Previous research has shown that the voice source is strongly correlated with speech quality [1][2][3]. However in many existing pitch modification algorithms only the impulse train excitation is modi- fied, while the voice source is normally included in vocal tract transfer function and remains unchanged during modification. We present a production based pitch-scale modification scheme, which modifies the voice source waveform extracted from the speech signal. Experiment on synthesized voice sounds shows that compared to previous methods, better quality can be achieved for a wider pitch modification range.