1)  How long does it actually take to calibrate a sensor?

First irony… during the actual process of accelerometer calibration, it takes longer to connect the signal cable and mount the sensor than to execute the actual reference measurement itself.  Modern automated calibration systems are able to control the excitation frequency at the specified amplitude and step through the complete frequency range in less than a minute.  A typical cycle time for a single axis of calibration may be on the order of five minutes with approximately 60% allocated to the physical effort of inspecting the sensor mounting surface, mounting, torque measurement, signal cable connection and ultimately removing the sensor.  The remaining actions take about 30% of the time for running the actual automated test sweep with measurement and the remaining 10% of time for data entry with model number, serial number, etc.  In modern automated accelerometer calibration systems, the data entry is also expedited with helpful features like built-in sensor databases to look up sensor details like nominal sensitivity and the high/low frequency range to automatically set up the test parameter and populate specification sections of the calibration certificate.  Further, automation like TEDS (transducer electronic data sheet) also speed the process, and eliminate typos as much of the information can be read from the digital memory chip within the sensor.

2)  What really needs to be done to “calibrate” an accelerometer?
Since there are no actual “adjustments” made during an accelerometer calibration, the calibration action serves as a control methodology to ensure confidence in the accelerometer performance and the documented scaling value for converting the raw voltage measured signal into meaningful engineering units.  At the bare minimum, the calibration should return a single point scaling value (in terms of mV/g for an ICP type sensor or pC/g for a charge mode accelerometer) at a reference frequency (typically 100 Hz or 159 Hz as reference).  While single point is the absolute  minimum, most commercial calibration also includes a frequency response from either 5 or 10 Hz (low frequency for calibration is actually exciter and reference dependent) to the sensor maximum frequency as defined by a maximum 5% deviation from the reference scaling value.  This frequency response should approximate a flat line across the entire frequency range allowing for the single point reference frequency scaling number to serve as valid for the entire range.  Some modern Digital Signal Analysis systems allow for inclusion of a full frequency response scaling curve, but in practice most users default to accepting the single value.

3)  Is that all there is to accelerometer calibration?
It’s a loaded question. Accelerometer calibration is generally about determining the scaling value at reference frequency and the flat frequency response described above, but when you send your accelerometer to certain manufacturers, you MAY get a host of other value-added services included in the same price (ask your manufacturer).  The additional services and no-charge hardware provided by PCB Piezotronics in Buffalo, NY, generally include, in addition to a base FRF calibration, testing these key specifications as a health check:

a. Discharge time constant measurement check
b. Resonant frequency check
c. Sensor bias check (if an ICP sensor)
d. Maximum transverse sensitivity
e. Refurbishment of the accelerometer kit:
i. Cleaning of the mounting surface
ii. New mounting base
iii. New mounting stud
iv. Small box of Petro wax
v. Replacement storage jewel case if needed
f.  In extreme cases, if the sensor exterior case is in bad shape, it may be a candidate for sandblast (or polish) and re-etch.

4)  What is uncertainty and why is this always shrouded in mystery?
Why the mystery?  Maybe it’s a little bit of math. Maybe it’s the debate fostered in round-robin testing from inter-laboratory comparisons comparing results.  Maybe it’s just a manifestation of our engineering stubbornness. Regardless of the reason, accelerometer calibration uncertainty is not difficult or scary.  Uncertainties are simply a statement of the possible range of measurement variation caused by the combination of equipment, environment and user practice.  Broken into two components, there are well-defined bias errors (like ADC uncertainty, reference accelerometer traceability uncertainty, DSP methods,  etc.) and experimentally determined random errors (like user skill and mounting variances, etc.).  Uncertainty gives a metric or “currency” that we can use to compare the performance between calibration labs across the world.  In general, a better overall uncertainty indicates better performance and discipline on the part of the calibration system operator.  This means better equipment selection in the measurement process; better procedures, performance and training; and ultimately, better knowledge of the accelerometer calibration process.  In the world of accelerometer calibration, your vendor should always be certified to a standard like ISO17025, as this provides the requisite level of confidence in the vendor hardware, practices and continuing critical processes like inter-laboratory comparisons.  When a vendor is certified, their uncertainty declarations are easily found through a link on the site of the accreditation body.  For example, here is the link for the A2LA certified uncertainty for The Modal Shop scope of accreditation.

5)  Is that all I have to know about accelerometer calibration?
Certainly not, but it’s a good start!  Sound calibration practices are all about control and confidence.  To provide your customers with quality measurements, it’s important to keep learning.  Your calibration system provider or your service vendor should maintain an experienced staff of calibration experts to be able to answer your questions by phone, visit or email.