This paper presents current developments of the attitude determination algorithm for Delfi-n3Xt, TU Delft next nanosatellite. Several novel quaternion filters using Sun vector and Earth magnetic field measurements and rate gyro outputs are presented. The quaternion measurement matrix associated with each line-of-sight measurement is shown to be rank deficient. This property is exploited in order to design reduced order measurement update stages in the filters. The measurement model reduction is designed such as to preserve the statistical information. The filter covariance propagation can cope rigorously with the multiplicative process noises. The paper also describes the development of the Sun vector determination algorithm, which merges the outputs of 6 body-mounted four-quadrant Sun sensors. For each sensor, a simple algorithm allows Sun vector determination while avoiding the use of uncertain physical parameters. This algorithm takes into account geometrical imperfections linked to manufacturing limitations. A thorough error analysis of the photodiodes measurement outputs is carried out. A spacecraft Sun vector determination algorithm is proposed and illustrated, in the absence of Earth albedo effect, via Monte-Carlo simulations and experimental validation. In addition, extensive Monte-Carlo simulations illustrate the good performances of the quaternion filters using spacecraft Sun vector and Earth magnetic field measurements. The novel reduced filter shows good performances in a challenging tumbling dynamics environment, where a standard additive Kalman filter fails to converge.