Magnetometer
From Theory of Measurements Wiki
Introduction
Pulsation magnetometers are designed to measure the variation of the magnetic field of the earth. A pulsation magnetometer consists of three search-coils, preamplifier and a computing unit with A/D converters.
Within the frame of statistical inverse theory a magnetometer can be modelled with the following first kind fredholm equation
By m we denote measurements, H stands for impulse response of the pulsation magnetometer and B stands for the magnetic field. Within the frame of this project, we study how accurately we can resolve the original magnetic and how we can make data fusion from different kind of magnetometers.
It is apparent, that within the frame of statistical inverse theory, we can invert the magnetic field from many measurements
By continuing this kind of deduction, it is also quite apparent, that we can measure all three different directions with various instruments and invert the original magnetic field.
Eventhough inverting the original magnetic field B sounds an easy to do task, we might end up in troubles, if the modeling has been done falsely. Therefore in this project we ask the following questions.
- How can one deduce the impulse response H for all i?
- How can one deduce the directions of the search-coils?
- How can we invert the magnetic field, if there is not adequately information? i.e. prior information usage in the inversion.
Below, we have done already some small scale projects and shown answers to part of our questions.
Done Projects
Response Curves
A short study on the impulse response deduction of pulsation magnetometers was carried out in 2005. A partial report can be found here PDF.
Magnetometer Direction Deduction
A pulsation magnetometer has got three search-coils, which measure different direction of the magnetic field. A small study on how to deduce these three directions is represented here Magnetomer Direction.
Numerical estimates of field values and magnetometer parameters
One possible numerical method is MCMC. An initial implementation can be found here.
Another option is Marquardt-Levenberg (ML) minimization.
Yet another possibility is to linearize to problem, and solve it using flips.
