Fluoroscopic imaging has become increasingly popular to investigate total knee arthroplasty kinematics non-invasively - 3D implant models are aligned with 2D image projections, and optimized via an edge-contour alignment technique. Previous studies have quantified the accuracy of this approach, however they do not adequately address the impact of image collection parameters. A particularly sensitive parameter is the pulse width, or exposure time per frame At longer pulse widths, a more motion is captured in a single frame; this can lead to image blur and subsequent degradation to image edge quality. Therefore, the comparative accuracy of relative joint kinematics as a function of pulse width and joint velocity needs to be defined. A limits of agreement approach was taken to define the mean differences between optoelectric kinematic measures (gold standard) and fluoroscopic methods at various pulse widths (1, 8, 16ms) and knee velocities (50, 100, 225°/s). The mean absolute differences between the optoelectric and fluoroscopic methods for 1ms pulse width was less than 1.5° and 0.9mm. Comparable rotational differences (1.3°) were observed for the 8ms pulse width but had larger translational differences (1.4mm). The 16ms pulse width yielded the greatest mean differences (2.0° and 1.6mm), which increased with knee flexion velocity. The importance of pulse width and velocity should not be overlooked for future studies - this parameter has proven to be a sensitive metric in the quantification of joint motion via fluoroscopy and must be identified and reported in future studies.