2014.B.1.2 Reliability for Interplanetary CubeSats
Carl Brandon (1)
- Vermont Technical College, USA
Interplanetary CubeSats by necessity have long to very long duration missions compared to LEO missions. As a result, they require very high levels of reliability, above that of LEO CubeSats. We recently (November 19, 2013) launched a 1U CubeSat as part of the ELaNa IV mission on the Air Force ORS-3 launch on a Minotaur 1 rocket from Wallops Island, VA. The results from the CubeSats on this launch are instructive as to the reliability of CubeSats. Of the 14 non Air Force CubeSats, only five (Vermont Lunar, KYSat II, NPS-SCAT, CAPE 2, and PhoneSat) were heard from initially. Additionally, Firefly was contacted five weeks after launch. Eight CubeSats (H2, TJ3Sat, Black Knight 1, DragonSat, SPA-1 Tailblazer, COPPER, SwampSat and ChargerSat have never been heard from. NPS-SCAT ceased communication after one week.
The CubeSats on this mission were given the option of performing a vibration test on the flight unit at a level of the square root of two times the recorded vibration spectrum of the Minotaur 1, or to test an engineering model at twice the vibration power spectrum level. According to Cal Poly, the most common reason for failing the vibration test is screw, nuts or bolts coming loose. For large (expensive) satellites, very carefully monitored construction of the flight unit to be an exact duplicate of the engineering model can be accomplished, at great cost. In the CubeSat world, with limited budgets, and inexperienced personnel, this is not possible. Testing an engineering model does not help at all if you forget to locktite a screw on your flight unit.
The I2C bus issue resulted from the inclusion of redundant pull up resistors on the electrical power system board. These, in addition to those on the motherboard, prevented the bus being pulled low, and thus data could not be transmitted by the bus. For several CubeSats, an ISIS antenna was used which was deployed by commands over the I2C bus. A bus failure resulted in no antenna deployment, and thus a dead CubeSat.
Almost all CubeSats are programmed in C, which has an error rate of about one per 100 lines of code. We used SPARK/Ada which has an error rate of about one per 10,000 lines of code. 100 times better than C.
- Download slides in PDF format here (4MB)
- Supplementary video: