The main lobe jamming suppression of array radar is a difficult problem in the field of radar anti-jamming. This paper proposes a blind source separation main lobe anti-jamming algorithm in sub-array domain based on joint approximate diagonalization of eigen-matrices（JADE）blind source separation. The algorithm provides lower complexity compared to the traditional omnidirectional array algorithm， especially for large arrays. The omnidirectional array is divided into subarray domain. A new blind source separation channel number pre-estimation method is designed on the basis of subarray domain using multiple signal classification（MUSIC） and constant false alarm rate （CFAR） to provide prior information for JADE. The separated signals are processed pulse compression and signal detection to achieve radar main lobe anti-jamming. Simulation analysis shows that the method can effectively demonstrate the separation of target signal and jamming signal， and the calculation amount is significantly reduced compared with the omnidirectional array， which has a certain engineering application prospects.