In 1993, Racal Decca HK generated a radar recording during a voyage on
a High Speed Craft(HSC) between Hong Kong and Macau. We were given the resulting tape for
investigation purposes. The prime objective of this exercise was to capture various
elements of the signals that a radar receiver would need, with the intent to playback such
a record subsequently. However, GPS, fluxgate compass and speed log signals were also
captured and subsequently analysed. Finding a reference system to measure GPS accuracy is
by its very nature difficult. Hence, in this case, an unusual artifice has been used to
get an idea of comparative accuracy over a segment of the voyage. This involves working
out the course between GPS positions at 10 second intervals. As the craft velocity was
about 15 metres per second, the distance travelled between such positions is about 150
metres. This is relatively large in relation to expected GPS error. Hence, the difference
between this and the heading recorded by the fluxgate signal is a simple indicator of
relative accuracy. As the craft are constrained between the narrow lanes, heading
deviations are few and small, and fluxgate stability is generally satisfactory. The
results appear to indicate that SA may not have been activated on that particular day.
Hence the larger errors would appear to accrue from other factors.
Figure 1: Difference between Computed Course (10s period) and Compass
Course
In Figure 1, the differences between the flux gate course and the
course obtained between GPS observations at 10 second intervals is plotted, against the
longitude. From this it can be seen that while the normative difference, such as in the
range between 113°55‘ E and 113°59‘ E， is less than 10 degrees, the differences at
other times can be greater than 30 degrees.
In Figure 2, the absolute value of this difference is represented by
the major axis of an ellipse. Some relationship with proximity to land can be seen. It is
important to appreciate that the large ellipses represent a continuing error in positions
as the course between positions at successive intervals of 10 seconds is compared against
the course at one point in time. No attempt was made to average the courses during the 10
second intervals as there were relatively few changes of course during the period.
