For integer types of TT_TWIDDLE, the twiddle values are stored internally in lookup tables as their true value shifted right until the sign bit (2s complement) is in the most significant position of the data type (int16 or int32). However, 2s complement is an asymmetric range, which means the twiddles close to point 1+j0 are saturated. This can lead to spectral leakage. The library unit therefore offers TP_TWIDDLE_MODE, which, when set to 0 will have twiddle values as described above, but when set to 1 will have twiddle values of half this magnitude, so that saturation does not occur. This will have higher noise than for the other mode, but does not exhibit the spectral leakage.