The single most important function of such a flight information system
is the display of an artificial horizon which allows the pilot to control
the aircraft attitude (i.e. pitch and bank). In a conventional attitude
indicator this information is provided mechanically by a gyroscope.
The ESFIS uses three identical solid state sensors to detect the
aircraft's pitch, bank and yaw (direction) separately.
The sensors are orientated at right angles along the three axes.

Consider the sensor orientated to detect rotation about the vertical
axis. The sensor output is actually proportional to rate of turn andnot to
absolute direction (the proportionality constant is referred to as
the ``scale factor'' and is nominally 25 millivolts per degree per
second). This means that when the sensor is static, its output is zero;
when the sensor is rotating clockwise, its output is positive; andwhen
rotating counter clockwise, its output is negative. One then integrates
the sensor output to determine an absolute change in direction.

The accuracy achievable with this method is determined mainly by how
accurately the scale factor is known --- other factors include power
supply stability, A/D converter errors, externally and internally
generated noise, sensor non-linearity, etc.

The sensor's uncorrected directional errors accumulate at the rate
of between 70 and 300 degrees per hour, i.e. after one hour of
operation the difference between the true direction and indicated
direction is typically between 70 and 300 degrees. When installed
within the current production instrument, the onboard
software corrects for a variety of errors (mainly temperature
dependence of the scale factor) so that the directional errors
accumulate at the rate of only 15 to 30 degrees per hour.

The purpose of this project is to identify and quantify the remaining
sources of error with the goal of reducing the accumulated error rateto
less than 8 degrees per hour.

An ideal deliverable would be a list of suggested means for reducing
accumulated error, together with evidence (analytic or simulation)
of the effectiveness of these means. Ideally, the means would involve
only improved software algorithms.

This summary prepared with the assistance of K. Devarasettyand G. Watson
of B. F. Goodrich Aerospace.