A low cost instrument based on a solid state sensor for balloon-borne atmospheric O₃ profile sounding

Abstract

The design of an instrument based on a solid state tungsten oxide sensor for making profile measurements of atmospheric ozone from balloon platforms is described. The sensor is operated at a constant temperature, typically 530 °C. The importance of a detailed consideration of the electronic design is demonstrated, with particular reference to the circuit to control the sensor heater. Calibration methods which are straightforward to implement are illustrated, and the results of a test flight alongside an electrochemical ozonesonde are shown. Quantitative agreement within 25% for most of the profile demonstrates the potential of this type of sensor for ozone sounding. Future improvements from manufacture to analysis are expected to improve on this.

---

I. Introduction

Gas-sensitive resistors have been available commercially since the 1960s, generally capable of detecting combustible gases at the ppmv (parts-per-million by volume) level.¹ More recently, materials which can detect ozone at ppbv (parts-per-billion by volume) levels have been discovered.^2–4 Given that ground-level average ozone concentrations are typically 10–50 ppbv, depending on location and season, there is a clear opportunity to replace the current cumbersome and expensive ozone-monitoring technology with instruments employing solid state sensors.^5–7 Absolute concentrations of ozone are even higher in the middle atmosphere and solid state sensor technology may soon rival electrochemical ozonesondes for measurement of ozone profiles.

The use of solid state ozone sensors for measuring ground-level ozone concentrations has been described previously.^5–7 A promising technique involves cycling between two sensor temperatures on a time-scale of a few minutes to obtain a more reliable measurement. This method has no significant disadvantages for ground-based measurements because hourly-averaged ozone concentrations are sufficient for many purposes. In contrast, a balloon-mounted sensor must measure the entire ozone profile within a period of about 2 h. It is essential therefore that the sensor is operated continuously, in this case implying at constant temperature. In order to ‘reset’ the sensor^1,5 prior to a flight, the sensor is run at high temperature (650 °C) for 2 min after first turning on but before the launch, followed by indefinite operation at the lower measurement temperature (530 °C).

The ozone instrument we have developed, based around a solid state ozone sensor, is small, lightweight, has low power requirements and is inexpensive. It has flown many times since 1998 alongside ECC ozonesondes on meteorological balloons and short-duration⁸ (several hours) atmospheric research balloons. During this time the instrument has been subject to continuous development and improvement. Section II of this paper describes the operating principle of the sensor while the design of the ozone instrument and of the sensor are presented in Section III. Calibration procedures are illustrated in Section IV, and an example of flight data is shown in Section V. A more detailed analysis of the flight data will be described in a separate publication.⁹

II. Sensor operating principle

The sensing principle of the tungsten oxide sensor has been described in detail elsewhere^1,3,4,6 and a brief summary only is presented here. At high temperature, vacancies in the oxide lattice at the gas-sensor interface are generated by thermal ejection of oxygen atoms, releasing two electrons into the conduction band of the semiconductor:¹⁰

O_o → V + ½O₂

where O_o denotes a lattice oxygen site and V denotes an oxygen vacancy. Vacancies can be filled by reaction with either oxygen or ozone:

V + O₂ → O_o + ½O₂

V + O₃ → O_o + O₂

Filled vacancies are effectively electron traps and as a consequence the resistivity of the sensor increases upon reaction with oxygen or ozone. The ozone reaction is much faster and for this reason ozone can be detected at ppbv-concentrations even in the presence of atmospheric oxygen. The resistivity of the device is also determined by the temperature-dependent intrinsic conduction of the tungsten oxide. At constant temperature however, the sensor resistance is controlled by the local concentration of ozone and oxygen.

The solid state ozone sensor described here responds to absolute concentrations of ozone rather than to mixing ratios. Ozone concentrations are expressed as partial pressures (in mPa) for the remainder of the paper since this unit is in common use amongst the ozonesonde community.

III. Instrument design

A. Overview

Fig. 1 shows the schematic instrument design. A Triangle Digital Services logger (TDS2020) or PIC microcontroller (18F452) controls the sensor temperature and reads data from the analogue-to-digital (A-D) convertor. In the case of large research balloon flights, the data is logged to flash memory and telemetry is not normally used as the instruments are recovered after each flight. For flights with ozonesondes on meteorological balloons, all data is telemetered back to ground by interfacing with the ozonesonde TMAX card¹¹ which accepts digital data prepared by the solid state sensor instrument. More comprehensive data is recorded in flash memory for those occasions when the instrument package is found by a member of the public and returned. Some flights on weather balloons have also been made with a simplified instrument design in which the flash memory and A-D convertor are excluded, digitisation being performed by spare channels on the TMAX card.

Fig. 1 Schematic diagram of the ozone instrument configuration for large balloon flights.

For all flights, housekeeping/diagnostic data in addition to the sensor signal are sent and/or recorded, generally including: battery voltage, circuit temperature, pressure and two heater circuit voltages (V_AB and V_BC, see Fig. 2). The latter provide a check on the heater stability and allow the sensor power consumption to be calculated. These data have proven to be extremely useful for instrument diagnosis. LiMnO₂ batteries (camera-type) and rechargeable NiCd batteries have both been used. The rechargeable batteries have the advantage that they are able to deliver relatively high currents without a significant drop in voltage, while the lithium batteries are much lighter.

Fig. 2 Sensor heater control circuit incorporating a Wheatstone bridge. Pull-up resistors and decoupling capacitors are omitted for clarity.

The total instrument weight is currently 0.7 kg, including batteries (LiMnO₂) and polystyrene packaging for thermal and physical protection and overall dimensions are 135 × 225 × 260 mm³. These values could be reduced significantly in a commercial instrument, as well as the cost per unit which is currently about £200 for the components only. The average power consumption is on the order of 1 W.

For large balloons, the instrument package is attached either directly to the flight train, or to another instrument on the flight train, with straps. The separation between the balloon and ozone instrument is typically 20 m or more. For flights on meteorological balloons, the sensor package is taped to the ozonesonde which attaches to the balloon with string. The separation from the balloon is normally about 2 m which is sufficient to ensure that the balloon itself does not affect measurements.

B. Sensor design and fabrication

Fig. 3 illustrates the design of the ozone sensing element. One side of the 2 × 2 mm² alumina tile carries a platinum track which acts both as a heater and a resistance thermometer. The other side supports inter-digitated gold electrodes, overlaid by the tungsten oxide sensing material, approximately 40 μm thick. The oxide layer has a highly porous structure, allowing gas to penetrate through the material. Each of these materials is deposited onto the tile using commercial thick-film screen-printing technology. 50 μm thick platinum wires are bonded to pads which form the ends of the electrodes and the platinum heater track. The ceramic tile is attached by the wires to a plastic base which has four pins for making electrical contact. The sensor package occupies a volume of approximately 1 cm³ and can be handled directly, with care. The CAP21 ozone sensor is commercially available from City Technology Ltd.¹²

Fig. 3 Schematic diagram of the (a) upper and (b) lower sides of the sensor element. The tungsten oxide layer is shown partly cut-away to reveal the electrodes. Connections are made to the pads at the ends of the tracks. Air flow impinges vertically downwards onto the oxide layer.

C. Sensor flow configuration

The sensors have been flown in two different flow configurations. The first uses a small pump to draw air across the sensor which is housed in a small Teflon or glass chamber. Both ECC ozonesonde pumps, giving ∼200 ml min⁻¹, and miniature rotary pumps (ASF Thomas, models G6/01-E and G6/01-K-LCL), ∼1000 ml min⁻¹, have been used for this purpose. The latter is preferred because the dependence of the signal on flow rate is much smaller at the higher flow rate. This configuration allows for relatively straightforward laboratory calibration. In the second flow configuration, the sensor is mounted in the free air flow which takes advantage of the upwards motion of the balloon to provide air flow across the sensor. Clearly, this obviates the need for the pump and reduces power requirements, thereby saving significant weight.

D. Electronic design

Flight instruments experience very large changes in ambient conditions, notably pressure by about two orders of magnitude and temperature by up to 100 °C. The instrument must be designed to cope with these changes without loss of performance. Operation at low pressure can cause electronics to overheat due to greatly reduced heat loss via convective air cooling. In the present case, the total power dissipated by the electronics is less than 1 W and no problems have been encountered at low pressure. While the use of industrial- and military-specification electronic components ensures that the instrument remains functional over a wide range of circuit temperature, the accuracy of a circuit required to perform a quantitative function may degrade as the circuit temperature drifts. In practice, the electronics are mounted inside a small steel box, providing electromagnetic screening and thermal mass, which in turn is contained within polystyrene to give physical support and thermal insulation. With these measures, the circuit temperature typically varies over no more than 30 °C for a flight on a large balloon lasting several hours from launch to landing.

The instrument electronics perform two primary functions which determine the overall performance: control of the sensor temperature and measurement of the sensor resistance. The stability of the sensor temperature is crucial to the success of a measurement. By measuring the sensor response to ozone over a range of sensor temperatures, it can readily be shown that a temperature error of just 1 °C (i.e. a change in temperature between calibration and measurement) gives rise to an error in the deduced ozone of about 15%. Given that the usual operating temperature is approximately 800 K and the roughly linear relation between temperature and heater resistance, the heater driver circuit must be accurate to better than one part in 800. This stability is achieved by use of a Wheatstone bridge circuit (Fig. 2). The sensor is resistively heated by passing current through the platinum heater track which forms part of the bridge. The platinum track has a well-defined resistance-temperature characteristic. The out-of-balance signal across the two arms of the bridge drives an operational amplifier, the output of which regulates the current through the sensor heater via a field-effect transistor. Use of the op-amp with negative feedback ensures that the bridge circuit remains balanced as long as the values of R₃ and C₁ are chosen so that the circuit does not oscillate. Measurements have confirmed that the imbalance is below 0.5 mV under varying sensor flow conditions (corresponding to different heater currents), implying a negligible sensor temperature error.

The bridge is balanced when V_B1C = V_B2C, and this can only occur if the following equality is satisfied:

where R_control is the resistance on the arm of the bridge between points B₁ and C. The heater resistance R_h can therefore be fixed at a value corresponding to a desired temperature by choosing the value of R_control (or indeed R₁ or R₂). For example, the value of R_control could be conveniently set with a potentiometer. However, potentiometers have relatively high temperature coefficients, typically 100 ppm °C⁻¹, and consequently the temperature sensitivity of the bridge circuit would be dominated by the potentiometer. Calculations suggest that the sensitivity of the set-point temperature to a 30 °C change in circuit temperature would be 3 °C, unacceptably large. Instead, the value of R_control is determined by high precision (0.1%) resistors which have temperature coefficients of 15 ppm °C⁻¹. The sensitivity to a 30 °C change in circuit temperature has been reduced to 0.7 °C in this way. Fortunately, high precision resistors are available over a wide range of values and it is possible to set the sensor temperature within 1 °C of the desired value. The stability of this temperature is far more important than the precise value.

For the circuit shown in Fig. 2, the value of R_control is determined by R_high, R_low and the states of FET_high and FET_low which are controlled by the digital microprocessor. This combination allows finer control of the sensor temperature, as follows. If FET_high is set to the conducting state, in which its resistance is <4 Ω, the current passing through the sensor heater will drop essentially to zero as the circuit attempts to satisfy the above equation but cannot because with R_control < 4 Ω the setpoint temperature is, in effect, well below ambient temperature. With FET_high (and FET_low) in its non-conducting state, in which its resistance is in the GΩ range, R_control = R_high and the sensor temperature will be set to the corresponding value. If FET_low is then set to its conducting state, R_control equals the parallel resistance of R_high and R_low. Thus, FET_high gives on/off control while FET_low allows two-temperature operation, alluded to above, the higher temperature being controlled by R_high and the lower by R_high and R_low in parallel.

More recently, digital potentiometers with low temperature coefficients have become available. With relatively minor modifications to the Wheatstone bridge circuit, these devices allow finer and more sophisticated control over the sensor temperature by microprocessor. For example, the sensor temperature may be ramped over several seconds when changing states, and this helps to prolong the sensor lifetime by avoiding thermal shock.

The second primary function performed by the instrument electronics is measurement of the sensor resistance from which ozone concentrations are deduced. This resistance is measured by passing a constant current through the device and interrogating the developed voltage. In order to avoid sensor drift, this voltage must be kept below 100 mV which necessitates a low current, typically 100 nA. This current is generated by means of a closed-loop operational amplifier in conjunction with a field-effect transistor. The advantage of this design is that the stability of the sensing current is determined by the properties of high precision resistors which, as previously mentioned, have low temperature coefficients. The voltage across the sensor is amplified to a more convenient range, 0–5 V, using standard amplification techniques. The gain of the amplifier circuit is also fixed by the values of high precision resistors. The expected error in the resistance measurement with this circuit design is less than 0.2% for a 30 °C change in circuit temperature, implying a negligible error in deduced ozone.

IV. Sensor calibration methods

Since the precise response of each sensor depends on many factors it is necessary to calibrate the sensors prior to use. It is also necessary to ‘burn-in’ each sensor before first use by heating it at high temperature (∼600 °C) for a period. Without this preparation the sensor response drifts significantly during the first few hours of operation.

The laboratory setup for calibration is relatively simple. Synthetic air from a cylinder is used in the calibrations, containing low-ppmv levels at most of water, hydrocarbons and nitrogen oxides. Higher grade air with sub-ppmv concentrations of contaminants has also been tested to determine whether the minor constituents have an effect on the results. No significant difference in the calibrations was observed. Known concentrations of ozone are generated in the air flow by a Dasibi multi-gas calibrator (Model 5008 W/D&P). This instrument photolyses oxygen at 185 nm to produce ozone and incorporates a 254 nm photometer to calibrate the ozone concentration. Some of the air flow is vented to the atmosphere at this point to ensure that the sensor is calibrated at atmospheric pressure. The remaining air is drawn over the sensor by means of a miniature rotary pump. The sensor response is recorded by means of one of the instruments used for short-duration flights and subsequently downloaded to a PC. Teflon tubing and nylon fittings are used in the flow line up to the sensor in order to avoid loss of ozone on exposed surfaces. The sensor itself is mounted in a glass or Teflon housing. Although there are facilities for calibrating at known humidity levels, it is preferable to use dry air for calibration of sensors for balloon flights. Absolute humidity in the atmosphere drops very rapidly with altitude and so water vapour can affect the sensor response only near ground-level. Furthermore, at the relatively high operating temperature normally used for balloon flights (530 °C), the influence of humidity is very small.⁷

At the simplest level, a sensor may be calibrated by recording the sensor resistance as the ozone concentration is stepped through a series of different values. Data for such a calibration is plotted as a time series in Fig. 4. The ozone concentrations range from zero to ∼20 mPa, covering the range of absolute concentration normally encountered in an atmospheric profile measurement. The levels are alternated between high and low values to avoid hysteresis effects. The ozone concentration is held constant for 15 min, and Fig. 5 shows a plot of the concentration against sensor resistance, taken at the end of each 15 min period i.e. a calibration curve. Also shown is a quadratic polynomial fit to these data points; under these operating conditions (constant sensor temperature of 530 °C) it is invariably found that a quadratic fits the data well.

Fig. 4 Sensor response to a sequence of ozone concentrations (shown in mPa in the figure) during a calibration run.

Fig. 5 Ozone concentration versus sensor resistance for the calibration sequence shown in Fig. 4, together with a quadratic fit.

It has been found that the scatter of data points about the calibration curve is smaller if a high concentration of ozone is passed through the flow line for a period, typically 2 h, immediately prior to the calibration sequence. It is believed that ozone adsorbs onto the walls of the Teflon tubing, depleting the concentration reaching the sensor until equilibrium between the gas flow and the exposed surfaces is reached. By the same token, when the ozone level is switched to zero at the end of the calibration sequence, adsorbed ozone is released back to the gas phase and the concentration seen by the sensor does not immediately drop to zero. These assertions are supported by several pieces of evidence. Firstly, if a calibration sequence is performed without the two hour preconditioning period, a calibration fit deduced from the later ozone levels suggests a greater response of the sensor to ozone than a fit deduced from the earlier levels. Secondly, it is often possible to fit the response of the sensor to a (nominal) step change in ozone with a double-exponential function, indicating two different timescales for the response. We interpret these as the intrinsic response time (∼30 s) combined with a response time dependent upon details of the apparatus such as the length of tubing between the ozone generator and solid state sensor, approximately one metre in our case. Thirdly, analysis of flight data indicates a faster response than in the laboratory. Configurations for flights involve shorter lengths of tubing or a much wider flow channel plus much higher flow rate. The calibration shown in Figs. 4 and 5 included a two hour preconditioning period.

A drawback of tungsten-oxide-based solid state sensors is that their response is prone to drift. This problem has been carefully addressed recently,³ and a method to correct for the drift proposed, involving measurement in zero-ozone air. It is possible to apply this technique in balloon flights by using a pumped configuration. The zero-ozone measurement can be made by turning the pump off periodically to stop the flow. Any ozone remaining in the vicinity of the sensor will be quickly destroyed by the sensor itself. However, it should be emphasised that problems with drift are a relatively minor issue for short-duration flights lasting a few hours (with the ascent generally taking about 2 h), assuming a recent calibration is available, since the sensor drift takes place on a much longer timescale.⁷

V. Flight data

A comprehensive analysis of flight data will be presented in a separate publication.⁹ Here, we illustrate results from one flight only to give an indication of sensor performance. Fig. 6 shows a comparison of ozone profiles measured by the prototype ozone instrument and a conventional ECC ozonesonde, flown on the same balloon. The solid state sensor profile is derived using a simple laboratory calibration, as described in Section IV, with no reference whatsoever to the sonde data. It is immediately clear that the correlation in fine structure is extremely good. Indeed, such structure tends to show deeper peaks and troughs in the solid state sensor data than in the ozonesonde data, an indication that the former has a somewhat shorter time response. Agreement is quantitative within ∼25%, except in the troposphere and at the lowest pressures. There are several reasons to expect discrepancies with pressure. One reason is that the sensor responds to the oxygen concentration as well as to ozone, leading to changes in the calibration with altitude. Falling pump efficiency (in terms of volume flow rate) may also have an effect at low pressure. Treatment of these issues is beyond the scope of this paper and has not been included in the comparison shown. These effects will be addressed in more detail in future publications.⁹ By comparison, it should be borne in mind that ECC ozonesondes have systematic errors of up to 17%.^13–15

Fig. 6 Performance of the solid state sensor relative to an ECC ozonesonde for a flight on 14th May 1999 from Aberystwyth.

VI. Conclusions and future work

This work has demonstrated the feasibility of balloon-borne ozone measurements using a lightweight semiconductor sensor with custom-designed control and recording electronics. The semi-conducting oxide sensors are highly selective to ozone (by virtue of very high sensitivity to this gas) and show excellent potential for atmospheric ozone profile measurements. An instrument based on these sensors would cost a fraction of the price of the ECC ozonesondes in current use, particularly if mass-produced. Future work will focus efforts towards a more quantitatively accurate measurement which requires correct treatment of the effects of reducing ambient pressure during a sounding.

Acknowledgements

This work was supported by the Natural Environment Research Council and the European Union. The authors would also like to thank City Technology Ltd and Prof. Geraint Vaughan (the University of Wales, Aberystwyth) for their contributions.

References

1. D. E. Williams, Sens. Actuators, B, 1999, 57, 1.
2. P. McGeehin, P. Gleeson and A. Walton, Solid state ozone sensors for ppb to ppm operation, Paper B4.3, Sensor 97 Proceedings, AMA Service GmbH, Wunstorf, Germany, 1997, Vol. 1, p. 171, http://www.sensor-test.com/main/dshkgxco/dvhswdps/page.html.
3. S. R. Aliwell, J. F. Halsall, K. F. E. Pratt, J. O’Sullivan, R. L. Jones, R. A. Cox, S. R. Utembe, G. M. Hansford and D. E. Williams, Meas. Sci. Technol., 2001, 12, 684.
4. D. E. Williams, S. R. Aliwell, K. F. E. Pratt, D. J. Caruana, R. L. Jones, R. A. Cox, G. M. Hansford and J. Halsall, Meas. Sci. Technol., 2002, 13, 923.
5. R. L. Jones, R. A. Cox, D. E. Williams, K. F. E. Pratt, P. McGeehin and J. O’Sullivan, Tungsten oxide sensors employed for atmospheric ozone measurements: comparison with UV analysers, Paper B3.1, Sensor 99 Proceedings, AMA Service GmbH, Wunstorf, Germany, 1999, Vol. 1, p. 159, http://www.sensor-test.com/main/dshkgxco/dvhswdps/page.html.
6. S. R. Utembe, G. M. Hansford, M. G. Sanderson, R. A. Freshwater, K. F. E. Pratt, D. E. Williams, R. A. Cox and R. L. Jones, submitted to Sens. Actuators B.
7. S. R. Utembe, PhD thesis, University of Cambridge, UK, 2000.
8. J. P. Pommereau, F. Goutail, I. Pundt, J. A. Pyle, F. Danis, G. Hansford, R. Freshwater, A. Robinson, R. L. Jones, N. R. P. Harris, A. Adriani, F. Cairo, L. Pulvurenti, G. Difrancesco, S. Kirkwood, H. Nilson, J. Arvelius, P. Woods, N. R. Swann, I. H. Howieson, S. Garcelon, T. D. Gardiner, T. Deschler and A. Buivan, Proceedings 14th ESA Symposium on European Rocket and Balloon Programmes and Related Research, Potsdam, Germany, 31 May–3 June 1999 (ESA SP-437, September 1999), p. 609.
9. G. M. Hansford, R. A. Bosch, S. R. Aliwell, K. F. E. Pratt, J. F. Halsall, R. A. Cox, D. E. Williams and R. L. Jones, manuscript in preparation.
10. V. E. Heinrich and P. A. Cox, in The Surface Science of Metal Oxides, Cambridge University Press, Cambridge, 1994, p. 20.
11. http://www.users.qwest.net/∼tlthompson/index.html.
12. http://www.citytech.com/.
13. R. A. Barnes, A. R. Bandy and A. L. Torres, J. Geophys. Res., (Atmos.), 1985, 90, 7881.
14. W. D. Komhyr, R. A. Barnes, G. B. Brothers, J. A. Lathrop and D. P. Opperman, J. Geophys. Res., (Atmos.), 1995, 100, 9231.
15. I. S. Boyd, G. E. Bodeker, B. J. Connor, D. P. J. Swart and E. J. Brinksma, Geophys. Res. Lett., 1998, 25, 2409.

---