Adaptive prediction of sample values for digital transducers

In the following, we present a method to scale down the dynamic requirements of current sensors by adaptive prediction of the next sample value. Using magnetic field sensors to measure current, a compensation field can easily be generated by a feedback current. If the compensation field corresponds in magnitude to the field generated by the primary current, the current sensor only needs to be linear within a much smaller range, thus, the resolution can be increased. The prediction is done by applying LMS-adaptation rules on previous samples of the resulting magnetic field. This works well for wide sense stationary and periodic signals, and it requires a certain learning time until accurate results are achieved. Applications for this system can be found in the field of current measuring devices for low voltage networks in power distribution systems