Where to Buy

Norbar Blog


Understanding ISO 6789 – Calibration Laboratories

20 Mar, 2017 | Return|

The 2017 edition of the standard is considerably modified from the 2003 edition.  The five Norbar articles are designed to explain the key differences and how the new standard is intended to be used. They are an overview, not a replacement for studying the standard. If you have questions we will be happy to try and help. Please use the email link at the end of each article.

This third article of five has been written with calibration laboratories in mind.  Further articles for, automotive end users and general industrial users will follow. It will help to first read article one on the general reason for change.


The new edition has the greatest impact on calibration laboratories (simply called laboratories for the remainder of this article). This is because Part 2 is specifically written to provide consistent requirements for laboratories and these requirements will make it easier for accreditation bodies to perform audits that are comparable throughout the ILAC cooperation www.ilac.org

Other organisations such as in-house quality control departments or hand torque tool service centres may wish to use this standard to provide traceable certification of hand torque tools. To reconfirm the statement from previous articles, the only way to produce a calibration certificate in accordance with ISO 6789 will be to follow Part 2. Part 1 only allows a declaration of conformance to be issued. The measurement method is the same, but other requirements are different.

Range of torque tools:

One of the key changes (Part 1 clause 5.1.3) is that the torque range now starts at the lowest marked value and calibration must start at this value.  This applies to both indicating tools of Type I Classes A, B & D and setting tools of Type II classes A, D & G. Tools of both Types in other Classes have a range specified by the manufacturer. Customers who have previously had a 10-100 N·m tool calibrated at 20 N·m will now find that the laboratory needs to calibrate it at 10 N·m.

Calibration system requirements:

In general, the requirements from the 2003 edition are carried into the new Part 1 and are then referenced in Part 2. There are however, several changes that affect laboratories.

The suitability of the calibration system is defined differently in Part 2 (clause 4.3) from the measurement system in Part 1 (clause 6.1). In Part 1 the maximum measurement error of the torque measurement device shall not exceed ¼ of the claimed maximum permissible relative deviation of the torque tool at each target value. In Part 2 the relative measurement uncertainty interval W ’md of the measurement device shall not exceed ¼ of the expected maximum relative uncertainty interval of the torque tool W ’. The measurement device must have a valid calibration certificate issued by a laboratory meeting the requirements of ISO/IEC 17025, or the national standards laboratory. This ensures that the certificate is internationally traceable with a suitable evaluation of measurement uncertainty.

The application of load is more clearly defined in the new edition (Part 1 clause 6.2.1) and it is emphasised that the calibration device must allow the tool to move to prevent “parasitic” loads or moments. The new patented Norbar counterbalance system achieves this, but older loaders may exert side or end loads on the tool.

Also, the time to attain the last 20% of the torque application during testing has been more tightly defined (Part 1 clause 6.2.4) according to the size of the torque tool. it does make sense that small torque tools need less time to complete the last 20% of their target torque.  That does however make it harder to measure whether the load application is within specification. The new Norbar Torque Wrench Calibrator (TWC) learns about the characteristics of the tool mechanism and adjusts the motor speed throughout the calibration to optimise the loading cycle. Torque screwdrivers have a minimum and maximum time requirement because they are often speed dependent. Achieving a duration of between 0.5 and 1.0 seconds for the last 20% of the applied torque is challenging.

Measurement error in place of deviation

The second article explained that in Part 1, the method of calculating the relative deviation observed at any given target torque has been returned to the formula used in the 1992 edition.

In Part 2 the ISO defined term of relative standard error is used to provide consistency within calibration laboratories. This is calculated using the same formula as the 2003 edition.

There is the possibility of confusion here. For example, an observed reading of 104 N·m at a target value of 100 N·m is defined as being +4% in Part 1, but -3.85% in Part 2!

To most torque tool users, if the tool is delivering 104 N·m at a set or indicated value of 100, it is intuitive to say that the tool is over-torqueing by 4%. However, the ISO standard regulations require that we use standard error in calibrations and so, using the appropriate formula, when we observe 104 N·m at a target of 100 N·m it means that the target is 3.85% low. Sorry, but that is the way it is.

This makes it very difficult to compare Part 1 and Part 2 results. Again, the logic is that a simple torque measurement is used create a declaration of conformance with Part 1, while the more demanding requirements of uncertainty budgets in Part 2 are needed to create a calibration certificate.

Evaluation of uncertainties:

The biggest change from the 2003 edition has to be the calculation of specific elements of uncertainty. These elements have been evaluated and selected by the drafting committee and are already used by many laboratories.  They do add significant time to the calibration of a torque tool, but they are a key element in claiming a traceable calibration has been performed. From our experience it takes approximately one hour to perform a calibration including the evaluation of uncertainties. It is therefore clear that the prices charged by some laboratories will need to increase in order to cover the cost of performing a calibration according to ISO 6789-2:2017.

The requirements for resolution are clearly defined and should not take significant time to evaluate.

Reproducibility evaluation time has been minimised by using results for the lowest specified torque value only, rather than requiring results at the 60% and 100% of maximum torque values. This slightly compromises the accuracy of the uncertainty budget at the two higher torque values, but does reduce the time taken to generate the budget.

Ratchets and removable square or hexagon drives can influence the result, depending on their tolerances from new and on wear that has occurred during use. The evaluation is therefore an essential element for both new and old tools.

One of the largest causes of uncertainty comes from inappropriate adaptors between the torque tool and the measurement device. Using commercial square driver converters, for example from 3/8” female to ½” male, will significantly increase the uncertainty of calibration.

Some torque wrenches are more “length sensitive” than others. The evaluation of uncertainty due to the force loading point is therefore important. Perhaps it is not obvious that the observed torque from the measurement device decreases as the force loading point moves towards the end of the handle. Hold many torque wrenches right on the end of the handle and the applied torque will be significantly lower than expected.

All of the above evaluations need to made prior to actually taking the readings that will appear on the calibration certificate. They are the Type B uncertainties due to the torque tool. There are two ways of avoiding this process every time.

Where a laboratory calibrates a significant volume of a specific model of torque tool, the standard allows the results of ten or more examples of that model number to be statistically combined and inserted in the uncertainty budget of future tools of that model. There still needs to be periodic complete evaluations of the uncertainty budget in case the performance of that model changes over time. It is also clear that the performance of new torque tools will perform differently from older used tools and therefore the data for old and new should not be combined. Laboratories therefore need to keep extensive records of the conditions and performance of all torque tools evaluated.

The manufacturer of a torque tool or a third party may provide the data, but care must be taken to ensure that the data can be replicated, by periodically performing the uncertainty evaluation and checking that the values being used are comparable to experimental data.

The repeatability, which is the only type A uncertainty considered due to the torque tool, is calculated from the calibration readings.

Finally, the calibration certificate for the measurement device will reveal the necessary information for inclusion in the calculation of both relative standard measurement uncertainty and the relative uncertainty interval.

Relative uncertainty interval

This may be a new concept to some laboratories. It defines the interval within which a reading could fall. It adds together the mean value of measurement error plus the relative expanded uncertainty plus the maximum error of the measuring device. The resulting numbers can be rather large and it is not unusual to see a torque tool with claimed maximum permissible error of 4% (under the Part 1 definition) having a relative uncertainty interval of 6%.

Calibration certificates

The documentation requirements of ISO 6789-2:2017 are also expanded from ISO 6789:2003. Where the laboratory is already working to ISO 17025 there will be some additional information items to add to the certificate.

For laboratories that do not work to ISO 17025, the certificate content is quite different to the simple certificate often in use now. There is a requirement to detail the items used, including interchangeable end fittings. Laboratories not working to ISO17025 must also detail the measurement device traceability. Finally the relative expanded uncertainty and the relative measurement uncertainty interval must be stated for each calibration point.

In summary, there is a lot in the new standard and if you calibrate hand torque tools, you need to purchase both parts and study them. If you have questions we are happy to help where we can. Please email ISO6789@norbar.com

Neill Brodey

Member of ISO working group on ISO6789