|
I guess it is just my naiveté but I am just not familiar enough
with what it takes to get a TSO rating on an instrument except that once it gets
it, the price goes way up
Joe,
I think the issue is that there
is a statistical interaction
among a number of variables totally out of your
control.
Testing tries to capture a larger portion of the
statistical
population so that you aren't dead because a few of
those independent variables aligned one day in a
way
that will result in malfunction.
Is there a vibration in your
particular plane
that the mechanical isolation the ahrs doesn't
damp out sufficiently?
Is there electical noise in your
airplane
that is somehow different enough to affect
your unit?
How does the unit respond to
static
discharge?
How does the unit behave if
there is a large
current near by?
Is there an attitude excursion
that is beyond
the capacity of the unit to
track?
Will you do something in your
airplane that
will cause the micro-code in the unit to execute
a code path that no one else has?
Is there a marginal wire
connection in the
chip that is going to turn into a fuse over
time?
The problem with these kind of
statistics is that
even though the failure mode may only occur 1%
of the time, when it does occur and you are
out
of backups, you are having a 100% bad
day.
It's not like you have a 99% good day every
day.
I would just note that the
cross-bow 425 was
built by an experienced company and it
failed;
not 100% of the time but enough that people
eventually walked away.
The garmin gmx200 - another
piece of digital
avionics - is on hold right now because it
"sometimes"
powers itself down. I bet that unit got
more
testing than most of these low unit volume
ai's.
The stec 55x auto-pilot has so
much misbehavior in
the field that Tom West got 300 replies when
he
made a post about pitch oscillation problems in a
c210.
IMO the state of avionics in the
field is that
certified avionics "almost work", in a
statistical
sense, leaving anything less well tested
somewhere
south of that.
Is an old vacuum gyro more
reliable? maybe not by
design but by shear numbers it's been
exposed
to a much richer set of statistical variables
without
doing something weird.
|