A method comprising: receiving at least two audio channel signals; determining, for a first frame, at least two parameters representing a difference between the at least two channel audio signals; scalar quantising the at least two parameters to generate at least two index values; adaptively encoding an initial scalar quantized parameter of the at least two parameters; determining whether the initial scalar quantized parameter has a value different from a predetermined value; adaptively encoding any unencoded scalar quantized parameters where the initial scalar quantized parameter has a value different from the predetermined value; determining whether all of the at least two scalar quantized parameters have values equal to the predetermined value where the initial scalar quantized parameter has a value equal to the predetermined value; adaptively encoding any unencoded scalar quantized parameters and generating an indicator that an output is one of fixed or variable rate coding where the initial scalar quantized parameter has a value equal to the predetermined value and at least one of the at least two scalar quantized parameters have values different from the predetermined value; generating an indicator that the output is the other of the one of fixed or variable rate coding where the initial scalar quantized parameter has a value equal to the predetermined value and all of the at least two scalar quantized parameters have values equal to the predetermined value; generating a single channel representation of the at least two audio channel signals dependent on the at least two parameters; and encoding the single channel representation.