Accelerometer Nominal Specifications

Does a Manufacturer's Spec Matter?

One of our Dynamic Sensors & Calibration Tips newsletter readers recently sent me an email communication about a concern they encountered with some pre-owned ‘Brand X’ accelerometers which they purchased from an e-commerce website. The reader explained that the sensors performed “fine” in the lab, but that the recent re-calibration certificate provided to them by TMS showed that although the sensors’ frequency response was fine, the reference sensitivity was out of tolerance from the manufacturer’s stated maximum deviation from nominal. Below is the discussion that followed:

Reader Q: The original manufacturer’s published specification for these accelerometers is 10 +/- 2% pC/g. The re-calibration certificate showed 10.3 pC/g and was flagged and labeled out of tolerance. Since we can set up our controllers to scale by the re-calibrated sensitivity, can’t a point be made that the manufacturer specification doesn’t matter? If they are 10.3 pC/g and we use that calibration factor, our shaker will run correctly since it knows its input. Can’t we just get the re-calibration certificate reissued for the known response and skip the “out of tolerance” notation? We do not see a need to dispose of these two accelerometers, although we can’t use them unless we have the valid calibration certificate. What are your thoughts on this?

TMS A: Regarding your questions, the answers are both “yes” and “no.” To better answer the questions, we have attempted clarity by dividing the questions and answers into two parts:

Q1: Isn’t the manufacturer specification kind of arbitrary, and as long as I’m scaling properly in my calibration or control system/software? Shouldn’t it work?

A1: Yes… if the scaling is correct, the conversion to engineering units should be correct and your measurement system should work fine.  But, it begs the currently unspoken additional question…

Q2: If the manufacturer specification is +/- 2% and now the sensor under test calibrates showing 10.3 pC/g, what happened?

A2: There may be a number of things that could have happened, some of more concern than others:

  1. Check with the original equipment manufacturer and verify original specs and original calibration data. The sensors could have been manufactured under different specifications than sensors of the same type are currently. Have the OEM run the serial numbers.
  2. ii.            Check with the original equipment manufacturer of the accelerometer to see if the sensing element is ceramic or quartz. The most stable vibration sensing element for long-term calibration reference stability is a naturally polarized quartz reference sensing element.
  3. Has the unit undergone a large thermal or mechanical shock?  Accelerometers constructed with ceramic sensing elements can also be susceptible to a shift in base sensitivity from large shocks.
  4. Has something changed in the way the reference accelerometer is calibrated?  A different mounting torque? Different vendor? Secondary method vs primary method? There will always be some variation in measurement from vendor to vendor.  The variation is acceptable if it is within acceptable uncertainty levels or unacceptable if it falls outside of their uncertainty levels.  Comparing lab to lab uncertainties is why reputable and quality metrology labs participate in Inter-Laboratory Comparisons (ILCs), statistically comparing uncertainties in a term called “E sub n” that ensures that reference values overlap within the comparison uncertainties, including sufficient overhead.

If any of the aforementioned has happened, we would recommend removing the unit as your working reference and acquiring a replacement.  The goal is to figure out if the reference accelerometer is stable, or still changing.  Any significant shift (found by either annual reference calibration, or a change caught by running a daily sensor verification) is a call for deeper and more regular examination of both the equipment and process.  Equipment changes can often be caught by the question, “What changed recently?” and by substitution, one component at a time.  Process, training or proficiency changes can be caught by daily validation runs, comparison between multiple operators and Inter-Laboratory Comparisons.  The following link talks more about Inter-Laboratory Comparisons (aside: we offer free participation for vibration calibration system owners.)

Later in a follow-up email to the customer, we discussed comparing and navigating the procedures and control methods in our two quality systems...

Q: How can I work with these sensors within our Quality System?

TMS A: We have researched the work done on the two Brand X accelerometers, and here is the challenge at hand…It seems both your organization and our organization are operating under accredited quality systems.  Brand X may not be.  We can’t find reference to a quality system or accreditation on the organization’s website.

Our ISO 17025 quality system (the global standard for quality systems of test and calibration laboratories) is audited by A2LA (American Association for Laboratory Accreditation), and by audited procedure we must report a non-conforming deviance from the original equipment manufacturer specifications.  All the published information for the specific accelerometer family on Brand X’s website specifies 2% max deviation from nominal sensitivity at reference.  By definition in our quality system, we must follow manufacturer published specifications.

One way to navigate quality systems is to get Brand X to provide you a letter stating that at the time of manufacture of your sensor, the organization’s specifications allowed for a large range of nominal sensitivities (ex. 3% or 5%). This statement of authority from the vendor would allow for a link in facts, giving traceability and authority to reissue your calibration certificate as “in tolerance.”


On Brand X’s calibration sheet online it says the specified uncertainty in the 10 Hz to 2 kHz is 2.1%.  Functionally, this means the uncertainty of the measurement is larger than the tolerance they are trying to hold to and it’s likely that many of their production calibrations would be outside the 2% of the nominal specification and is still accepted.  Perhaps this is why a 10.33 pC/g example calibration certificate is displayed on the company’s website. In general, this practice would raise a flag in our manufacturing and quality system.

General practice in the industry is to realize that both mounting and operator skill add a small contribution to uncertainty. In effort to try to get the unit under test to pass within tolerance, our technicians take and mount/remount the sensor three times in an attempt to capture a best performance mount that will pass.  In the case of the 10.21 pC/g, our technician assures that he would have tried three or more times when sensitivity is this close to the edge and was unable to get the unit to pass within 2% of nominal.

The short story is that the original equipment manufacturer may be the key to getting these sensors into use. Quality systems are in place to protect against “non-conformance” and the subsequent errors that can be propagated by their use. However, another route that some companies have taken with “out of spec” sensors is labeling them “approved for reduced functionality.”  This most often has been used when accelerometers fail the deviation at high frequency, approving them (with quality system auditor approval) to be used in the lower frequency ranges.  If you can show the sensor is stable over multiple quarterly calibrations, your quality system auditors may approve a variance for the wider tolerance in nominal sensitivity.  In the worst case, you might find that the accelerometers must be scrapped. Most sensor manufacturers could give you a competitive scrap allowance toward purchasing new units.