Originally Posted by FullWings
If you (as a controller) have sent two aircraft directly at each other and there is no further intervention (from you, pilots, TCAS, etc.) then they're going to collide.
Only if
- they're going directly head-on
- both are flying very accurately.
The latter is usually true these days with high-precision navigation, although some oscillations around the exact altitude are normal for every feedback-control loop. (I don't know about other types, but Airbus FBW aircraft's autopilots have a "soft-altitude" mode for cruise that allows up to 50ft deviation to reduce thrust changes for comfort and economy.)
As to the first point, of the two recent Mid-Airs, only in one were they going in directly opposite directions (GOL/ExelAire), in the other (Bashkirian/DHL)they were at around 90 degrees, and for a collision in that setup you have to get it exactly right, otherwise they'll miss.
Not what I said. What I am saying is that the penalty for assuming that TCAS has got it wrong (and being wrong in that assumption) is much higher than for assuming ATC has got it wrong (and being wrong in that assumption). Apply a bit of logic to that and you can see why I'd follow an RA. It's a sort of Pascal's Wager for aeroplanes...
I'm curious what statistic this assertion is based on.
The penalty for both might be a Mid-Air. The DHL crew in the Ueberlingen accident followed the RA to the letter. And were rewarded with a collision.
Consider the Ueberlingen scenario with a slight modification: The Tuplolev has, as allowed for dispatch, an unserviceable TCAS system, but fully working transponder.
Thus, the DHL Boeing gets a resolution advisoary, but the Tupolev gets none. The Tupolev does get, however, a manoevering instruction from ATC ("descend FL350 expedite", iirc). The difference here being between not following an RA versus not getting it. Educate me about TCAS: would the DHL crew have known about the Baskirian's TCAS being inoperable?
Bernd