Concurrent Product Design: A Case Study on the Pico Radio Test Bed

This paper presents a case study on the mechanical design and fabrication of the Pico Radio Test Bed: a wireless networking node produced from off–the-shelf components for experimentation with applications, networking, media access layer design, and position locating algorithms. Particular focus is placed on the systematic design process and resolving coupling design constraints between the mechanical and electrical domains. Three generations of the design are presented to demonstrate the evolution of the design as conflicts arise, problems are noticed, and requirements change. Introduction – Concurrent Design Concurrent design is also known by several other names, including simultaneous engineering, concurrent engineering (CE), and integrated product development. Even though these terms were not coined until the early 1980s, the concepts that CE embodies have been implemented as early as World War II (Ziemke & Spann, 1993). Noble states that “concurrent engineering is typically defined as the integration of both the product and the manufacturing design processes. The goal of this integration is to reduce the product development time, to reduce the cost, and to provide a product that better meets the customer’s expectations.” (Noble 1993) This definition, like many others, seems to fall into the common trap of reducing concurrent design to design for manufacturing (DFM). While DFM is a very important consideration for good design (Corbett 1991), it remains a subset of concurrent design. Perhaps a better definition is offered by Canty (Canty 1987), “Concurrent engineering is both a philosophy and an environment. As a philosophy, CE is based on each individual’s recognition of his/her own responsibility for quality of the product. As an environment, it is based on the parallel design of the product and the processes that affect it throughout its life-cycle.” This definition nicely states a key aspect of concurrent design (and focus of this paper): the parallel work of multidisciplinary teams. The Pico Radio Test Bed The Pico Radio project is one of the main topics of research at the Berkeley Wireless Research Center (BWRC). The goal of this project is to develop a series of meso-scale, low cost transceivers for ubiquitous wireless data acquisition that minimizes power/energy dissipation (Rabaey et al., 2000). This goal is to be achieved through a three-step implementation, the first of which is the Pico Radio Test Bed (PRTB) (da Silva Jr. et al. 2001). During this step, a macro-scale wireless networking node was produced from off–the-shelf components for experimentation with applications, networking, media access layer design, and position locating algorithms. To complete this phase, a multidisciplinary team comprised of computer science, electrical design, mechanical design, and manufacturing skills was assembled. This paper will focus on the collaboration between the electrical, mechanical, and manufacture paradigms used to produce the casing for this project. Design for Manufacturing The benefits of DFM are well documented (Boothroyd et al. 1994). To take advantage of DFM, the downstream manufacturing processes must be identified before the design phase begins. Injection molding was selected as the ultimate manufacturing process as it meets several of the key casing requirements. These requirements included: low per part cost, complex geometry, thin walled sections, and low weight, among others. Since only 200 or so casings were required, class B (short run) tooling was targeted. Once this decision was made, the documented DFM rules for injection molding (DFIM) (Bralla 1986, Wright 2001) could be followed. The primary DFIM rules that were considered in this design included: uniform wall thickness to avoid sink, warp, and residual stresses, no undercuts for a simple two-half mold, bosses and ribs designed to avoid sink, adequate draft angles, and few sharp corners to avoid stress concentrations. Injection molding requires a significant upfront capital investment for tooling. For this reason it is very important that the design be perfected and verified prior to tooling. This helps avoid any costly changes or retooling due to design errors. Rapid prototyping (RP) is ideal for this purpose as it is well suited to producing complex geometries that simulate the function of the finished part. RP is also useful for quickly checking unfinished design concepts, and verifying proper fit and function of parts (Lopez 2001). In order to simulate the function of the casing (particularly the screw bosses), Fused Deposition Modeling (FDM) was used to create prototype casings. The Stratasys FDM 1650 was used for this purpose, and the casings were made from P-400 ABS. The FDM process is capable of making very complex geometries, but quality can be improved by following some simple DFM rules (Montero, et al. 2001). Fortunately, most of the DFM rules for FDM are a subset of the DFM rules for injection molding (e.g. minimum wall thickness). For this reason, a part that follows DFIM rules typically requires only slight modification, if any at all, to be manufactured by the FDM process. Legacy Issues – The Marine Intercom Project As with most projects, the PRTB project did not “arise in a vacuum.” Some of the work was based on the marine intercom project that had been previously completed for DARPA. The goal of this project was to generate a new wireless intercom system for use in marine tanks (Brodersen 1999). It consisted of two earpieces and a handset. The earpieces contained a digital board, a power board, and a radio adaptor board connected to a radio. These boards represent the first generation of boards for the PRTB. Figure 1: Legacy – the Marine Helmet Project a) Helmet Assembly b) Earpiece with PCB To accommodate the flat-sided oval shape of the existing earpieces without wasting printed circuit board (PCB) area, the PCBs were also shaped as flat-sided ovals (see Figure 1b). It is interesting to note that this general flat-sided oval shape was retained throughout all versions of the PRTB, even though the shape could have changed to any form after the first generation. There are two key reasons why this shape was retained, even though it was no longer necessary for the design. First, it simply did not occur to anyone to change the board shape, even though an early change may have resulted in a simplification of the ultimate casing design. Second, everyone was overly concerned with other challenging aspects of the project to re-consider this simple aspect of the design. About halfway through the project, someone finally thought to ask why we were still using the oval shape. At this point, however, significant effort had gone into designing both the new generation PCBs and the casing. A change in shape at this point would have essentially required starting the casing over again. Legacy issues such as these are common in practice, particularly in the software engineering industry. A classic example of a software engineering legacy problem is the “Y2K” bug. Pescio presents several other examples, along with some approaches that can help mitigate these issues (Pescio, 1997). PRTB – First Generation The first generation of the PRTB casing was intended to be a quick prototype demonstrating the general approach to enclosing the board stack and to mounting other hardware. For this reason, it followed only DFM rules for the FDM process. At this point in the design, there were almost infinite options available, as the paring of the option space had not yet begun (Hazelrigg 1996). As previously mentioned, the first generation board stack was comprised of a power board, a digital board, and a radio adaptor board with an attached radio. These were connected together with board-to-board connectors to form the main board stack. In addition, a “dummy” board was included on top of the stack to provide protection to the connectors on the digital board, and to possibly carry a patch-antenna. The connectors for the dummy board were to protrude from the casing to allow for connection to external sensor boards (to be designed later). The power board provided switchable power supplies to several voltages, and connectors to the radio board and external hardware. The digital board contained a complete CPU subsystem, a Codec interface, and Xilinx external RAM and ROM to support protocol research. The radio adaptor board contained connector adapters for either the Proxim RangeLAN II radio, or the Bluetooth Single Chip radio. Additional hardware required to make the PRTB a self-contained system included: a battery, a power switch, and an antenna. Selecting some of these components proved to be somewhat challenging. The battery used in Motorola’s StarTAC cell phones was selected due to its small form-factor, its common availability, and the fact that a snap-fit mechanism is built onto it. A large, selfilluminating, case mounted power switch was selected to provide visual feedback on the state of the unit. The means for drawing power from the battery was neglected for the proof-of concept prototype. Motorola StarTac Cellular Battery (3.6V) On/Off Power Sw itch Pico Radio Test Bed Casing Cover Serial Port Window Figure 2: CAD model of PRTB version 1 assembly The casing shape was selected to be a flat-sided oval (to correspond with the existing PCBs), with scalloped indentations to position and laterally secure the board stack. This approach provided a quarter-inch gap around the PCB stack to allow for wire routing and antenna placement. Four long screw bosses protruded through cutouts in the boards to allow for fastening the lid to the casing with #4 screws. Cutouts in the casing side provided for serial port access, and a mounting location for the power switch. A set of ribs extended vertically from the bottom of the casing to support the board stack. The depth and position of these ribs was designed to accommodate either the Proxim or the Bluetooth radio boards. Finally, a mechanical battery mount was designed into the bottom of the case to accept the StarTAC battery. Several iterations of this feature were required to generate a tight fit with the battery. The casing was modeled in SDRC I-DEAS Master Series 8. The solid model of the PRTB version 1 assembly is shown in Figure 2. The lid simply followed the contour of the casing, and sat nested inside the top lip. It also featured two rectangular cutouts for the connectors on the dummy board, and a hole to access the reset switch on the digital board. These can be seen clearly in Figure 3, a photograph of the completed prototype. Note that the board stack is missing the dummy board, and shows the Proxim radio (the topmost board shown on the stack). Figure 3: Prototype Components for PRTB version 1 It was recognized early on that the casing (and as a result, possibly the boards) would have to change when the design was updated to follow the DFM rules for injection molding. Figure 4 shows several of the key features of the version 1 casing, annotated with the manufacturability issues that some of them pose. The key issue was that no draft had been applied to any of the features. In addition to slightly changing the shape of the entire casing, a draft angle would significantly increase the size of the long screw bosses at their base. This would cause two problems. First, the large plastic mass that would result at the base of the bosses would cause large sink marks to appear on the outside of the casing. Second, the large bosses would no longer fit through the existing cutouts in the first version of the PCBs. Figure 4: CAD model of PRTB version 1 casing features Battery End Sensor End Battery Mount Power Button Mounting Hole (undercut) Serial Port Window (undercut)