I meant leave it engaged all the way down the runway.
That would be dumb since an autopilot might do all sorts of wild things when it realises it cannot control the plane (because its wheels are on the ground) so could drive the ailerons fully one way. It will definitely drive the elevator trim fully aft and that has been implicated in some crashes. Most GA autopilots have no “weight on wheels” sensing so have no idea that you have touched down.
But in the sim it remains airborne, at about 10-20’ then climbs away.
But in the sim, I don’t get injured or die when I crash.
It is worth remembering that what works at one airfield, or in one aircraft, might not work everywhere, or in every aircraft.
I would be wary of an autopilot at these heights. rule one of transitioning to landing in low visibility is to be stable and do nothing but the smallest corrections, most people tend to overcontrol when they see the runway sligtly off at RVR 550, let alone 200m or less.
The last thing you want is the autopilot making an input, and then you overcorrecting.
Is it correct on a 3D LPV the GPS has multiple database waypoints and is, as it wear, ‘sniffing’ its way down the approach like a bloodhound?
At the fly over missed approach point the SUSP alert comes on, and when pressed will load the missed approach. I was certainly not aware that the 3D waypoints of the LPV go beyond the MaPT, or provided the equivalent of a Go around function.
It would be interesting to understand the radio nav theory behind this A/P taking you down to 10-20 feet and then climbing away at some point on the runway.
The sim is not going to be simulating the autopilot performance.
I wonder if anyone knows specifics about the LPV guidance which comes out of a WAAS GPS. Start with what data the hex block contains. Robert may well have a point if there is no guidance in the hex block past the MaPT, then what decides what happens? The hex block is published in the AIP. Is there a certification requirement? @ncyankee might know. @avijake should most definitely know, or know someone who does.
If you look at the standalone GTN trainer, and fly a coupled LPV approach, you’ll see the same behaviour. The GP goes down, along and up.
I guess that this is Garmin programming rather than FAS Datablock data.
Probably worth investigating the principle behind this? 3D ends at the missed approach point, does Garmin then switch to LNAV/V+ guidance without an annunciation that LPV is no longer available?
Although the GP is supposed to stop at the MAP (normally 50’ above the threshold) it definitely continues guidance beyond that.
I have no idea whether that’s in the TSO-C146() spec or introduced by Garmin.
Does anyone know what the IFD does?
It’s down to procedure design and is not unique to LPV although GNSS procedure design terminology is different. For a GNSS approach the Missed Approach commences at the end of the Final Approach Surface (FAS). The FAS (normally) finishes at the a point where the Minimum Obstacle Clearance measured above threshold elevation is achieved (note, this is not necessarily above the threshold, it is just using the threshold elevation as a vertical reference plane). From here, the Initial Segment is a level segment (x) based upon 15 seconds flight time for the largest category aircraft at maximum Vat. The end of this segment (Zi) is where the Missed Approach Slope (Intermediate Segment) commences.
So, if we take a Cat E aircraft at maximum Vat of 210kts TAS, that gives a level segment of 0.875nm (5316ft) before the Intermediate Segment commences. There is a (hidden) waypoint within the database coincident with Zi.
I think the databases uses Zi in order to manage resultant climb gradients but, for that, I would have to read ICAO Doc 8168 Vol II which, on a Saturday afternoon, isn’t overly attractive to me. 
