Hi there...
I'm working with some imu,there is something keep stuck in my head..
We all know accelerometer and gyroscope.
Gyroscope gives you angular velocity,
Accelerometer gives you acceleration.

Then if you Integrate angular velocity gives you angle,
Integrate acceleration gives velocity,Integrate velocity gives you position.

My point is,if i want precise position or velocity from accelerometer, I need higher sample rate to get more acceleration information?

No problem with Integrate Gyroscope,because it directly give me velocity,
and its result of integration is correct enough.

Drift is an issue for integrating any sensor - for a rate gyro the angle will drift over timescales of 10s to a minute, with accelerometers position is too unstable to use over more than 1s or so.
The signal already has a lot of noise in these MEMS sensors, no amount of extra sampling can remove it. They are also not great linearity anyhow.
Remember these are tiny devices measured in mm and microns in size, they cannot be used for inertial navigation, but are good enough for orientation measurement.

Drift and noise can decrease by filter、algorithm and eliminate if increase the numbers of sensor.

Things I wanna know is:
The acceleration from accelerometer,is the average from last read to at present?
Or just at the present time's acceleration?
If is the present time's acceleration, higher sample will give more information that demand.
Want the know the principle of it.
uh sorry for my bad english..