Torque wrench calibration: how to calibrate a torque wrench (ISO 6789 procedure)

Torque wrench calibration compares the torque a wrench delivers (or indicates) with a calibrated reference, usually a torque transducer, at several points of its range, and records how far off it is. The international procedure is ISO 6789-2:2017: test at the lowest value of the range (usually 20 % of maximum), at 60 % and at 100 %, take five readings at each point, and compare them with the wrench’s permitted deviation, typically ±4 % for click wrenches with a maximum above 10 N·m (ISO 6789-1, clause 5.1.5). At home you can check a click wrench with a known weight hung at a measured distance, and adjust it if it is far off. That check is useful, but it is not a traceable calibration.

This page covers both: the home check and what it can and can’t tell you, then the full procedure a lab or an in-house calibration room follows, with a worked example, the records to keep and how often to recalibrate.

Can you calibrate a torque wrench yourself?

You can check it yourself, and many click wrenches have an adjustment screw you can turn. Whether the result counts as a calibration depends on what you compare against and what you record:

The check is worth doing. The mistake is to adjust a wrench at one point and call it calibrated: a wrench that is right at 50 % of its range can still be out at 20 % or 100 %.

How to check a click wrench with a weight

Torque is force times the perpendicular distance from the center of the square drive. Hang a known weight from the handle and you know the torque at that point:

1. Clamp the square drive (or a socket on it) in a vise so the wrench is horizontal. Only the drive should touch the vise.
2. Mark the load point on the handle, usually the center of the grip, and measure the distance L from the center of the square drive to the mark.
3. Weigh the weight on a scale you trust. A bag of sand or a jug of water works if you know its actual weight.
4. Calculate the setting. In US units: torque in lbf·ft = weight in lb × L in inches ÷ 12. Metric: torque in N·m = mass in kg × 9.81 × L in meters. Example: 20 lb at 15 in = 25 lbf·ft; 10 kg at 0.5 m = 49.0 N·m.
5. Set the wrench, exercise it a few times, then hang the weight from the mark and lower it gently. It should click just as the full weight is applied. If it clicks before, it releases low; if it doesn’t click, it releases high.
6. Repeat three or more times, and at a second weight or distance near the low and high ends of the range you use.

To convert, 1 lbf·ft = 1.356 N·m and 1 lbf·ft = 12 lbf·in (NIST SP 811, App. B.8).

How good is this check? It depends on three things you control:

• The angle. If the rope or the pull is not at 90° to the handle, the torque is reduced by the cosine of the angle error: about 0.4 % at 5°, 1.5 % at 10° and 6 % at 20°. Keep the wrench level and the rope vertical.
• The length. A quarter of an inch off on a 15-inch length is a 1.7 % error in torque. Measure from the center of the drive, not from the edge of the head.
• The weight. Every percent of error in the weight is a percent of error in the torque.

Add them up and a careful weight check is good to a few percent, which is the same size as the ±4 % tolerance you are trying to check. That is why it can catch a wrench that is badly off, but it can’t prove that a wrench is within tolerance.

Torque wrench types, classes and tolerances

ISO 6789:2017 replaced the 2003 edition and comes in two parts (ISO 6789-1, foreword):

• ISO 6789-1 covers design and manufacture: classes, tolerances, conformance tests, and the declaration of conformance the manufacturer supplies with the wrench.
• ISO 6789-2 covers calibration: the method, the requirements for the calibration system, the evaluation of measurement uncertainty and the content of a traceable calibration certificate.

The 2017 edition dropped the factory “calibration certificate” because manufacturers’ certificates had not contained enough information to be traceable (ISO 6789-1, foreword, items g and h). Both parts are being revised; the 2017 editions are the current ones.

The standard sorts hand torque tools into Type I, indicating (they show the torque: beam, dial, digital) and Type II, setting (they release at a set value: click and break-over wrenches, preset tools), each with classes (clause 4). The class and the maximum torque of the tool set the permitted deviation (clause 5.1.5, Tables 3 and 4):

The two columns refer to the maximum torque of the tool, not to the value set: a 20–100 N·m click wrench is held to ±4 % at every setting. The limits apply to every reading, in each direction the wrench is made to work. If the manufacturer claims a tighter limit, the tighter limit applies (clause 5.1.5). The range starts at the lowest marked value, not at zero, for graduated wrenches (clause 5.1.3); below it, the wrench is not specified.

In the US, many manufacturers rate click wrenches to ASME B107.300, the American national standard for hand torque tools and torque testers (current edition 2021). Their product pages commonly state ±4 % clockwise and ±6 % counterclockwise (for example, Harbor Freight’s ICON 20–100 ft-lb click wrench, citing the 2010 edition). ASME B107.300 includes a nonmandatory appendix that correlates it with ISO 6789 (ASME B107.300-2021, table of contents). Use the tolerance on your wrench’s declaration or specification sheet, and write down which standard it comes from.

What you need to calibrate a torque wrench

• Reference standard

  A torque calibration system (transducer, loader and display) with a valid calibration certificate

  From an ISO/IEC 17025 calibration lab or a national metrology institute (Norbar, Calibration Laboratories; Torque Lab). Shop testers are fine for checks.

  ISO 6789-2, Annex C covers the calibration of the measurement device

• Accuracy ratio

  Calibration system relative uncertainty interval no more than ¼ of what is expected of the wrench (in practice, ¼ of its tolerance)

  For a ±4 % wrench, the system must be within about ±1 %. Check yours with the TUR calculator.

  ISO 6789-2, 4.3 (Norbar; Torque Lab; DKD-L 10-1 item 11)

• Loading

  Force at 90° to the handle, at the marked load point, with no side or end loads

  The wrench must be free to move as it turns. A motorized or geared loader gives a controlled rate; the standard sets a minimum time for the last 20 % of the torque, which grows with the torque value.

  ISO 6789-1, foreword d) and e); Norbar; Sushma Industries

• Adapters

  As few as possible, identified, and always mounted the same way

  Adapters add uncertainty. DKD suggests marking a fixed mounting position.

  DKD-L 10-1, 3.1 item 4; Norbar

• Environment

  18 to 28 °C, stable within ±1 °C during the calibration; humidity below 90 % RH

  Let the wrench stabilize in the room first. Record the conditions.

  ISO 6789 (Norbar; Torque Lab; Calibrate)

• Paperwork

  The wrench declaration or specification, your procedure and a record form

  You need the type, class and tolerance before you start. See how to write a calibration procedure.

The calibration system needs its own calibration. DKD’s guideline for torque wrench calibration devices asks for recalibration after 26 months at the latest, and after any overload, repair or mishandling (DKD-R 10-8, §6.2). It excludes fixed-length beams and discs as transfer standards for calibrating those devices, and it does not consider devices that only capture peak values capable of being calibrated under the guideline (§1). Your accreditation body or customer may set other rules.

Test points and number of readings

ISO 6789-2 tests the wrench at three points: the lowest value of the specified range (usually 20 % of maximum), 60 % and 100 % of maximum (Torque Lab; UT Metrology; Calibrate). For indicating wrenches and adjustable setting wrenches (Type I, and Type II Classes A, D and G), that is five readings per point, 15 per direction (AWS; Torque Lab). Fixed and non-graduated setting tools follow a different scheme; check the standard. If an older wrench has no defined lower limit, DKD allows starting at 20 %, provided the tool is marked with the range and the certificate says so (DKD-L 10-1, item 8).

Before the readings, the wrench is exercised at its maximum torque in the direction being tested, without recording: three times per direction for Type I and Type II Classes A, D and G, according to two summaries of the 2017 edition by makers of calibration equipment (AWS; Sushma Industries). The 2003 edition asked for one exercise on indicating wrenches and five on setting wrenches (Sushma Industries; Calibrate). The plan for the example wrench below:

Why three points? A setting wrench is a spring, a lever and a release mechanism, and its error can change along the range. A test at one point can’t show that: in the worked example below, the wrench passes at 60 % and 100 % and fails at 20 %.

Step-by-step torque wrench calibration procedure

The sequence below follows ISO 6789-2 as described by the sources above, plus the as-found and as-left logic of any calibration procedure. Adjust it to your wrench type and write your choices into your calibration procedure.

1. 1

   Identify and inspect

   Record ID, model, type and class, range, direction and tolerance. Check the square drive, ratchet, scale or display and handle for wear or damage.

   Damaged: tag out, record as found as not calibratable

2. 2

   Condition

   Let the wrench reach room temperature (18 to 28 °C, stable within ±1 °C). Record temperature and humidity.

3. 3

   Mount

   Fit the wrench to the calibration system with as few adapters as possible, free to move, with the force applied at 90° at the marked load point.

4. 4

   Exercise

   Load the wrench to its maximum in the test direction, three times for indicating and adjustable wrenches (AWS; Sushma Industries). Do not record these.

5. 5

   As-found readings

   At the lowest value, 60 % and 100 %, take five readings each. For a click wrench, set the value, approach it smoothly at a controlled rate and record the reference reading at the release. For an indicating wrench, apply the reference torque and record what the wrench shows.

6. 6

   Evaluate every reading

   Calculate each deviation and compare it with the tolerance for the type and class.

   Any reading outside: record as found out of tolerance, open an out-of-tolerance report

7. 7

   Adjust only if needed

   Use the manufacturer's adjustment method, and only past the threshold your procedure sets.

8. 8

   As-left readings

   After any adjustment, repeat the full sequence and record it as a separate set. If nothing was adjusted, the as-found results are also the as-left results.

9. 9

   Uncertainty, record and label

   Evaluate the uncertainty at each point (or apply your lab's validated values), issue the certificate or record, label the wrench, and leave it as its manufacturer recommends (never wound below the lowest scale value).

Calibration and adjustment are different operations: calibration compares the wrench with the reference, adjustment changes the wrench, and after an adjustment the wrench has to be calibrated again (VIM 2.39 Note 2; 3.11 Note 3). That is why the as-found readings come first.

DKD adds two practical rules for click wrenches: the run-in behavior belongs in the result, so once the initial exercises are done, don’t add preloads within a calibration step and don’t discard readings; and the full measuring program can reuse series for several uncertainty components (DKD-L 10-1, items 10, 14 and 15).

Acceptance criteria and measurement uncertainty

Worked example: a click wrench out of tolerance at 20 %

The example wrench, TW-031, is a 3/8-in click wrench, 20–100 N·m with 0.5 N·m graduations (ISO 6789-1 Type II, Class A, maximum above 10 N·m), so the tolerance is ±4 % at every point. Below are the five clockwise readings at each point, as found and after adjustment. The deviation of each reading is (reference reading − set value) ÷ set value, the Part 1 calculation. The values are invented for this example.

How to read it:

• As found, 20 N·m: all five readings are between 20.88 and 21.10 N·m, a mean deviation of +4.92 %. The limit at 20 N·m is 20.8 N·m (20 + 4 %), so the wrench was out of tolerance. It delivered more torque than set, at the low end of its range. The out-of-tolerance report has to ask what was tightened with this wrench at low settings since its last good calibration.
• As found, 60 and 100 N·m: inside the limit (mean +2.80 % and +2.02 %). A one-point check at 60 % would have passed this wrench.
• As left: after adjustment, every reading is inside ±4 %, with the largest deviation at 20 N·m (+2.75 %).

A real ISO 6789-2 certificate would also give the measurement uncertainty at each point. At +4.92 % the as-found result is outside under any reasonable uncertainty. Results close to the limit are different: whether you can call them a pass depends on the uncertainty and on your decision rule (Global ACI-TECH-1-002, formerly ILAC G8). Our guard banding guide explains the options.

What goes into the uncertainty

ISO 6789-2 builds the uncertainty from seven contributions (AWS white paper; DKD-L 10-1, §3.1):

Measuring all of them takes far more operations than the old standard: AWS counts up to 153 per direction for indicating and adjustable wrenches, against at most 35 under the 2003 edition (AWS; Sushma Industries). Labs don’t have to repeat all of it every time: DKD lets them estimate reproducibility for groups of similar wrenches (10 tools per group), output-drive effects for groups of sockets or inserts (10 per group) and adapter effects once per adapter (10 measurements), and Norbar describes combining results from ten or more wrenches of the same model (DKD-L 10-1, §3.1; Norbar). The certificate then states, for each point, the relative expanded uncertainty and the relative uncertainty interval, which combines the mean error, the expanded uncertainty and the error of the calibration system (Norbar). Norbar notes that it is not unusual for a tool claimed at 4 % to end up with a relative uncertainty interval of 6 %.

Build or check your own budget with the uncertainty calculator.

What to record on a torque wrench calibration certificate

Whether you issue a certificate or fill in an internal record, keep:

• Identification: wrench ID or serial number, manufacturer, model, type and class, range, units and direction(s) calibrated.
• Method: “ISO 6789-2:2017” (plus “in conjunction with DKD-L 10-1” if you use it; DKD-L 10-1, §3.2), or your procedure number and revision.
• Equipment: the calibration system and its traceability, and every adapter or interchangeable part used (Norbar; Torque Lab).
• Conditions: temperature and humidity.
• Results per point: set or reference values, every reading or the mean, the deviation or error, as found and as left.
• Uncertainty: the expanded uncertainty and relative uncertainty interval at each point (Norbar; Torque Lab).
• Decision: pass or fail against the stated tolerance, the decision rule used, and any adjustment made.
• Who and when: technician, date, next due date.

Our guides to calibration certificates and calibration records cover each field in detail. Print a matching status sticker with the free calibration label generator.

How often to calibrate a torque wrench

If you have no interval of your own, the default associated with ISO 6789 is 12 months or 5,000 cycles, whichever comes first, counted from when the wrench is put into use, not from the date on its declaration (Norbar’s Neill Brodey in Fastener + Fixing; UT Metrology; the same figure appears in Norbar’s description of the 2003 edition). Snap-on presents the same period as an ASME recommendation; we have not been able to confirm that in ASME B107.300. Either way, it is a starting point, not a rule.

Recalibrate before the due date if the wrench:

• has been overloaded by 25 % or more above its maximum (Norbar USA; Snap-on);
• has been dropped, repaired or mishandled in a way that may have affected it (UT Metrology; Calibrate);
• gives results that disagree with another calibrated wrench.

Then let your own data set the interval. If wrenches keep coming back in tolerance, extend it; if they come back out, shorten it. Our calibration intervals guide explains the methods.

Storage. Advice differs. Norbar calls the habit of winding a click wrench back to zero wrong: if you wind it back at all, stop at the lowest value on the scale, never below it and never at zero, and it reports production wrenches left at their setting for months without a problem; use and abuse matter far more (Norbar USA). Other manufacturers tell you to wind the wrench down after use. We have found no published figure for how much a loaded spring drifts. Follow your wrench’s manual, keep it in its case, and check it if in doubt.

When a torque wrench fails calibration

An as-found result outside tolerance does not only mean “adjust the wrench”. Every fastener tightened with it since its last good calibration may be wrong, in the direction and at the settings where it failed. In the example, the wrench over-torqued by about 5 % at its lowest settings: you would check which jobs used it near 20 N·m and decide whether that matters for those joints. Our guide to out-of-tolerance results walks through the impact assessment.

Common mistakes when testing a torque wrench

Related guides

• How to write a calibration procedure: the general structure this procedure follows.
• Torque wrench calibration hub: everything on torque wrenches in one place.
• TUR calculator: check that your tester is good enough for the wrench.
• Uncertainty calculator: build the uncertainty budget.
• Out of tolerance: what to do when a calibration fails.
• Calibration intervals: how often to recalibrate.
• Calibration label generator: print the calibration sticker.