When faced with a particularly difficult technical challenge, engineers may be tempted to solve it with higher quality components, tighter tolerances, or more processing power, all of which drive cost and complexity. While those solutions are appropriate in some cases, most situations require more nuanced solutions that fit better within development constraints. At Synapse, we know that these nuanced solutions always come from deliberately breaking the problem down into manageable pieces. Once we’ve done that, we may look at similar problems and corresponding solutions applied in a new way, or we may take deliberate steps to develop a reliable unique solution. Either way, we typically find that we can best address challenging tradeoffs between business constraints and physics if we dive into the fundamental phenomena underlying them and apply solutions that directly and surgically address them. For us at Synapse, these simple and cost-effective solutions are often inspired by work we’ve done across a wide range of industries.

We believe effective system design hinges on a clear understanding of business constraints—product cost, timeline, and development cost—and the tradeoffs they introduce in the solution space. This understanding motivates teams to focus on the most fundamental design drivers and think creatively to avoid unnecessary complexity. We find this especially true when it comes to complex electromechanical system applications that are intended to be produced in high volumes and have stringent performance, usability, and reliability targets. 

A Complex Challenge

Recently, we developed a system for high accuracy, low latency tracking of physical objects in three dimensional space with a strict set of performance and cost requirements. The system determines the distance between photosensors based on the time it takes for a laser sweep to pass between them, and includes a complex set of design variables and interdependencies. The laser sweep is created by an optical assembly driven by a motor, and performance degrades with distance between the laser source and the photosensors. In order to achieve performance targets at longer distances, we needed to minimize jitter, or the inconsistency of the motor assembly’s angular velocity. We initially planned to implement a fluid dynamic bearing motor specifically to address this challenge, but since the angular velocity could only vary by a tiny fraction (~0.0003%) between rotations, this selection still left an 80% performance gap to be closed, and would have increased the COGS by about $4.00. Given the production volumes and other factors at play, we also knew we couldn’t increase the overall device cost by more than $2.00 to solve this problem—clearly a significant challenge.

Revealing the Simple Solution

We evaluated a few common options for addressing this problem, and decided to approach it as a signal processing challenge, opening a toolkit that we wouldn’t typically consider for a problem like this. We performed deliberate experiments to identify the principal sources of jitter, which were natural motor variation and measurement noise. Our key insight was that increasing control over the motor variation also increased the effect of measurement noise. One concept we evaluated based on that insight was a sensorless speed control algorithm based on motor current and voltage, and we found we could expand the bandwidth for control of the motor variation while tightening the bandwidth for our measurement noise, allowing us to find an optimal balance between the two sources of jitter. We determined that we could implement this solution with a slightly more powerful MCU than we were already planning to use, which incurred an additional cost, although still below the $2.00 limit.

This solution balanced the tradeoff between performance, device cost, and development effort. We broke the problem down into manageable chunks to determine what we needed to learn. We quickly, and deliberately experimented to understand our system and sources of variation. We creatively identified a relatively simple solution to address the principal source of variation. And when we tested the solution to verify it, we achieved the performance and reliability levels we expected.

When we developed the AirFloss with Philips Sonicare, our challenge was to deliver an air & water mixture with enough force to demonstrate effective interproximal cleaning, while meeting specific reliability and cost targets. We began with a blue-sky investigation of possible solutions, including intuitive ones that involved separate actuators for the air and water. But, upon quick realization of the part count impact on device cost, reliability, and control system complexity, the team decided to explore concepts using a single motor to simultaneously drive water and air pumps. Once we developed some basic mechanism concepts to convince ourselves that they could fit inside the industrial design size targets, we were able to focus on the merits of different components and materials in our architecture evaluations.

We conducted experiments to validate this approach, and settled on a basic architecture where the motor simultaneously drives a peristaltic pump for the water and a spring-loaded piston pump for the air. Both are delivered to a mixing component in the head, then the nozzle. This solution not only allowed us to deliver the fluid power needed within an aggressive industrial design envelope, but also achieve a unit cost that supported the device’s value proposition while likely saving development time needed to address part count-driven reliability challenges.

Finding Balance

We love the challenges that arise from helping our clients innovate in their markets, especially when a new, unique solution is required that doesn’t have an obvious roadmap. Challenging tradeoffs between business constraints and physics are best solved with low-complexity solutions that emerge only after we develop a deep understanding of the problem. As engineers, resisting our instinct to quickly choose a complex concept—often the path of least resistance—is critical to developing the insights needed to uncover the simple, balanced solution.