European Harmonisation and Progress in the Field of Dimensional Metrology

Transcription

1 1 European Harmonisation and Progress in the Field of Dimensional Metrology 1 Introduction H. Haitjema and P.H.J. Schellekens Eindhoven University of Technology The field of dimensional metrology is, contrary to its somewhat static image, subject to dynamical changes in various aspects. In the field of primary standards and calibrations at the highest accuracy level, there is a trend towards mutual recognition not only based on confidence, but accompanied by a system of key comparisons which gives some justification of that confidence. This development is the finishing touch for an European - and even world-wide - laboratory accreditation system where national organisations recognise each other based on measurement intercomparisons and peer reviews. This gives a world-wide network of laboratories where traceability and a harmonised approach to measurement uncertainty are established. In the field of standardisation, within the ISO a thorough review and development of the written standards concerning dimensional metrology is made in the framework of the GPS masterplan [1] in which the whole chain between product documentation and measurement standards can be covered. An essential standard is the ISO [2] which implies that the uncertainty and traceability which is about to be generally agreed for measurement standards has to be applied for all industrial measurements also. In the meantime, the demands of industry increase so rapidly that the metrology research laboratories have difficulties following it. Some of the new developments in this field are described in this paper. 2 Mutual recognition 2.1 Mutual recognition of national metrology institutes Until a few years ago, the mutual recognition of national metrology institutes (e.g. NPL in Great Brittain, PTB in Germany, NIST in the USA, NMi-VSL in the Netherlands, etc) was only implicitly arranged. Historically, all countries which signed the metre-convention were supplied with primary standards by the BIPM. Most physically this is still the case for the kilogram, for which the international prototype still is at the BIPM in Paris and where all national prototypes are compared with, about each 10 years. Until 1960 for the metre the situation was similar as the standard metre-bar was present in Paris too. Since primary standards can be realised independent, the role of

2 2 the BIPM has evolved to a more coordinating one; except for own research the BIPM organised comparisons of primary standards on a voluntary base. Mutual recognition of national metrology institutes was based on the confidence that each institute could well derive its secondary standards and transfer them via calibration to accredited laboratories and the industry, or on a more subjective base ( if we know the people ). Another indirect way of mutual recognition was via the European Accreditation organisation: where within Europe there was a mutual recognition agreement between national accreditation organisations, this implied that traceability to and certificates of - the national metrology institites concerned were mutual accepted too. Where national metrology institutes are not only concerned with realising primary standard, but also offer calibration services at lower uncertainty levels, often overlapping the field of accredited calibration laboratories which had to fulfill much higher demands before being accredited, this was not a fair situation. Also problems arised when accreditation organisations over the world tried to achieve some mutual agreement; here the poorly defined requirements for national metrolog institutes are a problem To overcome this, recently [13] a mutual recognition between national metrology institutes was agreed where institutes who sign agree: to participate in internationally organised key comparisons which are organised world-wide and which are representative for high-level calibraton services offered by metrology institutes. to undergo some kind of evaluation according to ISO 25 /EN 45000, using a system of peer review for the technical aspects The key-comparisons are to be followed by intercomparisons on a more regional base, e.g. for Europe by Euromet, the European cooperation of national standards laboratories. As example of a result of such an intercomparison is shown below. The figure shows the results for a tungsten carbide gauge block with a 80 mm nominal length with their uncertainty based on 2s (covering a 95% confidence interval). Although it is a real example this is not relevant for the discussion. As the results are all obtained independently, with participants realising their primary standards independently, they should be statistically consistent. However, at first sight (and also after a thorough statistical analysis) it appears that participant no 7, and maybe participant numer 6, have overlooked or underestimated some systematic deviation in their measurement. How this situation is to be treated further, how a reference value can be established and what the consequences should be for participants 6 and 7 is still to be figured out.

3 3 dev. from nominal length / nm mm TC gauge block participant number figure 1. Example of a result of a (key) intercomparison Note that within the European accreditation organisation a situation like this would be more straightforward: the reference value including an uncertainty is attributed to the participating (e.g. participant 1) national laboratory, and for a result like this, laboratory 6 and 7 would get a limited amount of time to find and solve their problem, or loose their accreditation. The mutual agreement between national metrology institutes will be based on many comparisons like this, of which the results will be generally available via the BIPM website Mutual recognition between accreditation organisations The mutual recognition inside Europe for national calibration, testing, and certifying organisations is now establish for 16 countries and, outside Europe, agreements exist with organisations in Australia, South-Africa, New-Zealand, USA, Singapoge and Hong-Kong. It is maintained by a system of peer review and intercomparisons which is an example for organisations at other continents, for regional metrology organisations (e.g. Euromet) and the system for national metrology institutes described in section 2.1. A problem for these kind of organisations are the demands for the paperwork and formalism both for themselves and for the laboratories, where the technical aspect, the art of measurement can be overlooked. Nevertheless, within technical expert committees useful work is done, not only organising intercomparisons, but also in the making

4 4 of some technical guidance (e.g. document EA-10/05: coordinate measuring machine calibration) which clearly indicates how traceable measurements can and can not be achieved with a coordinate measurement machine. Over ISO-standards these documents have the great advantage that they are readily available from the internet ( The strict separation between national institutes and accredited laboratories is somewhat artificial where primary standards can be realised independent. There are examples of national laboratories which calibrate dial gauges and vernier calipers, e.g. in order to earn some money or when there is no alternative. On the other hand there are accredited laboatories with facilities where some national laboratories can be jealous of. E.g. the Eindhoven University of Technology has an accreditation for the calibration of laser interferometer based on a home-built primary length standard, with a lowest uncertainty in length-measuring capability of 0.5 nm. The Mitutoyo corporation in Japan has made its own laser-based red and green primary standards and has about worlds best calibration facility for line scales. Still their certificates are not automatically recognised in Europe. For this reason they have a separate acceditation from the Dutch organisation (NKO) for the calibration of gauge blocks, also by interferometry, with an uncertainty comparable to those shown in Figure 1. On the other hand, the calibration facilities of several national metrology institutes (NLP-UK, laboratories in Sweden, France and Denmark) are accredited by their own national accreditation organisation. In the field of electrical standards there are firm debates where companies have made (or bought) their own quantum-based standards and question the necessity to be traceable to a national standard. 3 Harmonisation in approaches 3.1 Developments in standardization Within ISO, in the field of written standards, most of the gaps and contradiction are being treated in the framework of the Geometrical product specification masterplan [1]. All standards concerning dimensions and dimensional metrology are categorized according to the geometrical characteristic of the feature (e.g. radius, angle) and a chain-link which ranges from the product documentation-definition of tolerances to measurement equipment requirements, calibration requirements and measurement standards. All new standards are considered with this masterplan in mind, thus avoiding and taking away contradictions between standards which could (and can) easily be found. An essential standard is the ISO [2] which implies that the uncertainty and traceability which is about generally agreed for measurement standards (see section

5 3.2) has to be applied for all industrial measurements also. This is a bit like selling the skin before the bear is caught; the matter of uncertainty estimations is just about solved for calibration instruments, which can be classified and generalised, and now the same demand is put on any industrial product measurement which, in general, can be classified and generalised hardly so that, in fact, a separate uncertainty estimation for each separate product must be made. A way to solve this may be the virtual instrument concept Harmonisation in the estimation of uncertainties Following some developments in the early eighties, in 1993 the bible of uncertainty estimation appeared: the GUM [4]. With the wide international acceptance of this standard (among others both ISO and BIPM supported this document), there was no choice but following this document. Various attempts are made to extract simplified (or: practical applicable) versions from this document. Within the European Accreditation organisation, the former document 19 was replaced by EAL-R2, now called EA-4/02. Within ISO, for the uncertainty estimation in measurement and calibration, ISO/TS :1999 [5] appeared. The general approach is, in very brief: the estimation of a standard uncertainty (standard deviation) of each contributor to the uncertainty estimating the influence of the uncertainty of each contributor to the uncertainty of the parameter which is measured adding up all independent influences quadradically multiply the result by a factor of 2 to obtain a coverage factor of 95% Of course, still thick books can be written on this matter, but this basis will not easily be left. 4 New developments in research In this section, a small selection of new developments is given, where the Eindhoven University of Technology is involved in. This overview is not complete and topics already covered elsewhere in this conference are omitted. 4.1 The virtual instrument concept As mentioned in section 3.1, hardly general rules can be given for the uncertainty in the measurement of a product. To overcome this problem, the virtual instrument concept was developed. This concept means that a measuring instrument, except giv-

6 6 ing a measurement value, gives also an estimation for the uncertainty in that value. This means that for each measurement a complete uncetrtainty budget is made automatically. This is done by generating a number of virtual instruments with deviations the actual instruments could have, simulating the same measurement with many virtual instruments and make a statistical analysis of the results. This was done at first with about the most difficult but also most urgent case: a coordinate measuring machine [6,7]. Here the problem is both most complex and the industrial need maximal: CMM s are used because they are universal and can measure anything and so any product; where, on the other hand, the estimation of the uncertainty for any product is most difficult. The virtual CMM is simulated by modified Monte-Carlo techniques. Another case where this approach is necessary is the field of roughness measurements [8]. Here the situation is depicted in table 1, which can be made up similarly for CMM-measurements. typical uncertainty in Ra Wavelength calibration of interference microscope Calibration of groove-depth standard Calibration of standardroughness tester Calibration of type C or D roughness standard Calibration of workshop roughness tester Measurement of workpiece roughness National standards laboratory < 0,01% 2 nm + 1% Accredited laboratory 5 nm + 2% 5 nm + 2% 5 nm + 4% 5 nm + 5% Workshop laboratory 10 nm + 8% 10 nm + 8 % Table 1- Traceability chain and typical uncertainties in the field of roughness measurements?? As any workpiece is different, each roughness structure will be different too. With the virtual roughness tester developed in our laboratory, an uncertainty budget is made for each measurement. An example is given in table 2. The table shows that many uncertainty parameters are taken into account which all work out different on any roughness parameter and which will be different for any specimen.

7 7 Nominal uncer- u: u: u: u in Ra tainty Rz RSm Rsk 13,8 nm u 120 µm 8,4 µm 1,26 x-axis 1 1% 0 0 0,08 µm 0 z-axis 1 1% 0,14 nm 1,2 nm 0 0 λ c 80 µm 2% 0,06 nm 0,03 nm 0,13 µm 0,007 λ s 2,5 µm 20% 0,96 nm 10 nm 0,61 µm 0,051 F 1 mn 50% 0,01 nm 0,06 nm 0 0,0003 Radius 2 µm 50% 0,26 nm 1,7 nm 0,02 µm 0,093 sample rate Noise offset inhomo geneity 0,05 µm 100% 0 0,15 nm 0,003 µm 0, % 0,05 nm 0,4 nm 0,2 µm 0, ,2 nm 2,3 nm 0,25 µm 0,094 Total (1s) 1,0 nm 10,5 nm 0,71 µm 0,142 Table 2 - Uncertainty budget of a Rubert-Song roughness standard. This is somewhat different from a Monte-Carlo technique, although in the end the result is (and should be) the same. 4.2 Displacement measurements with nanometre accuracy Semiconductor wafer illumination machines require positioning which is accurate down to the nanometer-level. The accuracy of laser interferometer systems approaches this level and new systems are in development. Also, capacitive and inductive probes claim to measure with this accuracy. Figure 2 depicts the calibration set-up of the Eindhoven University of Technology [9]. Its working principle is as following: one of the mirrors of a Fabry-Perot cavity is displaced, which gives the displacement to the object to be calibrated. A slave-laser is stabilised on the Fabry-Perot cavity.

8 8 iodine stabilised laser mixing prism avalanche photodiode mirror driver frequency counter moving mirror slave laser control loop F-P interferometer photodiode sensor bending prism Figure 2. Set-up for displacement calibration with nm-accuracy. When this cavity changes its length, the slave-laser changes its frequency. This frequency change is measured by comparing the slave laser frequency with the frequency of a primary length standard: the I2-stabilised He-Ne laser. As an example the calibration of a capacitive sensor is shown below in figure 3 4 m n n i o t i d evia y r it L inea Fabry Perot mirror displacement in µ m Figure 3 result of the calibration of a capacitive sensor

9 Figure 3 clearly shows that repeated measurement reveal the nanometre-reproducibility of both the capacitive prove and the calibration system. The non-linearity of the capacitive probe is clearly indicated Accurate 3-D mechanical probe systems. Where coordinate measuring machines become more accurate and smaller [9] there is an increasing need for accurate probe systems. Here an overview of a few designs is given. Figure 4 gives the design for a 2-D-probe as it was applied by Pril [10]. A Laser Diffractive Grating Unit (LGDU), as used in comon CD-players, is used for the detection of the probe movement in the x- and z- direction. A typical uncertainty of 10 nm in both directions was shown. Figure 4. Schematic of 2-D probe with nanometre-resolution Figure 5. Schematic of 3-D probe with integrated resistive sensors

10 10 The same system of elastic hinges is used in a different set-up which is etched in silicon. It is shown in figure 5. Its real size is about 5 x 5 x 5 mm and here also the typical uncertainty is about 10 nm. A similar design was used by Peggs [11] who used capacitive sensors as the sensing element. The idea to use a grating to split the beam was combined with the probe as it was developed by van Vliet [12] and has lead to the design as sketched in figure 6. Figure 6. Probe design with opto-electronical detection systems An incoming laser beam is split in 4 beams, which are each detected in 2-D on a PSD detector. Calculations show that very high resolutions should be possible. A prototype is under development. 5 References [1] Geometrical Product Specification (GPS) Masterplan. ISO/TR (1995). ISO, Geneva. [2] GPS Decision Rules for proving conformance or non-conformance with specifications (ISO :1998)

11 [3] Mutual recognition of national measurement standards and of calibration and measurement certificates issued by national metrology institutes. BIPM, Paris, 14 October 1999 [4] Guide to the expression of uncertainty in Measurement (GUM), ISO, 1993 [5] ISO/TS :1999 Geometrical Product Specifications (GPS) -- Inspection by measurement of workpieces and measuring equipment -- Part 2: Guide to the estimation of uncertainty in GPS measurement, in calibration of measuring equipment and in product verification. ISO, Geneva. [6] E. Trapet and F. Wäldele, 'The virtual CMM concept', in: Advanced Mathematical Tools in Metrology II, P. Ciarlini et al (ed.), World Scientific Publishing Company, [7] B.W.J.J.A van Dorp, F.L.M. Delbressine, H. Haitjema, and P.H.J. Schellekens, Calculation of measurement uncertainty for multi-dimensional machines, using the method of surrogate data, presented at AMCTM 2000, May 2000, Caparica, Portugal. [8] Han Haitjema and Michel Morel, 'The concept of a virtual roughness tester', in: M. Dietzch and H. Trumpold (ed) Proceedings X. International Colloquium on Surfaces, 31 jan-2 feb 2000, pp Shaker Verlag, Aachen, [9] H. Haitjema, P.H.J. Schellekens and S.F.C.L. Wetzels, Displacement sensor calibration up to 300 µm with nanometre-accuracy with direct traceability to a primary standard of length", accepted for publication in 'Metrologia 36 (2000) ' [10] W.O. Pril, K. Struik and P. Schellekens, Development of a 2D CMM probing system with nanometre resolution, Proceedings ASPE annual meeting, 16, (1997). [11] G.N. Peggs, A.J. Lewis; S. Oldfield Design for a compact high- accuracy CMM, CIRP Annals 48, (1999) [12] W.P. van Vliet, P.H. Schellekens Accuracy limitations of fast mechanical probing, CIRP Annals 45/1, (1996) 11