Our third sprint saw some changes in the mechanical design. Last sprint, the mounts for the interior parts were held in place solely by friction against the sides of the box. However, since we still needed to access the insides, we couldn't glue the box all the way together, and the friction mounts pushed the box apart. In an attempt to remedy this, we added slots in the sides of the box for the mounts to fit into. Unfortunately, the round bottoms of the slots allowed the mounts to torque out of place, so this was not particularly effective either. It was also not aesthetically pleasing. Due to a lack of communication or a CAD review, the holes did not line up well with where the mounts fit, so a lot of sanding was done to make this part usable. However, adjustable springs were added to the stepper motor mounts to hold the wheel firmly against the page. Overall, the mechanism is now capable of turning pages.
During this sprint, the beat detection software was tested with live music. As expected, the system worked reliably when the pianist played at a consistent tempo. However, it soon became clear that consistent playing is not a valid assumption to make with live music. Continued testing showed that the page would often turn too late, specifically full measure later than it was meant to turn. This was solvable by hard-coding the turning difference, but this is not optimal because this value may vary for different pieces or different pianists.
Overall, we have achieved our Minimum Viable Product. Our system was capable of turning pages in a binder through the use of a hands-free beat detection method. There are still improvement to be made, however. In particular, the beat detection software must be revamped to more accurately detect beats in a live setting. Also, the box must be redesigned to better accommodate the holders for the stepper motors.
Our Main Takeaways
This sprint, we ran into complications due to a lack of communication. In particular, several people worked on individual parts of the CAD. We did not effectively communicate or have a CAD review before the parts were printed/laser cut. This led to improper fit and time wasted on sanding. In the next sprint, we are going to have a detailed CAD review prior to producing our parts.
During this sprint, we found that our current beat detection software is not fully reliable for live music. The current system relies on a combination of beat detection and time-based turning. For the final sprint, we will remove the time-based component and focus on counting the beats that are detected.
The PowerPoint that we presented in class can be found here: