Continued efforts to identify compatible or incompatible blends have demonstrated a need to separate overlapping curves or peaks in a thermal trace. One way to do this is by using the derivatives of these curves. While the function is available on most modern thermal analysis units, its usefulness is often limited by noise in the system. Two techniques which can help to greatly reduce noise are digital filtering and Fourier transformation of the data. Both approaches have been used on identical sets of noisy data. The FT method appears to offer greater potential for producing useful derivatives from noisy composite curves.