Low-Noise Orthogonal Fluxgate Using Flipped Current Joule Annealing

Abstract

It has been shown that annealing of amorphous magnetic wire used as a core for orthogonal fluxgates in the fundamental mode can significantly decrease the noise of the sensors in the $1/f$ region. This is due to an increase in the circumferential anisotropy due to the dc current flowing through the wire during annealing. This method, however, presents some drawbacks: first, it requires an infrared furnace and precise compensation of the magnetic field inside it. Second, it is very slow, because it requires the cooling of the whole furnace before removing the wire. Most importantly, while the $1/f$ noise decreases, the noise floor increases. In this paper, we present a method that allows the simultaneous reduction of $1/f$ noise and noise floor. This method is based on Joule annealing by means of a very large current in the wire. The current is periodically flipped with 0.25 Hz frequency in order to avoid an excessive increase in the offset. The annealing is performed in a four-layer shielding to avoid the presence of an external dc field. Annealing for 1 min, 1 wire-based sensor returned a considerable noise reduction both in the $1/f$ regions from 2.5 to 1.5 $\text {pT}/\sqrt {\mathrm{ Hz}}$ ) while the noise floor was unchanged at $650~\text {fT}/\sqrt {\mathrm{ Hz}}$ . For larger annealing time, however, the noise floor rose. We tried to compensate this problem by increasing the number of wires to four, but also in this case, we achieved the best noise reduction (from 1.7 to $0.75~\text {pT}/\sqrt {\mathrm{ Hz}}$ at 1 Hz and from 470 to $350~\text {fT}/\sqrt {\mathrm{ Hz}}$ noise floor) with 1 min annealing. By the use of thermocamera, we discovered that the problem of long-time annealing was that the sensor head support was warming up too much. Therefore, we repeated the experiment by annealing 21 min as a series of 1 min annealing followed by 3 min of no current for 21 times to let the sensor head totally cool to room temperature after every annealing period. In this way, we achieved the lowest noise of 630 $\text {fT}/\sqrt {\mathrm{ Hz}}$ at 1 Hz and $400~\text {fT}/\sqrt {\mathrm{ Hz}}$ noise floor.