DPCNews 029 - Quality of some calibration services

October 2010

1- Introduction

The quality of some calibration services brings us to draw the attention of the calibrating service companies and their customers on some questionable practices from the Quality Assurance point of view.

In many cases, these practices are only the consequence of a lack of requirements in the standards. It would be high time when these specify the minimum figures/data of needed requirements, while knowing that the customers may always ask their calibrating service companies for more if these companies can technically do it, if it is justified by the customer’s specific application, and if ... the customer agrees to pay the price for that!

Let us come back to the "justification according to the customer’s specific application". The proverb according to which "he who can do more can do less" is not always justified for measuring equipment:
is it useful to be able to measure temperatures within about a 1/10°C for PT purpose? It requires that the reference thermocouple be calibrated within about 0.025°C, taking into account the tolerance calculations and the deviations when going from a reference unit to the working unit. Is it reasonable? Anyway, this comes at a price!

The customer, the user, you, in fact, has to define what is useful for him (her) and to write it in a procedure!! Once again, if it is written, the auditor has no margin enough to tell you that he does not agree with your data. If not written ...

2- Certificate of verification of digital radiometers and luxmeters

The ISO 3059:2001 standard, currently under systematic revision, states in its chapter 7 that ‘‘the calibration of irradiation and illuminance meters shall be verified’’.

This statement is far too much vague and may lead to abuses ... in the direction of the cheapest tender, for instance!

A certificate of verification is certainly the most suitable document.

We all have already seen certificates of verification on which only one measuring result was displayed, sometimes right in the middle of the page to take up all the available space.

When asked whether the meter has a linear response, the services company answers, through a technician, that as it is a digital meter, its response can be only linear. How many times did we get such an answer from unknowledgeable people? Every time we are flabbergasted!

To prevent such a problem, the standard should comprise something as a requirement stating that the certificate of verification shall display three figures of (UV-A) ultraviolet irradiance measurement and three figures of (visible light) illuminance measurements at least all over the range of figures used “on the field”, plus a verification of the zero, of course! That means that, even if, say, a luxmeter has a capability to measure 20,000 lx, if this luxmeter is used in a UV-A booth to check that the visible light is less than 20 lx, it is worthless to calibrate it above, say, 50 lx!!

While the ISO 3059:2001 standard states specific (UV-A) ultraviolet irradiance and (visible light) illuminance figures, it does not give any tolerance limit for these figures.

Further, there is any clue neither for the measurements' accuracy nor for the measurements' uncertainty.

Of course, we agree that this is not as important as it is for other meters such as UT precision thickness gauges.

Furthermore, the calibration setup device of recently designed radiometers/luxmeters is inside the interchangeable sensors, contrary to what it was in the previous generation. Therefore, calibrating the display unit and the sensor as an alone system is no longer useful, at least for these new meters.

According to the standard, the UV-A radiation used to calibrate sensors shall be a monochromatic 365 nm radiation, i.e. 365 ± 0 nm.

The French Laboratoire National de Métrologie et d’Essais (LNE/LCIE - Bureau VERITAS), equivalent to the American NIST, uses monochromatic filters: the 365 nm wavelength is only slightly "polluted" by other wavelengths ..., very far from the industrial world.

Further, keep in mind that the emission spectra of a mercury vapour source, of a luminescent tube or of a xenon source are not similar, which is not without any effect on the measurement.

Unless when using a laser beam, it is somewhat impossible to get a perfectly monochromatic wavelength: a utopian request! Therefore, the best which can be achieved is to get the technically possible narrowest bandwidth at a cost in relation with the customer’s needs. Measuring the UV-A emission of a distant star with a satellite-born radiometer has nothing to do with measuring a UV-A source for a non destructive testing application!

Some light-emitting diodes allow for a 10 nm bandwidth at mid-height, a very good result. The bandwidth on the calibration benches of some service companies is even smaller.
We suggest writing in the standard that the UV-A radiometers calibration be carried out using a radiation centred on the 365 nm (360 < λ < 370 nm) wavelength with a 10 nm maximum bandwidth at mid-height (cut-off).

However, some service companies use a prism to get the 365 nm radiation. It means lighting the sensor with a gradation of (UV-A) ultraviolet wavelengths: 360 nm on one side, 365 nm in the middle and 370 nm on the other side!!! This does not comply with the standard.

Some Quality Assurance Managers ask their service companies to supply them with their calibration procedure. Service companies agree to show them on site, but generally do not send a copy, as they see the procedure as a part of their expertise.

3- Verification of tangential magnetic field meters

The ISO 9934-3:2002 standard does not state any periodic verification of tangential magnetic field meters.

The only requirement is in § 9.3.2: ‘‘measurement accuracy better than 10%’’.Nothing more!

Here again, we all have already seen certificates of verification with only one measurement figure.

Here again, the same question about the linear response of the meter leads to the same stupid answer.

To toughen the Quality Assurance of magnetic field meters, the standard should specify that tangential magnetic field meters shall be verified at the frequency recommended by the manufacturer, using a reference meter whose calibration is traceable to National or European standards. This periodicity shall not exceed 12 months. A certificate of verification shall be supplied and give the results of three magnetic field measurements at least all over the range of figures used on parts, plus a verification of the zero, of course! To prevent what is detailed in the “General Remarks” chapter, it is also worthwhile to limit the number of steps between a National or European reference meter and the meter under verification/calibration.

When using a Hall effect probe to measure tangential magnetic field, you may have noticed two different phenomena:
• By rotating the probe 180° around its main axis, we often get a magnetic field figure different from that before the rotation. This is due to a non-symmetric position of the sensor in the probe. Nevertheless we know a manufacturer who does it accurately. 
• Tilting the probe when on the surface leads also to errors which may be tremendous in some cases:
- If the field lines are not tangent to the surface.
- If the surface of the sensor in the probe is not perfectly perpendicular relative to the surface.

If a Helmholtz coil is used to calibrate the probe, this problem is almost undetectable as the location of the probe in the coil as well as the perpendicularity of the magnetic field to the sensor are perfectly reproducible.

The sensor exact positioning in the probe has no influence on the calibration if the magnetic field is very consistent in the area where the sensor is put. Using a Helmholtz coil assures of a very consistent magnetic field in the centre of the system!

However, here, as for many other calibrations/verifications, there is a huge difference between the way that we use the sensor in an industrial background and the way it is calibrated. For example, the tangential magnetic field on a part’s surface is far from being as homogeneous as the field inside a coil. The lines of magnetic field cannot be compared to parallel lines. This is very dependent on the part shape and the magnetization method. Even if the sensor is small - the Hall effect sensor in a probe is only very few square millimetres large - we cannot say the magnetic field which runs through the sensor is homogeneous, as in a coil designed for that purpose. As a further drawback, depending on units, the sensor is more or less far from the tip of the probe (the surface put on the part). The perfect sensor (infinitely small and in close contact with the part) does not exist. Then, the difference with calibration conditions is large ... and varying from a part to another one, due to very different shapes on which a measurement is performed.

Add above and beyond the fact that the calibration in a coil is carried out in a homogeneous field and is not related to the distance sensor/tip or to the sensor’s size. That is why two different sensors, even after calibration on the same bench, by the same operator, may display quite different figures of a magnetic field measured on the same part, all other conditions being the same. Further, the mechanical wear of the probe’s tip, which, little by little, moves the sensor closer to the tip, will lead to drifts not brought to light by a calibration in a coil.

Staying attuned to tangential magnetic field meters, there is a point that we have not yet raised. Verifications/calibrations are always carried out using sinusoidal (AC) signals, or half-sinusoidal, when half wave rectified direct current (HWDC) is used.
True, magnetic fields used for MT are AC or HWDC for 99%. Some tangential magnetic field meters are supplied with a certificate of calibration, which gives only figures got from a calibrated permanent magnet, traceable to National standards. How can one be sure to guarantee that these meters are calibrated for the full useful range? Some manufacturers suggest to check the unit using a reference permanent magnet previous to any measurement. Some units used for a daily check of the probe are based on permanent magnets, thus true DC fields. Several magnets, of different magnetic flux densities, or a single magnet whose distance to the probe may be adjusted to preset values, may give some warning of the mechanical wear of the probe. However, these pieces of equipment do not allow for knowing if the electronics of the field meter runs as it should be: for example, are the 100 or 1,000 measurements per second well achieved? A true DC field, by definition, gives identical peak and RMS values, which do not vary 50 times per second (60 times in the USA and Canada). These permanent magnets are a piece of information, for tangential field meters.

This is different for probes measuring residual fields, which are true DC fields. When measuring residual fields, the information given by these permanent magnets is more relevant. A good example comes from MT inspection of needle or ball bearings, which are later put close to navigational instruments in aircraft: the residual magnetization shall be less than 0.1 mT. In this application, calibrating the meter every six months or every year is not enough to be sure that the meter is right. For that purpose, one may use reference magnets of 900 A/m (the lowest figure for an acceptable sensitivity), 1000 A/m and 1500 A/m to check the consistency of the values displayed on the meter. So-called “Zero Field Chambers” are used to check the zero of meters. Keep in mind that anywhere we are, we are in the Earth magnetic field, a true DC one. This must be taken care of when measuring very low magnetic fields or flux densities.

Thyristors are widely used on fixed or mobile MT current generators for current control. Thyristor "chops" the sine curve after a while (100 or 120 times per second), the current drop to zero being within some milliseconds. However, it sends intensity "peaks" which can be very high. A slow time response of the meter cannot help in measuring these peaks, which will be "averaged" when displayed. Furthermore, the wave shape given by a thyristor is very far from a sine curve as soon as the current does not run 100% of the time - i.e. when the unit is not used at its maximum rate, which is almost always the situation in workshops.
Therefore, verification and calibration are carried out in "perfect conditions":
• The sensor in a quite large solenoid so that it “sees” a perfectly homogeneous field.
• The use of perfectly sinusoidal waveforms.
• The probe not in contact with a part, hence no way to be aware of the abrasion (mechanical wear) of the probe due to numerous contacts with parts.

However, the meter is used in completely different conditions!

We have no technical answer to these concerns, but we think it useful to point them out.

4- General remarks

We also have to raise another problem: calibrations/verifications "in cascade".
Let us imagine a company with 3 or four plants, each with one, two or even three PT lines. The Central Laboratory is "the reference" for radiometers calibrations/verifications.
This laboratory sends its "reference radiometer" to a service company, which sends it back to it, calibrated with a total uncertainty of 8%, at 1,500 µW/cm². This is a very good service provision.
The Central Laboratory then requires every plant to send one radiometer, which will be “the reference” for this plant. The laboratory performs a “calibration”, giving no tolerance...that the laboratory would be unable to calculate! Just for your information, the UV-A source is a standard 100 W mercury vapour bulb; we know that such a source is “unstable” in that its emission quickly varies (even within a second); nevertheless, this peculiarity is not a problem when carrying out inspections.
Every plant now has its own reference, which is to be used to “calibrate” the radiometer of every PT line. How is this “calibration” performed? Right, you have understood.

Let us give you a true example: three radiometers coming from three different process lines, each in a different plant are sent to the service company; in fact, a technician from this service company, a bit curious about the comparisons, got the agreement of the users to compare the “calibration” results.
The radiometers, checked on arrival, displayed a more than 50% difference between the lowest reading and the highest reading! Nevertheless, every unit comes with a certificate stating that it has been compared with an in-house reference unit, n° xxx, and that it is within a 10% range against this in-house reference.
By the way, 10% ... what does it mean?
Further, every calibration step multiplies the deviations and then, it is a given to see such differences! Every user believes that his unit is "accurate" and that he works in the right viewing conditions.
What can an auditor do when faced with such a situation?

Another point: if the range of a meter far exceeds the user’s need, there is no need to calibrate the unit on the overall range. Let us go to an example: a user has a luxmeter with four scales: 0-20 lx, 0-200 lx, 0-2,000 lx and 0-20,000 lx.
If this luxmeter is used only to measure illuminances lower than 20 lx, it is useless to verify or to calibrate the ranges which go beyond 20 lx. If the measured illuminance is above 20 lx when using the 0-20 lx scale, if the 20 lx figure has been calibrated, the unit will display an “over-range” warning, or automatically go to the 0-200 lx scale; even if it is not calibrated, the user knows the reading is above the allowed maximum of 20 lx. That is enough to check viewing conditions. A corrective action is required.
Should this luxmeter be used for inspections requiring an illuminance of 500 lx, or even 1,000 lx minimum, then the 0-2,000 lx scale is sufficient.

As far as we DEMONSTRATE that indeed the mandatory minimal figure is used, it is useless to calibrate the upper scale! However, of course, the limits MUST be written in the test report: which scale(s) has (have) been calibrated, or if there is only one scale, which range has been calibrated: between zero and, for example, 1,500 lx.
Is it that important that the meter displays with accuracy a 1,800 lx reading measured on the working area, even if it has not been calibrated for this figure? We know for sure illuminance is above 1,000 lx, if 1,000 lx is the mandatory minimum. This is the important point.
Another way to lower verification/calibration costs: do only what is purely needed for the user’s application.

Conclusion

In order to improve the quality of some calibration/verification services, it would be necessary, when the above-mentioned standards are revised, to specify some requirements such as the ones we mentioned. It would be a good idea to think of how the meters are used on site. For UV-A radiometers and luxmeters, the emission spectrum of the source or the sources is of the utmost importance for the consistency of results. The response curves of these meters to the wavelengths are not similar. How can one be confident in the technical data sheet of a cheap luxmeter’s manufacturer who, as a response curve of its meter, gives a -small- copy of the response curve of the standard eye according to the CIE (International Commission of Illumination) … curve that even very expensive units are almost unable to duplicate?

For sure, there are many different UV-A sources, and even visible light sources, with very different emission characteristics. Nevertheless, the marketing of 365 nm diodes, or 450 nm diodes, usable in industrial conditions, should simplify the situation within few years.
This is why it is SO IMPORTANT that the company which carries out a radiometer or a luxmeter calibration states on the verification/calibration report which reference source has been used.

This example of lighting sources can obviously be applied to many other meters. It is therefore, a good idea to check how the unit will be used, which is the accuracy that is really needed. The more we are requiring on accuracy and tolerances, the more the reference unit which is used for calibration/verification shall be accurate, therefore, expensive. The service is more expensive. Is it really useful? It is a question worth some thousands or dozens of thousands euros or dollars; the answer is “THINK FIRST TO SAVE A LOT OF MONEY”.

A reminder of vocabulary according to:

• French standards for measuring instruments traceable to National standards: AFNOR NF X 07-010 and AFNOR NF X 07-011 standards.
• ISO 10012:2003 - Measurement management systems -- Requirements for measurement processes and measuring equipment

Calibration:
It is only a recording of deviations, without any judgment (about conformity) neither adjustment.
It is the user’s duty to decide whether the unit meets his needs and to draw the correction curve.

Verification:
Verification makes one sure that deviations between figures given by a reference unit and those given by the unit under test all are lower than the maximum accepted deviations as defined in a standard, a procedure or any specific document.

Result of such a verification is:
• Put back to service.
• Adjusted and put back to service.
• Repaired and put back to service.
• Classification downgraded.
• Decommissioning.

Therefore, after verification, and adjustment if needed, someone, generally the organism which verified the unit, decides to declare that the unit meets the specified requirements or not. Usually, a correction curve is not useful, as the unit is adjusted if needed.

Reference

• ISO 3059:2001 Non-destructive testing -- Penetrant testing and magnetic particle testing - Viewing conditions, International Organization for Standardization, Geneva, Switzerland, 2001.

• ISO 9934-3:2002 Non-destructive testing -- Magnetic particle testing -- Part 3: Equipment, International Organization for Standardization, Geneva, Switzerland, 2002.

We, Pierre CHEMIN and Patrick DUBOSC, welcome any comment, any idea. If you have some examples you would like to see discussed here, please give us all the useful indications. If you require confidentially, we would modify locations, names and some parameters to prevent any traceability.
Nevertheless, we are convinced that our site may be a kind of surge-valve: the topic is NOT to target this company, or that auditor; but it is always to make users think, to make them ask themselves, or others, the right questions.
We may also give advice, once again on a confidential basis if needed: please, feel free to ask questions, to document our data basis: about Material Safety Data Sheets (MSDS), about environment, a chemical name you don't understand, a Penetrant process you have heard about, etc.
We have plenty of examples, some being out of all the specifications/standards, which led to the discontinuities detection, when the "current, normal, processes" prevented discontinuity finding.