This work aimed to develop an efficient R1ρ dispersion imaging method for clinical studies of human knee cartilage at 3 T. Eight constant magnetizations (Mprep) were prepared by tailoring both the duration and amplitude (ω1) of a fully refocused spin‐lock preparation pulse. The limited Mprep dynamic range was expanded by the measure, equivalent to that with ω1 = ∞, from the magic angle location in the deep femoral cartilage. The developed protocol with Mprep = 60% was demonstrated on one subject's bilateral and two subjects' unilateral asymptomatic knees. The repeatability of the proposed protocol was estimated by two repeated scans with a three‐month gap for the last two subjects. The synthetic R1ρ and R2 derived from R1ρ dispersions were compared with the published references using state‐of‐the‐art R1ρ and R2 mapping (MAPSS). The proposed protocol demonstrated good (<5%) repeatability quantified by the intra‐ and intersubject coefficients of variation in the femoral and tibial cartilage. The synthetic R1ρ (1/s) and the references were comparable in the femoral (23.0 ± 5.3 versus 24.1 ± 3.8, P = 0.67) and the tibial (29.1 ± 8.8 versus 27.1 ± 5.1, P = 0.62), but not the patellar (16.5 ± 4.9 versus 22.7 ± 1.6, P < 0.01) cartilage. The same trends were also observed for the current and the previous R2. In conclusion, the developed R1ρ dispersion imaging scheme has been revealed to be not only efficient but also robust for clinical studies of human knee cartilage at 3 T.