Brent posted a question concerning what
happens to the solution out of the Xbow when GPS is lost, and doesn’t the
presence of the GPS imply that it is needed.
Having managed the development of a
GPS/Mems IMU navigator for an unnamed military airborne platform I have some
background in the subject. Just having some background does not provide all the
answers, as the “graceful degradation” of the solution when GPS is
lost is dependent on the implementation. Xbow is your best source for
definitive information, but I don’t know how willing they are to share
this information, as I’m sure they spend big bucks developing and
refining their implementation.
GPS and IMU (inertial measurement units –
fancy name for a package of gyros and accelerometers) have different and
complementary strengths and weaknesses. GPS primarily measure range to the
satellites, with relatively small “bias errors”. Their weakness is
a lag in response to dynamics (turns are the primary issue) and relatively high
random errors. IMU respond quickly to dynamics, but have many error sources,
most of which are slowly varying biases (I know-that’s an oxymoron)
Almost all integrated navigators combine the
measurements in a Kalman filter. This is a software implemented data filter
that estimates the best fit solution (State Vector (SV) in the parlance of the industry)
to the measurements is receives. The SV out of the Xbow includes attitude
(pitch/roll/heading) and velocity (speed, VSI and drift) and time
If the measurements are perfect then the
solution is very good, but the measurements are never perfect. Kalman filters
do a good job to correcting random errors, but bias errors lead to errors in
the SV. To counter this most Kalman Filters also have estimate several bias
errors “states”, mostly attributed to the IMU. These can include accelerometer
biases, gyro drift rates, turn-on-to-turn-on errors, ….. The list is much
longer than I can remember. The GPS measurements allow the filter to estimate
these error states. That is one reason we wait before moving the aircraft, and it
take some time and software to be able to reinitialize in flight. The
instruments have to stabilize, and the filter has to eat a sufficient number to
measurement updates to allow the “error states” to stabilize.
If the GPS goes away in flight many
of these estimates continue to correct the IMU measurements, allowing for a
relatively good solution for some time. (You should get a GPS failure warning to
tell you to transition to the back-up instrumentation.) Just how good, and for
how long, is the implementation dependent variable.
Brent’s suggestion to disconnect the
antenna in VFR conditions and watch what happens is a good one. Be away that
IMUs exhibit “g sensitivity”, and acceleration due to turns and
speed changes will affect the IMU measurements and therefore the SV.
Perhaps Mike at Crossbow can provide some
data on performance after GPS failure in flight.
I don’t know if this is any help, or
just the ramblings of a retired engineer. As Grayhawk says…”use at
your own risk”
Paul Bricker
N63PB
Trying to get into primer