The Challenges Of Maintaining And Improving The Uncertainty Of An Industrial Humidity Calibration Laboratory Speaker Jack Herring, Michell Instruments, Inc., 11 Old Sugar Hollow Road, Danbury CT 06810, USA phone (203) 744-6881, fax (203) 791-2047, email jack.herring@michell.com Author Andrew M. V. Stokes Michell Instruments Limited, Lancaster Way, Ely, CB6 3NW, UK, phone +44 1353 658004, fax +44 1353 658095, e-mail andrew.stokes@michell.com Abstract The paper describes the issues associated with maintaining a busy industrial humidity calibration laboratory and the challenges faced in trying to maintain the laboratory measurement uncertainty whilst achieving a throughput of thousands of sensors per year. The humidity calibration facility of Michell instruments is essentially split into two parts a UKAS (EA) accredited laboratory for high level measurements with direct traceability and audit path to NPL and NIST standards, and a commercial laboratory providing lower level tertiary calibration of tens of thousands of dewpoint sensors per annum. The UKAS section focuses on excellence rather than on volume hence the processes and procedures are largely manual and quite time-consuming, whilst the commercial section is there to handle large volumes of sensors automatically and with the minimum of human intervention. The two therefore require very different approaches in terms of equipment, procedures and analysis and the uncertainty levels achieved reflect the type of operating model used in each case. Humidity calibration systems used by secondary laboratories have tended to be constructed either as clones of National Standards, using a two-pressure or two-temperature generation method, or as simple divided flow systems utilising calibrated vertical tube flow meters. The former are very expensive to produce and have certain limitations in terms of usability. The latter are cheaper to produce, but also suffer from inflexibility and difficulty in automation. Furthermore, these systems tend to offer varying flow rates dependent on the generated humidity level. The humidity calibration system described in this paper provides accurate and highly repeatable humidity generation using a combination of liquid and gas mass flow controllers. It allows automated use through the integration of a precision chilled mirror dew-point hygrometer that provides both the control feedback to the generator and traceability to National Humidity Standards. The paper describes the two processes, provides a practical consideration of the component uncertainties and explores ways in which these uncertainties can be refined and minimised through improved procedures, better equipment and careful operation. Also provided are detailed calculations of the liquid and gas mass flow ratios used to derive appropriate humidity levels in the measurement chamber. A novel technique to ensure sensitivity and stability of the generated humidity is described, along with the techniques employed to ensure homogeneity of the humidified air. The paper also describes the physical design and construction challenges that were overcome in producing a fully integrated system. An uncertainty budget for the whole system is provided, indicating the key contributory factors and suggesting ways in which the measurement uncertainty can be minimised.
1 Background Calibration is the comparison of a measured value with the true value. It has a formal definition under ISO standards: Calibration: the set of operations which establish, under specified conditions, the relationship between values indicated by a measuring instrument.. And the corresponding values of a quantity realised by a reference standard Adjustment is the setting of a minimum deviation from measured value and true value. Again, according to ISO: Adjustment: the operation intended to bring a measuring instrument into a state of performance and freedom from bias suitable for its use National Standard: A standard recognised by an official national decision to serve, in a country, as the basis for fixing the value of all other standards of the quantity concerned. The national standard in a country is often a primary standard. Details of several prominent national standards can be seen in figure 1, below. Figure 1. Prominent international standards laboratories. Traceability: The property of the result of a measurement standard, generally international or national standard, by an uninterrupted chain of comparisons All definitions taken from ISO 10 102-1: (1992)
Regular calibration of humidity sensors is important not only to maintain traceability, but also to ensure the correct operation of the process concerned. Most humidity instruments need calibration at least every year to ensure correct performance - some need calibration every few months. Of course, calibration can be taken to mean either verification or adjustment, followed by certification Figure 2. Calibration traceability hierarchy. The term uncertainty is defined as the tolerance band in which (at k=2) a measured value will be in agreement with the ACTUAL value for 95 % of the measurements made. It is effectively a confidence band for the measurement and can be calculated by combining various experimental and theoretical data on the system and instruments in use, according to well established formulae. Figure 3. Normal distribution chart. The uncertainties of the instrument in use and the calibration reference are both important. Whereas a traceable laboratory typically has an uncertainty of the order of ±0.2 C dew point (k=2 coverage) a typical field instrument may have an uncertainty ten times worse. The calibration of a hygrometer will result in two components - an off-set (often termed its accuracy) at a given point and an uncertainty value. So for example, if it has an off-set of +0.2 C dew point and the calibration uncertainty is ±0.1 C dew point, then you have 95 % confidence that the actual value = (reading -0.2) ±0.1. Michell Instruments Limited has been manufacturing and calibrating chilled mirror and capacitive (impedance type) dew-point hygrometers for more than thirty years. During most of that time, the basic method of calibration has been consistent. This process consists four component parts:
o Preparation of a supply of clean, dry compressed air with a moisture content below the minimum measurable range of the instrument or sensor to be calibrated o Generation of a range of dew-point temperatures across the operating range of the instrument or sensor to be tested, using a single- or multi-stage divided flow method o Provision of a suitable test chamber to house the instruments or sensors under test o Verification of the whole system performance on a continuous or sampling basis using a traceable reference hygrometer In the early years the performance of these calibration systems was limited by a number of factors, for example the quality and stability of the dry air source, the sensitivity limitations of the mixing system and the uncertainty of the reference measurement instrumentation. In more recent times individual components of the calibration system have been optimised to produce a better overall uncertainty for the particular calibration process, or product type, for which they have been designed. Figure 4. Typical calibration system overview. Although based on the same physical principles, the often disparate needs of the various types of calibration dictate very different forms. This is particularly true in the case of a laboratory set up for UKAS calibrations and also set up for volume calibration of mass produced hygrometers. The different challenges and their solutions, as implemented by Michell Instruments, are discussed in the remainder of this paper. 2 UKAS Laboratory: Quality Over Quantity Figure 5. UKAS calibration hierarchy. Referring back to our definitions, traceability is the property of the result of a measurement standard, generally international or national standard, by an uninterrupted chain of comparisons. Clearly, the longer the chain the more uncertainty is introduced. The more uncertainty there is in a measurement the greater the potential error.
Therefore it follows that the aim of a secondary laboratory, such as one operating under the UKAS accreditation scheme, is to provide calibrations with as short a chain of comparisons back to the appropriate national standards as can be practically achieved. In the case of Michell Instruments UKAS laboratory this means that the comparisons are performed directly against our in-house reference standard chilled-mirror hygrometers. These in turn are periodically calibrated against the primary national standards in the UK (NPL gravimetric hygrometer) and USA (NIST). Michell first obtained UKAS (then NAMAS) accreditation in 1987 and have since developed and refined our SOP s and established a large database of calibration history on the various reference standard hygrometers, some of which have calibration history right back to the mid 1980 s. The UKAS part of our facility now has XX reference standards, Y traceable to NPL and Z to NIST. The laboratory has increased capacity from its original one to one comparison to the level now where up to five working standards or customer instruments can be compared to the reference standard simultaneously. The process is however still largely a manual one and is labour intensive, and therefore expensive. Figure 6. Michell Instruments UKAS accreditation. The performance limitations of a calibration methodology are determined by a combination of factors including: Physical facilities of the laboratory used temperature control air quality stability of electrical power supply Quality of staff training
attention to detail repeatability variations in staff Calibration equipment Accuracy Repeatability Traceability In any laboratory these factors can be quantified statistically and combined to give a measure of that laboratory s capability as an uncertainty budget. The methods for calculating an uncertainty budget are well documented and beyond inclusion in a general paper of this type. For illustration, the uncertainty budget at -90 C dp is comprised of the following: Components of Uncertainty Calculation at -90 C dp Parameter Divisor Uncertainty Standardised Calibration of standard 2 0.41 0.205 Drift of standard 1.732051 0.5 0.288675135 Repeatability 1 0.01 0.01 Bridge calibration 2 0.014 0.007 Bridge repeatability 1 0.005 0.005 Bridge drift 1.732051 0.0225 0.012990381 Contamination 1.732051 0.015 0.008660254 Temp. gradients in condensate 1.732051 0.01 0.005773503 Temp. gradients in mirror 1.732051 0.01 0.005773503 Pressure difference 1.732051 0.01 0.005773503 Mirror temperature fluctuations 1 0.0319 0.0319 Non linearity 1.732051 0.08 0.046188022 Sampling uncertainties 1.732051 0.39 0.225166605 Standard Uncertainty 0.42394224 K=2 0.84788449 Figure 6. Uncertainty budget calculation for -90 C dp. The overall uncertainty of the UKAS section of the dew point calibration laboratory is shown most clearly in the chart 1. On the following page.
Chart 1. UKAS measurement uncertainty. 3 Tertiary Calibrations: Achieving High Volume While Maintaining Quality Figure 7. Tertiary calibration hierarchy Michell Instruments production output exceeds 30,000 hygrometer sales per annum, each of which requires a traceable calibration. Given the limitations of the UKAS equipment previously mentioned, this requires a different approach to be taken. What is required here are calibration systems that can deal with multiple IUT s simultaneously with minimal additional uncertainty and, preferably, minimal operator intervention. The issues associated with the design, validation and calculation of uncertainty for such a system are detailed below. A primary enabling step that allowed Michell Instruments to develop a new calibration technique was their development of a range of sensors and instruments for which the calibration process is much easier to automate. In the past low-cost sensors tended to have used simple analogue trimming circuits to enable zero, span and linearity to be adjusted. More modern transmitters and instruments have benefited from digital communications and significant on-board memory capacity that in turn facilitate automated calibration processes.
A further factor in the need to improve reliability and automation of the calibration process is the response speed of sensors and systems to changes in applied humidity. At the lower humidity limit of the early systems (typically -75 C frost point, equivalent to 1 part per million) the response time for a system to a dry-down from ambient conditions might have typically been 24 to 48 hours. Current demands for instrumentation calibrated down to - 100 C frost point or below (13 parts per billion) demands much longer dry down times due to molecular absorption, even on carefully prepared and optimised electro-polished stainless steel surfaces. It is common for 120 hour (or longer) dry down periods to be used in order to ensure full system equilibrium prior to a calibration. These elongated calibration periods increase the need for system reliability, not only to ensure proper calibration, but also to ensure satisfactory throughput of product to meet customer demand. Typical stabilisation times for a calibration system such as that described in this paper are shown below. Dewpoint Temperature Minimum stabilisation time -100 5 days -90 12 hours -80 10 hours -70 8 hours -60 4 hours -50 2 hours -40 1 hour -30 1 hour -20 1 hour -10 1 hour 0 1 hour +10 1 hour +20 1 hour Figure 8. Calibration system stabilisation times For tertiary calibrations the objective was to produce a calibration system with the following key criteria: overall uncertainty better than ±2 C frost point at the low end and better than ±1 C frost/dew point at the high end; greater set-point stability; fully automated; increased capacity; self diagnostic; automatic sensor verification and self-adjusting. 3.1 Available Technologies: Pro s & Con s 3.1.1 Dry gas supply Two alternatives were considered dry compressed air and liquid nitrogen. Ultimately it was decided that dry compressed air would be used, primarily on the grounds of safety, ease of implementation and familiarity. Michell has always used dried compressed air and found the method to be extremely reliable provided a sensible preventative maintenance programme is in place for compressors and dryers. A two stage drying process of oil-free compressed air at 7 barg is used to give a dry air feed to the generator of around 13 ppb moisture content, approximately -100 C frost point when expanded to atmospheric pressure. In theory a nitrogen source, dried through a nitrogen cold trap, can generate much lower moisture levels, but this was regarded as an un-necessary additional complication to the project. Practical experience has shown that the dry air system, now implemented on eight calibration systems, has resulted in zero down-time during more than two years of continuous operation.
3.1.2 Humidity Generator A feasibility study was undertaken to consider three options for humidity generation: twotemperature generator, two pressure generator and mass flow mixing system. These three technologies are amongst the most widely implemented within National Standards laboratories and commercial calibration laboratories worldwide. A mass flow mixing system was chosen because of the ability to produce a system with flow rate flexibility, easy control and relatively low cost in a short time scale. The humidity generator is the main subject of this paper and will be described in more detail in the following sections of this paper. 3.1.3 Test Chamber and Sampling System Considerable applications experience and expertise were utilised in designing a sensor calibration manifold and sampling system to ensure maximum integrity of the generated humidity at the test point. The design of this part of the system is described in more detail further on in the paper. 3.2 Calibration Integrity and Traceability Although not strictly necessary, based on the nature of the humidity generation method, it was decided to use a chilled mirror dew-point hygrometer as an on-line reference and as the control element of the closed loop generation system. This hygrometer provides traceability through an un-broken chain to NPL and NIST standards. 4 Generator Design 4.1 Basic Concept: The basic design concept was to use a novel combination of cascaded air mass flow controllers in conjunction with a proprietary liquid mass flow controller with controlled vaporisation/mixing, in order to allow the generation of any humidity level between -100 and +20 C frost/dew point, equivalent 13 ppb to 23,600 ppm. This represents a dynamic concentration range of 2 x 10 6. As a result, extremely careful selection of mass flow controllers was necessary to ensure an appropriate sensitivity and reproducibility of generation. 4.2 Selection Of Suitable Mass Flow Controllers: A complex mixing system comprising three stages of mixing of wet and dry air flows was developed in conjunction with the Dutch manufacturer Bronkhorst Hi-Tec, using their model EL-Flow liquid and gas mass flow controllers and a Bronkhorst CEM Controlled Evaporator and Mixer unit. Figure 9 illustrates the basic working of the system.
Figure 9. MFC based dewpoint generator schematic. A pressurised pure water source is fed through the Liquid Mass Flow Controller, LFC, at a controlled flow rate from 0 to 0.945 litres per minute. In the first stage of mixing this liquid water is mixed with a source of dry air (at approximately -100 C frost point) controlled by air mass flow controller GFC1 at a rate of between 0 and 5 litres per minute. This mixture then passes through the CEM evaporator/mixing unit and into an air receiver that acts as a further buffer, improving output stability of the total system. A controlled amount of this pre-mix air is then mixed in the second stage using GFC2, with more dry air through GFC3, the excess flow from the first mixing stage being exhausted the electronic pressure controller EPC1 and a rotary vent valve. In a similar arrangement, GFC4 and GFC 5 provide third stage mixing of pre-mix (2) and dry air, with the excess again vented through EPC2. The resultant mixed gas is then ready for delivery at a controlled flow rate to the sensors under test. Figure 10. MFC based dewpoint generator.
Whilst relatively simple in principle, practical realisation of a good working system relies both on the careful selection of the appropriate controllers and a sophisticated electronic control system, which in the case of this generator system is developed as a Windows based automated routine using C++ programming language. Table 1 on the following page lists the six mass flow controller elements used in the generator with their operating flow range and reproducibility. Controller Range Worst Accuracy, Nl/min LFC 0 to 12.5 g/h 0.125 g/h <0.0125 g/h GFC1 0 to 10 Nl/min 0.1 <0.01 GFC2 0 to 2 Nl/min 0.02 <0.002 GFC3 0 to 10 Nl/min 0.1 <0.01 GFC4 0 to 0.2 Nl/min 0.02 <0.002 GFC5 0 to 10 Nl/min 0.1 <0.01 Table 1. MFC selection. Reproducibility, Nl/min The liquid mass flow controller is calibrated in grams per hour. This can be converted to a volumetric flow rate of the evaporated water as follows: 1 gram mole occupies a volume of 22.4 litres at stp, so 18.001 g occupies 22.4 litres Therefore a flow of 12.5 grams per hour is equivalent to 22.4 x12.5/18.001 = 15.555 litres per hour. So, the Controlled Evaporation method is an effective method for precisely delivering a saturation water vapour flow at a rate as low as 0.0026 litres per minute. 4.3 Calculation Of Theoretical Generator Settings: Whilst the design of the generator allows for either open-loop or closed-loop operation, it has been set up to operate in a semi-open loop configuration whereby the reference hygrometer is used to validate the generated humidity and control the overall calibration cycle, but not to provide direct closed loop control of the mass flow controllers. This latter feature can be considered as a design modification that would further improve the system in the future. In the open-loop configuration, mass flow controller settings are developed based on a simple theoretical model using volumetric calculation of the liquid water and multi-stage dry air mixing stages. A theoretical generated frost-point temperature is calculated from the mass flow controller settings as follows, assuming dry gas feed is at 13 ppb moisture content. In this example the target generated frost point is -90 C: Stage 1: Wet flow is 0.308 g/h water = 22.4*(0.308/18.001) = 0.3833 litres per hour Dry flow is 5.001 l/min = 300.06 litres per hour at < 0.01 ppmv Resultant flow is effectively 10 6 *0.3833/300.06 = 1277.41 ppmv (-17.64 C f.p.)
Stage 2: Wet flow is 0.055 l/min @ 1277.41 = 0.055*1277.41/5.055 = 13.8986 Dry flow is 5.000 l/min @ 0.013 = 5.000*0.013/5.005 = 0.0130 Resultant flow is therefore (13.8986 + 0.0130) = 13.9116 ppmv (-58.02 C f.p.) Stage 3: Wet flow is 0.041 l/min @ 13.9116 = 0.041*13.9116/5.041 = 0.1131 Dry flow is 5.000 l/min @ 0.013 = 5.000*0.013/5.005 = 0.0130 Resultant flow is therefore (0.1131 + 0.0130) = 0.1261 ppmv (-88.46 C f.p.) Three-stage mixing at even the lowest humidity levels provides a good repeatability, being less than 0.2% of controlled value for each flow controller. Table 2 illustrates the relevant flow settings for all six mass flowmeters required to generate a -90 C frost point. Mixing Stage Flow settings, Nl/min [g/h] Target frost point, Repeatability, C C Stage 1 GFC1=5.001, [LFC=0.308] -17.64 ±0.02 Stage 2 GFC2=0.055, GFC3=5.000-58.02 ±0.01 Stage 3 GFC4=0.041, GFC5=5.000-88.46 ±0.01 Table 2. Three stage mixing of a -90 C frost point. This gives a system repeatability, excluding variations in dry air moisture content, of ±0.025 C frost point, calculated in quadrature based on maximum repeatability errors at each mixing stage. 4.4 Inclusion Of A Buffer Air Reservoir: Initial test results with the system indicated that a problem existed in achieving a stable generated humidity. It was discovered that operation of the liquid flow meter at very low flow rates, even with the addition of the Controlled Evaporator and Mixer, was giving rise to an erratic evaporation rate associated with the evaporation of individual water droplets. Whilst not expected to be a problem for the calibration of impedance type hygrometers, this cyclic response with a peak-to-peak variation of approximately 0.5 C at +10 C dew point and 1 C at -20 C frost point was large enough to cause total system uncertainty levels to exceed the required levels. It was therefore decided to incorporate a secondary mixing chamber after the first stage premix in order to provide a time averaging of this fluctuation and reduce its peak-to-peak effect from the above levels down to 0.04 C at +10 C dew point and 0.2 C at -20 C frost point.
The subsequent remixing ( x 2) of this first stage air mix improves the fluctuation further down to levels that are insignificant compared to the overall system uncertainty. Whilst selection of an appropriate air reservoir was important from the perspective of pressure rating, for safety reasons, the materials of construction were less important as the lowest pre-mix frost point at this stage is approximately -20 o C, rendering adsorption/wall effects insignificant. Charts 2. and 3. show the effects of the air reservoir on the output stability at -20 C frost point, plotted over a twenty minute period with no air dryer changeover, in order to de-couple any variation that may be caused by imbalance between the efficiency of the two molecular sieve desiccant columns. Chart 2. Generator output stability at -20 C frost point without air reservoir.
Chart 3. Generator output stability at -20 C frost point with air reservoir. 4.5 Design Of The Sensor Calibration Chamber: In the past, Michell had used multi-sensor calibration chambers machined from a solid block of stainless steel, with drilled and plugged gas flow paths to minimise dead space and internal volume. These blocks were expensive to produce and had limited capacity due to the design, accommodating a maximum of 24 traditional sensors or 12 transmitters, due to the larger housing size required for transmitter electronics. For the new humidity calibration systems it was decided to employ a new design of sensor calibration chamber, to accommodate the larger transmitter format and allow a much larger number of sensors to be calibrated simultaneously. The general arrangement of a typical calibration manifold is shown in figure 11. Figure 11. A calibration manifold (Note the system is loaded from both sides).
A vertical format stainless steel manifold design was implemented, giving each of the eight calibration chambers the capacity to house up to 250 transmitters during any calibration cycle. The calibration manifold is welded construction with high integrity VCR coupling gas inlet and outlet ports. Each transmitter is secured into the calibration tube using its sensor guard thread, and is sealed to the manifold with an individual Teflon o ring. As a further enhancement, each calibration chamber is mounted within a sealed environment during the calibration process. The cowling is purged with air at -100 C frost point to minimise humidity gradients between the inside and outside of the sensor calibration chamber thus negating any leakage effects that might result from a poor seal at the sensor o ring. Tests have indicated that this can reduce the differential (air out air in) by more than 80%, so improving the calibration integrity of the sensors under test and therefore reducing the overall uncertainty of calibration. Figure 12 View of a section of the laboratory The air path from generator to sensor calibration chamber is maintained in electro-polished 10 mm OD stainless steel tubing with a minimum of bends and welded angle connections. A 5 metre tail pipe after the calibration chamber, in 6mm OD stainless steel tube, is used to reduce back diffusion effects. 5 Verification of System Performance This calibration system is essentially a practical device for the calibration of capacitive type dew-point sensors over the range -100 to +20 C dew point. As such, it is designed to give a reasonable uncertainty in context of the nature of the sensors being calibrated. The intention in design was to produce a highly reliable system that would allow an overall calibration uncertainty for the finished product (capacitive sensors) of ±1 C dew point from +20 to - 60 and ±2 C in the range -60 to -100 C frost point. Table 3 on the following page illustrates the results of a series of 95 calibration runs, from which the generation uncertainty and overall system uncertainty were calculated. It should be noted that the overall combined system uncertainty incorporates the uncertainty of the reference hygrometer, calibrated in Michell s UKAS accredited laboratory over the range -75 to +20 C dew point. The reference hygrometer uncertainty below -75 C frost point is determined from previous experimental data.
Target Dew Point, o C -100-90 -80-70 -60-50 -40-30 -20-10 0 10 18 Mean Actual Dew Point, o C -98.59-89.80-80.24-71.30-60.11-50.25-40.09-29.81-20.11-10.05 0.14 10.40 18.08 Standard Deviation 1.13 0.74 0.48 0.67 0.25 0.23 0.23 0.17 0.19 0.14 0.25 0.17 0.16 Uncertainty (95% CI) 2.25 1.49 0.97 1.33 0.51 0.46 0.45 0.34 0.38 0.29 0.51 0.34 0.32 Max Value, o C -95.8-87.8-79.5-70.1-58.3-49.7-39.8-29.4-19.7-9.7 0.8 10.8 18.6 Min Value, o C -101.7-91.8-82.1-73.7-60.7-51.2-41.1-30.6-20.5-10.4-1.1 9.9 17.7 Spread, o C 5.9 4.0 2.6 3.6 2.4 1.5 1.3 1.2 0.8 0.7 1.9 0.9 0.9 UKAS Inst. Uncertainty (95%CI), o C 1.00 0.82 0.64 0.48 0.34 0.32 0.31 0.29 0.27 0.25 0.24 0.22 0.20 Combined System Uncertainty (95% CI), o C 2.46 1.70 1.16 1.41 0.61 0.56 0.54 0.44 0.47 0.38 0.56 0.40 0.38 Table 3. Summary of actual test results and calculation of combined system uncertainty. As can be seen from the results, the target combined uncertainty from -60 to +20 C dew point exceeds the design brief by an average of around 50%. Whilst the combined uncertainty currently exceeds the ±2 C target at lowest frost point (not generated), it is expected that significant improvements will result when the reference hygrometer can be calibrated against a national standard to these levels and further enhancements are made to the system to improve low end reproducibility. 6 Summary A commercial calibration laboratory exists to maintain a strong link with national standards, and by appropriate means to be able to offer calibration services with the best possible overall uncertainty. The practical challenge is to be able to meet the requirements of low volume traceable calibration and also to be able to provide high volume calibrations with minimal detriment to uncertainty. In this paper we have seen how a high volume humidity calibration system is able to provide practical and reliable calibration of capacitive dew-point sensors. Figure 13. Unloading calibrated transmitters
In conjunction with a fully automatic, computer control program, these systems are able to generate a range of dew point levels, verify them in real time against a traceable reference chilled mirror dew-point hygrometer, measure and record the data from the 250 sensors under test and validate correct programming of sensor memory with the appropriate calibration data. This allows the fully automatic calibration of capacitive dew-point sensors in a controlled environment with no human supervision or involvement. We have also demonstrated that these systems have, in the main, achieved their design brief and have delivered a combined uncertainty of less than ±2 C dew/frost point across an extremely wide measurement range of -100 C frost point to +20 C dew point. These systems have been fully operational for several years now, each one capable of performing in the region of 40 calibration runs per year, providing a total capacity of more than 80,000 calibrated dew-point sensors in a 12 month period. An extended validation programme has demonstrated both the reliability and long-term stability of the system. So, in conclusion, we have achieved our objective and have in one laboratory systems based on similar technologies that are capable of low volume, high labour content, traceable calibration on the one hand and high volume automated calibration on the other. Bibliography Bronkorst Hi-Tech BV, Mass Flow Meter Manual, 29 Dec 2000 Bronkhorst Hi-Tech BV, CEM Controlled Evaporator Manual, 13 Oct 1997 G W C Kaye and T H Laby, Handbook of Physical and Chemical Constants, 14 th Edition, 1973 The Institute of Measurement and Control, A Guide to the Measurement of Humidity, 1996 B Cretinon and J Merigoux, La Mesure de l humidite dans le Gaz, 2000 C Nordling, J Osterman, Physics Handbook, 1980 ISO 10 102-1: (1992)