In 2006, Goemans, van der Stappen, and Goldberg introduced a new three-parameter part feeding primitive, termed a blade, accompanied with a quasi-static algorithm that identified blade designs for given parts. Here I present software and hardware based experiments that compare the algorithm to physics based simulation and reality. I create a rigid body simulation in COSMOSMotion and determine it to be theoretically more accurate than the algorithm but at least 24 times more computationally expensive. I complete 30 physical experiments with three different parts and multiple blade designs on modular test hardware designed for the purpose. I find that the algorithm agrees perfectly with the experiments 65% of the time and completely disagrees only 10% of the time. The remaining 35% of the cases involve partial agreement between the two. I theorize that the discrepancies between the algorithm and the physical experiments are due to part momentum and perturbations in the parts position that can change the effective blade parameters. I then summarize observations of how certain blade parameters can affect feeding performance. The work in this report resulted in several revisions to the design algorithm that are detailed in Automated Feeding of Industrial Parts with Modular Blades: Design Software, Physical Experiments, and an Improved Algorithm.