Inertial Navigation Systems (INS') are used as a primary mean of navigation in mostly all of the unmanned and autonomous systems. INS accuracy is categorized as grades in which the navigation grade is the most accurate and commercial grade is the least. The requirement of any used INS is to provide high accuracy information on the position, velocity, and attitude over a certain period of time. The problem of using low grade INS is that their accuracy degrades rapidly with time. To provide accurate estimation of navigation information, modeling of the sensors' noise components is required. The sensors' noise components are categorized in deterministic and stochastic parts. Deterministic noise such as bias and scale factor is easily removed in a process of laboratory calibration. Stochastic noise is the hardest part and needs special processes to be modeled and removed. Allan Variance (AV) technique is a time domain method which can be used to characterize various types of stochastic noise terms appear in inertial sensor data by performing certain
operations on an amount of data. In this paper, the relationship between different parameters which affect the operation of low cost Micro Electro Mechanical System (MEMS) inertial sensors such as sensor bandwidth and sampling rate is explored using Allan variance technique. Test results show that by carefully choosing internal inertial sensor settings, the
sensor stochastic noise can be accurately modeled and hence, navigation processing is highly improved.