CMMs (Coordinate Measuring Machines) are measuring instruments widely used in dimensional metrology (fig. 1).

Their measuring principle is that a geometric feature of size (e.g. plug diameter or slot width), or of form (e.g. disk roundness or flatness) can always be evaluated in software as far as 3D coordinates of a sufficiently dense sample of points of the object surface are measured. This paradigm reduces the virtually infinite variety of geometrical measurands - as many as the possible geometries of technical surfaces - to a conceptually single one, the 3D coordinates of points. The complexity of the measurement task is then shifted from the measurement to the software evaluation of the sampled coordinates; the software environment is a decisive feature of a CMM and of its competitiveness.
CMMs span a wide accuracy range, from front–end industrial applications to high accuracy calibrations of geometric standards. They are very relevant in industry, particularly in Italy (fourth producer and seventh consumer of CMMs in the world); however, traceability is still an issue which challenges coordinate metrologists. Today, the best general procedure to achieve traceability is acceptance and reverification testing according to the international standard EN ISO 10360–2. Several laboratories are accredited in Italy by the SIT to do so, to ensure traceability to the highest possible level. In other fields of accreditation, the procedure requires candidate laboratories to prove their competence experimentally by intercomparisons. This is usually not done for CMMs, for the following reasons:

• The circulation of a CMM is not possible due to unfeasible transportation and installation; comparisons must be on-situ on the same CMM;
• The measurand is a CMM performance as defined by EN ISO 100360–2. This standard is admittedly an educated trade off between thoroughness and feasibility. It relies on some freedom left to the purchasing party in determining the actual measurement procedure at the time of testing, to force its counterpart - the seller - to deliver a good machine anyhow. A CMM tested twice according to the standard may produce different test results due to different procedural choices of a party;
• CMMs are sensitive to the environment, even in the presence of software compensation. EN ISO 10360–2 testing prescribes that the environmental specifications set by the manufacturer - usually expressed in ranges - are met, so a CMM tested twice may come to different test results due to different environments.

To overcome these difficulties, a working group of accredited and candidate laboratories was set up in 2005 under the leadership of the INRIM, to investigate how an experimental assessment could be done. It was decided to run a (subset of) the testing on the same CMM, all participants in a row; this meant convening at the same laboratory where a CMM was available, bringing one's own equipment: gauge blocks, step gauges, their supports, thermometers, etc. After two failed attempts, a third successful exercise was carried out in 2007 at the INRIM. Data reduction and analysis took some time and was completed in 2008. To the authors' knowledge, this was the first exercise ever of this type worldwide. The challenge was, in the first place, to set up a scheme and a paradigm:

• Each participant was given half a day to carry out a subset of the EN ISO 10360–2 test; due to the participant numbers, this took a full week's activity;
• Each participant followed its own procedure, using its own equipment and personnel, much like at a customer's site. As a result, gauge blocks (fig. 2) and three different models of step gauges were used: the consistency of the EN ISO 10360–2 across different permitted standards of size was challenged;

• In spite of all attempts to preserve the same measurand for all participants, two main instability sources were identified: CMM repeatability and drift over the week's exercise, and the impossibility to hit exactly the same measurement points in the CMM volume for all participants, due to the different standards of size (nominal values, misalignment and misplacement), so called proximity error;
• These two measurand instability sources were evaluated experimentally. An additional step gauge was permanently fixtured onto the CMM table and measured repeatedly after each participant; the standard deviations of these measurements evaluated repeatability and drift. Further, the same gauge was repeatedly measured at the end of the entire exercise, deliberately misaligning and misplacing it slightly to mimic participants' gauges; the standard deviations of these measurements evaluated the proximity error;
• The reference values were taken as the simple means among participants: in fact, not even INRIM would have been able to provide a better reference value, as the CMM and the environment were the same for all;
• The standard deviations of repeatability and drift, and of the proximity error were added in quadrature to the uncertainty of the mean, to give the reference value uncertainty.

The high accuracy CMM used confined the repeatability and drift, and the proximity error, to a minimum, relevant only for small sizes which suffer from proportional uncertainty components only slightly.
In spite of the success, it was agreed that this exercise was still to be taken as a drill, a sort of final rehearsal before actual experimental assessment.
In conclusion, the comparison of CMM performance verification according to EN ISO 10360–2 was pioneered (fig. 3), a proper scheme and paradigm were set and the most relevant error sources were identified. SIT is now ready for a full experimental assessment of the laboratories' competence, as is usually done in other fields of accreditation.

EN ISO 10360-2:2001 - Geometrical Product Specification (GPS) - Acceptance and reverification tests for coordinate measuring machines (CMM) - Part 2: CMMs used for measuring size.
SIT/Tec-016/09: Accreditamento SIT di verifiche di prestazioni di CMM secondo la UNI EN ISO 10360-2:2005 - Requisiti Tecnici http://www.sit-italia.it/SIT/Documenti/Tecnici.htm