br Results and discussion The switching acceleration for eac

Results and discussion The switching acceleration for each of several switches is shown in Fig. 6 for the “cut and bend” design, and in Fig. 7 for the molded design. Each switch was left in the stressed stable state (shown in Figs. 3b and 5b) for a ace inhibitor of 2–3weeks, and each was tested several times over that period to determine its switching acceleration on that day. The data shows very low variation, particularly when compared with previous tests using plastic switches, which showed an average increase in threshold acceleration of 54% after only 72h in the stressed state [13]. By comparison, the largest variation between data points in the “cut and bend” design is 2.8% (the diamonds in Fig. 6), and all the variations in Fig. 6, from top to bottom, are 1.6%, 0.8%, and 2.8%. The largest variation in the molded design is 14.9% (the diamonds in Fig. 7). Note that switches represented by the other symbols in Fig. 7 are much smaller, with variations of 5.7%, 8.2%, 14.9%, and 3.2% from top to bottom. Also note that there is no clear trend toward increasing accelerations over time for any of the switches, unlike the previously tested plastic switches which indicated continued plastic deformation. Measured variations for these metal switches was likely due to unintended bending of the flexor elements as the switches were loaded and unloaded from the testing centrifuge. Each individual switch tested showed very low shift in the acceleration threshold over time; however, Figs. 6 and 7 both show considerable variation in switching acceleration from switch to switch. For the cut and bend switches, the maximum variation was about 23% compared to the average switching acceleration of about 17m/s2. For the molded switches, the maximum variation was about 86% compared to the average of about 32m/s2. To better understand the source of this variation, we modeled how the switches would respond to variations in the flexor angle in the switch. Induced strain due to cooling of the molten tin may also be a source of variation for the molded switches. We modeled both of these effects using the model previously described [15]. Briefly, Polysome/polyribosome model solves for the bending of the thin beams using an elliptic integral approach to solve the governing differential equation. The resulting equations are solved numerically by applying a given displacement to the beams, and calculating the forces that result. Note that more complex models have also been developed for similar bistable mechanisms [16,17] but we have found the more simple model is sufficient for understanding most of the performance variations we measured. Based on the model in [15], Fig. 8 shows the percent change in switching angle for a single flexor in the “cut and bend” design if that flexor is at a different angle than the designed angle of 4°. The data was generated using a model of flexor motion. The results show that, for this design, the force can vary by as much as about 23% for a deviation in flexor angle of only 0.75°. Analysis of the leg angles shown in Fig. 3a suggests that, in this image, the flexors vary by as much as about 0.6° compared to the designed value. Hence, variation in flexor angle likely accounts for most of the variability seen in Fig. 6. Similarly, Fig. 9 shows the percent change in switching angle for a single flexor in the molded design at a different angle compared to the average leg angle of about 6°. For this design, the force can vary by about 38% for a deviation in flexor angle of about 2°. Analysis of the leg angles on fabricated switches (such as the switch shown in Fig. 5a) shows that this amount of variation is typical for this design. Hence, flexor angle variation can account for much of the variation shown in Fig. 7, though not all. Other sources of variation are possible, however. For example, the switches represented by the triangle and the X\'s in Fig. 7 were fabricated by using a steel flux clean on their ends before pouring the molted tin, while the other two switches received no flux. The switches fabricated using flux seem to exhibit better bonding between the steel and tin, which may account for their lower variation with respect to each other.