In reality yes - the simplest designs can be simplified to thinking of just a capacitor with one sprung 'plate' that is more stretched or compressed when accelerating, so the recorded capacitance value is varied, and as part of the first use, the rest state value is recorded. However the better sensors hide this from you as the calibration is performed in the factory - in the case of ones with a digital output the 'cal zero' value for each axis of that sensor is burnt in to a PROM and for ever after subtracted from the actual reading to  create the numbers fed to the user.

(in reality there is either a table or some quick calculations going on, as you need a range of offsets, to allow for the sensor also responding to other stimulii - temperature gradients, varying supply voltages, package  stress, and other things - all of these also need to be measured and corrected for to get a 'pure' acceleration response.)

It is also possible to get bare sensors with an analogue output where you do this sort of correction yourself, or there is some cal-step for the user to do each morning, like turning the device over to reverse the sense of 'g'. Very small errors in horizontal alignment can have interesting effects. The 'roll your own' calibration is normally only done when extremes of speed, temperature or ultra low power use are required, or, oddly for very low cost things not needing great precision such as car alarms

The main problem is drift especially for deducing things like path over ground- a very small error in the readings (perhaps temp has drifted from the cal temp or there is something else that gives a small voltage offset) if uncorrected integrates over time to a spectacular position error. 'proper' nav systems combine accelerometers with other sensors such as gps and or magnetic compass, to try and re-anchor the co-ordinate frame as often as possible. Home rolled systems need to achieve this drift correction another way.

Mike

In reality yes - the simplest designs can be simplified to thinking of just a capacitor with one sprung 'plate' that is more stretched or compressed when accelerating, so the recorded capacitance value is varied, and as part of the first use, the rest state value is recorded. However the better sensors hide this from you as the calibration is performed in the factory - in the case of ones with a digital output the 'cal zero' value for each axis of that sensor is burnt in to a PROM and for ever after subtracted from the actual reading to  create the numbers fed to the user.

(in reality there is either a table or some quick calculations going on, as you need a range of offsets, to allow for the sensor also responding to other stimulii - temperature gradients, varying supply voltages, package  stress, and other things - all of these also need to be measured and corrected for to get a 'pure' acceleration response.)

It is also possible to get bare sensors with an analogue output where you do this sort of correction yourself, or there is some cal-step for the user to do each morning, like turning the device over to reverse the sense of 'g'. Very small errors in horizontal alignment can have interesting effects. The 'roll your own' calibration is normally only done when extremes of speed, temperature or ultra low power use are required, or, oddly for very low cost things not needing great precision such as car alarms

The main problem is drift especially for deducing things like path over ground- a very small error in the readings (perhaps temp has drifted from the cal temp or there is something else that gives a small voltage offset) if uncorrected integrates over time to a spectacular position error. 'proper' nav systems combine accelerometers with other sensors such as gps and or magnetic compass, to try and re-anchor the co-ordinate frame as often as possible. Home rolled systems need to achieve this drift correction another way.

Mike