When you specify the sensitivity of your accelerometers (10 mV/g or 100 mV/g, for example), does the sensor output in g pk or g rms units?
The short answer is neither. Whether our accelerometer reads out in g pk or g rms units is entirely up to the user and the data acquisition equipment being used. The accelerometer knows neither peak nor rms units, and in fact does not even know what a "g" unit is. The accelerometer itself is just an analog transducer designed to output an electrical signal proportional to input acceleration. The sensor is calibrated to output-known electrical amplitude, given certain, known mechanical input (acceleration). This is the very definition of the sensor's "sensitivity" (with units of mV/g).
If you were to connect the accelerometer to an oscilloscope (and the accelerometer were properly powered), then subject it to various accelerations by waving it around, you would see the output (just a voltage signal) also varying as the accelerometer is waved about. If you recorded this voltage signal somehow, you could then go back and measure the waveform's amplitude. You might measure the waveform's zero-to-peak (or just peak) amplitude, its rms amplitude or even its peak-to-peak amplitude. In any case, you could characterize the waveform by any of these measures. But notice that in none of these cases did the accelerometer play a role in what units were used. You, as the user and with your data acquisition equipment (the oscilloscope), determined the amplitude and whether it was in peak, rms or even peak-to-peak units. The accelerometer simply provided a dynamic, calibrated voltage signal.
You may have noticed that, on some of our calibration certificates, we state the accelerometer's exact, calibrated sensitivity such 9.595 mV/g @ 100 Hz, 10 g pk. We are NOT saying the accelerometer outputs in g pk units. We are simply showing that the sensitivity of the accelerometer was determined (on a vibration shaker) at a frequency of 100 Hz and at an acceleration amplitude of 10 g pk. But it should be pointed out that, had the acceleration amplitude been set to 10 g rms (or even 5 g pk, for that matter), the sensitivity would have been very nearly the same, if not identical.
See below for a sampling of unit equivalencies (applies only to sinusoids):
0.707 x peak value