Is it advisable to use a derivative block to get acceleration from velocity data sampled at 5ms?? Is there any block to filter the velocity data and then use FFT before taking derivative and then use IFFT? will FFT work for discrete velocity at 5ms??

I am using velocity data from CANALYSER tool for creating Matlab files and then use it for feeding inputs to a derivative block to get acceleration. The results are not encouraging so I plan to filter my velocity data using FFT and then use derivative before using IFFT to get acceleration. Will this work or I use a low pass filter after FFT block and then IFFT block before taking derivative. I use Velocity as discrete signal sampled from CAN bus, will FFT block work for this?? Is there any standard trusted way to derive acceleration from velocity data? Shall I continue with Time domain analysis or use frequency domain to get better results? Can I use a standard low pass filter block like butterworth and operate upon my discrete velocity values?

[deepak raut]

I am using this for an automotive application where I intend to use velocity data from Vehicle speed sensor on front axle to get data on CAN bus. This velocity data from CAN bus message sampled at every 100ms(tunable) will be used for getting acceleration values to help predict vehicle mass and grade using State space model. My state space model and filters are ready but the acceleration values to be fed to the model is behaving badly to help the model predict mass and grade. Only if I get a suitable acceleration value for each sample time,I shall be getting the model to work well. I am not very clear on Time domain and frequency domain methods and their advantages. Also which analog or digital filter methods to use. I just have velcoity sensor data on CANbus which is discrete and with noise