View Single Post
  #4  
Old February 22nd 04, 05:10 PM
Robert M. Gary
external usenet poster
 
Posts: n/a
Default

Jonathan Goodish wrote in message ...
In article ,
(Robert M. Gary) wrote:
I think you are missing something. If both AIs start showing different
results you will be suspecious and start asking questions. As I noted
a realistic AI failure situation can be very difficult to detect.


How is this fundamentally different than if one AI disagrees with the DG
and turn coordinator?


Because, a real world situation is that the AI dies very, very slowly.
There isn't that immediate "hmm, somethings wrong" like with a dead
vac (alarms going off etc). The TC has enough bounce in it that there
is really no way to detect small differences like 5 degrees when you
are in real IMC. Once you start bouncing around in the clouds the TC
starts bouncing back and forth. It will keep you upright but is far
from close enough to determine 5 or perhaps even 10 degrees off. The
DG you may or may not notice. In anycase two AIs being off is a pretty
quick and certain thing to notice.

Couple that with the fact that you have to decide which is right, the
TC or the AI. If you have 3 its easier to pick on that might be bad.