This paper describes the efforts we devoted to improve Ambisonics decoders in Csound. Current version of the existing opcode, namely bformdec1, has some limitations that should be surpassed in order that the decoders better fulfill the Ambisonics criteria. In particular, the implemented decoders have no near–field compensation and do not use a different decoding matrix for low and high frequencies. These issues are addresses in a new implementation of the opcode, namely bformdec2, that also adds some features, such as additional loudspeaker array configurations (rectangle, hexagon) and a binaural output for headphones.