1. direct connection  of A/D to the measured output (D/A to the input) with careful connection of the common grounding.

2. Realisation of sampling interval hrough "direct interrupt control".

3. Control of the noise level through the ground point measurement.

Dear Mr. Bhui,

You can still detect any kind of faulty input to the UKF system equations using the innovation sequence and an appropriate chi-square test applied to the innovation covariance of the filter.

For such detection your filter must be sensitive to the faults in the inputs (e.g. a filter with low process noise covariance, Q).

If there is a possibility for both measurement and input faults then you may need also a fault isolation process as you need to detect two types of faults both from the innovatiaon sequence. I suggest you to check the following publication in this case for the details of a fault isolation process

http://www.emeraldinsight.com/doi/abs/10.1108/00022660510585992

We applied the same isolation process for an estimation problem with the UKF in

http://onlinelibrary.wiley.com/doi/10.1002/acs.2393/abstract

If you have trouble accessing the full texts I can send them via e-mail.

Best regards