Chapter IV: Large Scale Flexible Arrays
A.3 Project Planning
Figure A.4: NASA Technology Readiness Level Table [166]
Type of Component COTS Space-grade
White LED[99] [72] ($0.50) ($15)
Voltage Regulator[73] [109] ($3) ($500) Microcontroller[107] [108] ($5) ($1000)
4-Layer Circuit Board[122] FR4 substrate ($100) Polyimide substrate ($500) Table A.3: Price comparison for several COTS and space grade components.
The most costly mistake (in both money and time) for the MAPLE project came when selecting a camera to photograph the payload in space. While not critical to the scientific objectives of the mission, photographs provide information about the condition of the core systems and allow the work to be presented in compelling way. The MAPLE team first selected a camera with flight heritage but limited documentation (Gumstix Ironstorm [59]). After investing significant time trying
Figure A.5: NASA Mission Classes from [81]
and failing to develop working software5the MAPLE team switched a more simple Arducam Camera [8]. While lacking flight heritage, the Arducam were compatible with available voltages and digital interfaces of the SAMD21 motherboards. The Arducam’s were intended to work with Arduino and as such the MAPLE team had to completely re-write drivers to work from scratch with our motherboard. This tied up significant work-hours on a project with limited personnel. While the Arducam functioned well enough to survive thermal tests, multiple Arducams could not be made to reliably run from shared I2C and SPI lines.
After months of development and testing with Arducam, we decided that these issues could not be resolved. We then switched to Raspberry Pi High quality Camera [125]
and Camera Module V2 [124]. The Raspberry Pi and its camera have flight heritage and as much or more documentation than any electronic component6. In retrospect this seems like an obvious choice but at the time we considered the Raspberry Pi a non-starter because it could not be powered by the 3.3V regulator on the aggregation panel. Additionally, the Raspberry Pi boards were substantially larger than we wanted to fit into MAPLE. The DC power issue was solved by feeding the Raspberry with 5V directly from same line powering the aggregation panel.
A few more mechanical holder components needed to be designed and tested to
5The Gumstix Ironstorm reached end-of-life at almost the exact same time it was chosen for use in MAPLE. Available software resources were removed the internet mid-project.
6Raspberry Pi is the go-to hobbyist low level computer platform, meaning a multitude of tutorials and examples are easily accessible.
accommodate the new boards and cameras, including a revision of the aggregation panel. Fortunately, the digital interfacing and control (which had taken weeks to establish unreliably with the Arducam and Gumstix cameras) was accomplished in a single afternoon. This mistake cost $8000+ and 100s of hours of effort, required another round of mechanical and thermal testing, and delayed the Flight Model by almost 3 months.
While the MAPLE camera saga is long and winding, we can reduce it to a few fun- damental lessons learned about selecting digital components. Do not underestimate the importance of your team’s prior experience and infrastructure with certain hard- ware. Even a small amount of experience can help get over the “activation energy”
that often derails digital systems. Our camera experience also demonstrates the importance of picking well documented components that are in the proper portion of their product life cycle.
The parts, processes, assembly firms, and testing firms needed for the project should be identified as early as possible. If aspects of the exploratory phase are left until later in the project the budget will creep up as undercounted costs rise and the schedule will inevitably slip as the team waits for quotes and external firms slowly complete their tasks. Doing work in-house and using lab and university facilities instead of contracting an external firm will save time and money. In general, the MAPLE team completed as much testing as possible in-house but paid for professional external manufacturing and assembly. Appendix *Budget* documents the expenses of the MAPLE project. Tests by external firms account for 24% of the total budget and could have accounted for 50% if not for use of Caltech resources. External facilities often have a per hour or per day cost and it is important to come fully prepared and ready to maximize value of the time.
The above discussion and the later budget breakdown omit what is likely the greatest expense: launch costs. While launch costs have decreased substantially, booking passage to space for a small/medium payload will likely cost 10s or 100s of thousands dollars. While there are exclusive opportunities at lower costs, the launch method and price should be determined at the beginning of a project in the event they become prohibitive in cost or lead time.
MAPLE used three versions of the electronic hardware: first a benchtop model for testing functionality, then an engineering model (EM) used for environmental testing, and finally a flight model (FM). The benchtop and EM models are shown in Fig. A.6. The three model approach is adaptable to most projects. As the benchtop
provides the first opportunity for verification of components and performance and allows software development to begin, it should be completed as early in the schedule as is feasible. If the core technology is starting at 4+ TRL, previously constructed subsystems may be able to be re-used as part of the bench top model. Even if it differs partially in form and functionality from the later models, early warning of incorrect footprints, misread datasheets, or more serious issues are critical. Once a prospective component is confirmed, a quantity sufficient for the EM, FM, and back-ups should be purchased. This avoids delays from long lead times and the unfortunate scenario of a forced switch of components because the part used for the benchtop or EM is out of stock. The EM which follows the benchtop must be similar enough in form and function to the FM such the environmental testing results are compelling. The schedule should have sufficient slack such that if the EM fails during testing there is sufficient time for revision and re-testing. The board designer should purchase components and boards for at least one “extra” copy of the EM. If the EM passes all environmental testing than the copy becomes the FM. If the EM fails environmental testing than the copy can be modified and used as the new EM.
The FM delivery date to the launch provider provides the point around which the schedule should be anchored.
Figure A.6: (a) MAPLE bench top model. Electrically functional but not ready for environmental testing. (b) Electrically functional MAPLE engineering model which has been successfully environmentally tested. FM assembly is in progress as this document is being prepared.