Multicarrier phase-coded radar waveforms show significant potential in broadband radar applications by integrating phase coding with orthogonal frequency division multiplexing (OFDM) technology. However, their inherent high autocorrelation sidelobe levels limit system performance. To address this challenge, this paper proposes a two-stage joint optimization waveform design method. In the first stage, we construct an AC-MCPC signal by introducing chaotic coding in the time domain and applying a hamming window in the frequency domain, achieving effective sidelobe suppression. In the second stage, to achieve even lower sidelobe levels, we further propose the AC-MCPC-g signal. While retaining chaotic coding in the time domain, we employ a genetic algorithm in the frequency domain to optimize the window function parameters, thereby further reducing the sidelobe levels of the AC-MCPC signal. The results indicate that the AC-MCPC signal has significantly reduced sidelobes compared to the MCPC signal, while the AC-MCPC-g signal has achieved further suppression based on the AC-MCPC.