[ main page ] [ back ]

2012 : Elaboration of a Fault-Tolerance Strategy for Space-borne Digital Signal Processing Applications

Author(s)
Bernhard Fuchs
Abstract
Programmable Digital Signal Processing (DSP) is of paramount importance for the success of contemporary space missions in support of earth observation, astro-sciences and telecommunications. The present thesis aims at contributing to solving the problem of combining intrinsically contradicting requirements such as high signal processing performance, data throughput and flexibility on the one hand with robustness and high availability in the hostile space environment on the other. For economic reasons, the problem is tackled on systematic rather than technological level giving preference to pure software solutions implemented on commercial off-the-shelf (COTS) processing platforms. The applicability of commercial components to space applications is very limited due to ionizing radiation which may cause permanent modifications of the used materials and consequently of the electrical characteristics, referred to as total-ionizingdose (TID) effects, as well as single-event effects, experienced as soft errors in form of bit-flips or detrimentally as destructive latch-ups. Consequently, spaceborne signal processors are either fast but highly optimized for particular applications, thus, inflexible or programmable, slow and based on outdated semi-conductor technologies and processor architectures. The latter is due to the fact that for a commercial component to become a real space component it must have ample heritage, be screened or even modified for satisfying space-quality assurance requirements and it must be applied long enough to justify these investments. On the other hand, modern deepsub- micron processes are not only superior with respect to low capacitances and, thus, high processing speed but also with respect to TID and latch-up insensitivity so that they can be easier qualified for space. However, the vulnerability with respect to single-event upsets remains, resulting in intolerably low system availability. The objectives of the present thesis have been the selection of a modern programmable DSP as well as the identification, derivation, evaluation and experimental validation of faulttolerance (FT) mechanisms such as software-based FT, an external FT-controller as well as a combination thereof, to be applied to this component in order to establish a spaceborne DSPsystem satisfying payload dependability requirements. Typical applications of such a system are data and image processing, including filtering, decimation, coding and spectral analysis. Careful selection of FT-methods as well as optimal alloying of FT- and DSP-algorithms has been shown to be crucial for maintaining the full performance of either algorithm class and for ensuring software-product maintainability. Statistical measurements performed with the most promising candidate FT-mechanisms integrated along with typical DSP-algorithms have shown that software-only solutions, although economically attractive, fail in providing the required availability in combination with the desired processing power, while these goals can be fully met, if software FT-techniques are combined with an external FT-controller.
Bibtex
@mastersthesis{ fuchs:2012,
  author =      "Bernhard Fuchs",
  title =       "Elaboration of a Fault-Tolerance Strategy for Space-borne Digital Signal Processing Applications",
  address =     "Treitlstr. 3/3/182-1, 1040 Vienna, Austria",
  school =      "Technische Universit{\"a}t Wien, Institut f{\"u}r Technische Informatik",
  year =        "2012"
}
Download


[ main page ] [ back ]