We also give thanks to the WHOI CTD Group for access to and support in the CTD Calibration Facility; and, specifically, the personal efforts of Gary Bond without which we would not have completed this publication in time.

REFERENCES

1. Robert A. Weller and David S. Hosom, "Improved Meteorological Measurements From Buoys and Ships for the World Ocean Circulation Experiment", Oceans' 89, Seattle Washington, September 1989.
2. N.L. Brown, "A precision CTD microprofiler", Proc. IEEE Oceans '74 Conference, vol.2 no.2, pp. 270-278, 1974.
3. A.M. Pederson, "A Modular High Resolution CTD System with Computer-Controlled Sample Rate", Proc. MTS 1984 STD Conference and Workshop, San Diego, February 1984.
4. Vishay Resistive Systems Group, "Technote 13 - Design Guide for Use of Precision Resistors", February 1, 1981.

Figure 1

OUTLINE DRAWING PLATINUM RESISTANCE THERMOMETER

Figure 2

BLOCK DIAGRAM ELECTRONIC CIRCUITRY

Figure 3

OUTLINE DRAWING OCEANOGRAPHIC

TEMPERATURE SENSOR MODULE

Figure 4

OUTLINE DRAWING ATMOSPHERIC

TEMPERATURE SENSOR MODULE