"Filtering of Telemetry Using Entropy"

N. Huber, T. Carozzi, B. Popoola, and P. Gough
Space Science Center, University of Sussex


Spacecraft instrumentation produce vast datasets which contain scientifically interesting signals mixed with noise either inherent in the measurement process or in the nature of the phenomenon being studied. The scientific return from a space mission could be enhanced if an algorithm can be included on-board the spacecraft to determine the information content, and thus whether the data is “interesting enough” to transmit.

This work suggests a way of employing the entropy measure, as defined by Shannon [1], of a dataset as a cut-off filter for data transmission. It is suggested that an accurate calculation of the entropy inherent within a dataset can be utilised to determine whether it contains “enough information” to make it scientifically interesting. Random, noise-like datasets exhibit the highest entropy, while “interesting” datasets with an underlying structure would have lower entropy. If the entropy of a dataset is measured to be too high, that dataset can be considered as closely resembling noise, and hence can be flagged as “uninteresting”.

In a system with a limited Telecommunications budget, or if data-mining techniques are to be used on the receiver side, having this knowledge in advance can prove to be of critical importance. From preliminary research, we have established that this is the case for the telemetry transmitted by the CLUSTER spacecraft, launched by ESA in July 2000, where portions of telemetry that were most scientifically important coincided with those having the lowest entropy measure. In one particular dataset, the important portions amounted to less than 40% of the telemetry from this instrument. Hence, it was concluded that, if this filter had been employed on-board, there would have been a decrease of telemetry by up to 60% for that channel.

We decided to implement the filtering of telemetry using entropy in FPGA’s, as they are a very well suited platform for the parallel algorithms required [2][3]. An FPGA implementation can also process multiple datasets simultaneously, which would be an advantage especially for space instrumentation, since the entropy of multiple channels may be required. Other factors, such as reconfigurability, cost and a developer friendly environment, make FPGA technology probably the strongest platform to employ for this application.

We will present our preliminary results, based on CLUSTER. We will also present the implementation in FPGA’s of the various entropy calculation algorithms, along with their considerations and complications. Finally, we will show how this entropy filter is to be implemented on board an upcoming instrumentation mission that our team is involved in, as well as indicate further directions for this work.


  1. SHANNON, C.E. (1948): "A Mathematical Theory of Communication", Bell Syst. Tech. J., 27, 379-423, 623-656.
  2. BURG, J.P. (1967): "Maximum Entropy Spectral Analysis", Paper presented at the 37th Annual International SEG Meeting, Oklahoma City.
  3. G. GOVINDU et al. (2004): “A high-performance and energy-efficient architecture for floating-point based LU decomposition on FPGAs”, IEEE Proceedings: 18th International Parallel and Distributed Processing Symposium.


2005 MAPLD International Conference Home Page