Accelerometer calibration
The calibration simulations were based on scenario 3d_H and included solutions based on both absolute and differential DORIS observations. The accelerometer calibration assessment was based on having one accelerometer in the center-of-mass with a GRACE-type noise. The biggest contribution for precise accelerometer calibration comes from the kinematic orbit coordinates (GNSS). The addition of DORIS or ll-SST observation hardly improves the accelerometer calibration by POD.
Drag-free flight leaves a very small non-gravitational signal to be observed by the accelerometers, which makes the estimation of accelerometer scale factors less crucial and also very unstable. A proper maneuver scheme allows a very accurate kinematic orbit based calibration of the scale factor of the accelerometers: very accurate values can be obtained for the X and Z axes (accuracy generally better than 0.001), and more reliable estimates for the Y axis scale factors are obtained (better than 0.05). Periods of a few hours with thrust of the order of 50 nm/s2 might be sufficient. Accurate estimates of scale factors can be obtained as well when not flying drag-free (e.g. during the commissioning phase), especially for the X axis (much better than 0.001 during solar maximum, better than 0.01 during solar minimum). For the Y and Z axes, the performance is an order of magnitude worse, even more so for the Z axis during solar minimum.
A proposed implementation is to have 3 accelerometers on board of each satellite, with 1 accelerometer in the center-of-mass of the satellite and the other two symmetrically located with respect to this center-of-mass. This possibly leads to a √3 noise reduction when for example using a 3-accelerometer common-mode in case of accelerometer calibration by POD. The exact impact of such a possible reduction is yet to be assessed. Heritage from GOCE shows that similar results are obtained for 2-accelerometer and 1-accelerometer calibration by POD, but also that other calibration schemes, e.g. by comparison with star tracker observations, are feasible for at least the accelerometer scale factors.