Internal low pass filter in acc LSM303 before A/D conversion

anyone knows if in accelerometer (LSM303DLHC, DLM chip) there is any low pass filter before A/D converter ?

Im using accelerometer in IMU (tried IMU9 v1,v2, AltImu-10). It is installed on a mobile robot. No matter what I try, the output signal is very ragged. The signal shape does not look any close to the movement of the robot (~200ms acceleration forward, ~800 ms forward smooth move, ~100ms slowdown). In the ‘forward smooth move’ the picks in X axis are many times greater that during accelerating or slowing down (robot moves along X axis). The overall signal shape does show there is any high acceleration/slowdown. When robot is not moving there is low noise (signal amplitude is about 10-100 times smaller than during movement).

I tried ODR 50 Hz up to 200 Hz (higher ODR result in data overrun on RaspberryPI). I assume there is no great implementation error because algorithms that need to measure Gravity vector works fine (2 fusion method - Madgwick, other simpler one).

I suppose there can be an aliasing problem. So there is my question if there is any low pass filtration before sampling?
If there is none, can anyone suggest an accelerometer (digital with I2C) that does have internal low pass filter or maybe a different solution?


I did not see any indication in the datasheet for the LSM303DLHC of a low pass filter inside it, but there is one mention on page 23 of the LSM 303DLM datasheet about a low pass filter that is used to compute the output data in low-power modes. Unfortunately, all of the information we have about those sensors is what can be found in their datasheets, so I do not know anything more about that filter. The newer LSM303D accelerometer and magnetometer includes an anti-alias filter that might be helpful, but there is not much documentation on it in the sensor’s datasheet.

If you post the output values you are getting from the sensor, I might be able to let you know if they seem normal. Are you doing any averaging of the readings in software?


Thank you for reply. The signal I described previously is attached in log file. The movement (mostly along x axis) starts about 150th sample.
I am going to integrate the data for short movements.
It seems I will try the LSM303D
roll,pitch,yaw,ax,ay,az.log (10 KB)

The readings you are getting look reasonable. If you are not already, incorporating the gyro data from the AltIMU into your program using sensor fusion should give you more stable results for a moving platform.