arXiv:2611.18405v1 Announce Type: new Abstract: In massive MIMO-OFDM systems, peak cancellation (PC) reduces the peak-to-average power ratio but introduces in-band distortion, where compensation signals transmitted from a subset of antennas can suppress this distortion at the users. However, their effectiveness depends on the antenna partition and compensation precoder, while nonlinear power amplifiers (PAs) generate additional distortion. This paper proposes a joint optimization method for antenna partitioning and distortion-aware compensation precoding. To enable this joint optimization, we derive a covariance-based analytical model of PC and PA distortion to predict the per-user, per-subcarrier signal-to-interference-plusdistortion- and-noise ratio (SIDNR). The predicted SIDNR is used to select the compensation precoder regularization, the disjoint antenna partition, and the number of compensation antennas. Numerical results demonstrate that joint optimization of the antenna partition and compensation precoder regularization improves BER and SIDNR-based spectral efficiency compared with fixed-partition designs, particularly at high signal-to-noise ratios.