We introduce a block polar quantization (BPQ) procedure that minimizes a weighted distortion for a set of sinusoids representing one block of a signal. The minimization is done under a resolution constraint for the entire signal block. BPQ outperforms rectangular quantization, strictly polar quantization, and unrestricted polar quantization (UPQ) both when assuming the Cartesian coordinates of the sinusoidal components to be Gaussian and for sinusoids found from speech data. In the case of speech data we found a significant performance gain (about 4 dB) over the best performing polar quantization (UPQ).