You are here : Home > News > Correcting the errors of integrated circuits used in space

Scientific result | Electronics

Correcting the errors of integrated circuits used in space

The reliability of integrated circuits is a key point for space applications. They are becoming increasingly complex and dense, which heightens their vulnerability to cosmic particles. To remedy this, researchers from the Inac are proposing a new integrated circuit structure, which will be more robust in a space environment.
Published on 19 December 2017

Both avionics and the space industry require radiation-resistant integrated circuits. Indeed, ionizing particles generate interfering currents that can alter certain signals and introduce malfunctions. Calculations can even be interrupted, especially if resetting is too frequent. This highlights the need for circuits with a small surface area and low consumption, which can operate even in the presence of these disturbances.

The proposed innovative structure is composed of two chains of identical functional blocks, designed according to an asynchronous or clockless mode. In the absence of any error, the two chains provide the same result. A comparator is placed at the outlet of each pair of twin blocks. When it detects a difference between a block and its double, it sends an error signal that drives a so-called "handshake" protocol. In the event of an error, the circuit is temporarily paused at the location where the error was detected. The affected blocks then repeat the calculation until the disturbance automatically disappears over time by electrical dissipation. The error is thus corrected and the calculation chain is liberated. This mechanism is made possible by the virtual insensitivity of these asynchronous circuits to the delays.

A non-volatile memory zone composed of MRAMs allows the outputs of each functional block to be recorded without errors. If an error does occur, magnetic tunnel junctions restore the last correct value of the previous level. Incorrect data are then deleted and the calculation can be performed again.

The functioning of the circuit thus resists an error slipping in at any level, without requiring a systematic resetting. The energy consumption of the circuit, as well as its size, is reduced in comparison to the current state of the art.

This work, conducted with the Laboratoire d'informatique, de robotique et de microélectronique in Montpellier and the CNRS, was supported by the Cnes.

Top page