Shown in the same graph is the transfer function calculated from the injection in the detection lab, measured with a single magnetometer. Note that up to 150 Hz, the transfer function has the same slope as expected from actuating the mirrors. This is probably caused by the NI mirror being actuated by the coil in the the detection lab. Only at higher frequencies, the coupling might be via the electronics themselves.
Fig 2 shows the magnetic noise projection, obtained by multiplying the transfer functions by the magnitude of the magnetic field in quiet conditions. As shown by the old measurements, the coupling can change by about an order of magnitude between the different orientations of the field. This could lead to errors in the noise projection if the field would be pointed exactly along the insensitive direction during the noise injection, but along a sensitive direction for an environmental noise source (or the other way around). This seems like a pathetical case, but expect errors of a factor of 3 or so. Also note that this projection is only accurate for a source located in the far-field, so sources inside the oven will be underestimated. The magnetic coupling from the electronics in the detection lab seems to be far from limiting.
Fig 3 shows a plot of the coherences between the magnetometers of the different buildings and the dark fringe. For the central building, lines can be observed at 10.6, 11.3, 19.6, 28.8, 34.8, 41.6, 41.9, 47.0-47.6, 50 and 52.4-53.0 Hz. Some of these seem to be moving in frequency. The end-buildings only show coherence for some 10 Hz harmonics and calibration lines. Fig 4 shows a zoom of the noise projection around 10 Hz, which shows that the 10.6 and 11.3 Hz line are well predicted as magnetic noise from the central building. A zoom around 50 Hz yields the same conclusion for the 47 and 53 Hz lines. Also note that the amplitude of the 50 Hz line is predicted within a factor of 2.
As a comparison, fig 6 shows the old noise projection based on the measurements with the coil placed next to the different ovens. The projected noise was clearly to high. Its calculation uses a dipole model and depends on the dimensions of the coil, the calibration of the current probe and the calibration of the magnetometers. The new projection needs no model or calibration value. The new method thus seems to be both easier and more accurate.
Conclusion: the magnetic noise is currently not limiting the sensitivity, except for a few lines around 11 and 50 Hz; The 50 Hz line itself can be explained pretty well by magnetic coupling; The noise floor is not very far below the sensitivity and even above design, but this might be simply sensor noise. There is no evidence of coupling via the detection-lab electronics.
