Using oscillating magnetic fields for indoor positioning is a robust way to resist dynamic environments. This work presents the hard- and software-related optimizations of an induced magnetic field positioning system. We describe a new coil architecture for both the transmitter and receiver, reducing inter-axes cross-talk. A new analog circuit design on the receiver side attains an acceptable noise level and increases the detection range from 4m to 8m (the covered area is increased from $50m^2$ to $200m^2$). The median positioning error is reduced from 0.56~m to 0.25m in the near field with fingerprinting methods. Experiments in office and factory areas (including robotic and industrial equipment) demonstrate the system's robustness in large areas. This work aims to enlighten researchers working on the same topic with constructive optimization directions on their own induced magnetic field-based systems.