hi,
fft operation is basically a repetitive multiply and add operation.There are certain stages depending on the length of the input vector.
In the computation of each stage,one bit is increased in the resulting data. In the fix point calculation, this will overflow the data registers. To avoid this overflow of data registers, we shift the resulting output of the multiply and add operation and before storing the result, we shift the output data resister one bit right.
But in this process you get the output scaled by a factor 2^n if n is the no. of stages.