Currently, most intelligent photoelectric dry separators are designed for two-product separation, which lack flexibility when processing raw coal with high middlings content, leading to potential coal resource losses. To address this issue, the three-product intelligent dry separator, through multi-stage recognition and separation mechanism design, can simultaneously separate clean coal, middlings, and gangue. Existing research on three-product dry separators primarily focuses on mechanical structure optimization or image recognition algorithm improvements, while the impact of coal quality characteristics (such as particle size composition and density composition) on separation efficiency remains unclear. This study investigates and compares the coal quality characteristics and corresponding separation effects of two feed materials using the TDS32-300 three-product intelligent photoelectric dry separator. Experimental results show that the probable deviation of the photoelectric intelligent dry separator is significantly higher than that of dense medium separation equipment, and the E-value used for evaluating the performance of dense medium separation equipment is not suitable for assessing the separation performance of photoelectric intelligent dry separators. The properties of the feed material significantly affect the separation efficiency of the three-product dry separator. A higher content of large particle sizes improves separation efficiency, and a higher content of >70 mm particles is more conducive to separation. When the feed material contains a high proportion of intermediate density particles, the separation accuracy of middlings and clean coal by the three-product dry separator is significantly reduced. The findings of this study provide valuable insights for the refined separation and system upgrade of three-product intelligent photoelectric dry separators.