Some ILSs are better than others. Also, with regards to AP performance, wind conditions play a certain role. I know it shouldn’t be done, but very recently, I have flown an ILS, on the autopilot (DFC90) to about 6-8 feet or so off the ground. I am quite sure that if I had trimmed the plane (SR22) for 70 knots and kept going, it would have resulted in an acceptable landing. Poor man’s autoland, so to speak… 
I don’t know what it is but some people just don’t have to patience to wait it seems.
And some don’t have the imagination to understand that beeing dead is the maximum delay. I met many of those too. And one of those on my right seat when i was still a beginner (“just fly on, i have an IR rating”) almost killed me. I decided to not listen, turned around in the snow storm, landed at some remote airport and took the train home. It was also the last time I flew with that guy.
Most of the time when the visibility is low due to fog/mist, the wind is not strong or even virtually non-existent.
Most of the time when the visibility is low due to fog/mist, the wind is not strong or even virtually non-existent.
Except at some coastal airfields, like St Mawgan or Jersey
Talk of coupling GA autopilots to (near) touchdown unnerves me. There is no way that you can get sensible G/P info below about 35-40ft (you can’t change physics ;) ) and consequently, anyone letting their GFC700/DFC90 or equivalent fly the flare is asking for trouble, as by this point the autopilot is chasing electronic nonsense.
For information, we operate with a calibrated Radalt and DGPS/SBAS navigation system that is permanently accurate to <20mm. We never let the autopilot fly the approach as it is a simple single unit which only tracks the relevant signal in space, it does not have any roundout/flare function and is not failsafe (never mind fail active/passive). I don’t think there is currently any system out there which provides an autoland (Cat IIIa or IIIB) function for GA aircraft.
That “poor man’s autoland” comment was a bit tongue in cheek.
But you can believe me, I was no higher than about 10 feet when I sensed that the AP was getting a bit twitchy and then I disconnected. But up to that point, it was rock stable. This was a CAT2 installation. Thinking about it, yes, it would probably not work on the last 10 feet, since the signal probably disappears and then then autopilot might do nasty things.
Anyway, as I said myself, flying that far down with the AP coupled is not something one should do on a routine basis.
The LOC should not ever disappear, because Cat 3 autopilots track the LOC all the way to the runway. That’s why the LOC transmitter is at the far end of the runway, otherwise you would lose the signal as you cross the threshold 
The GS transmitter is part-way down the runway, near the touchdown zone I think, and is offset to one side of the runway (obviously, otherwise you would crash into it ) so is not usable all the way down to touchdown. Cat 3 systems use a radar altimeter below about 150ft to generate a virtual glideslope effect (so the ground before the runway has to be suitably smooth for Cat 3 to be allowed) and then there is a fairly primitive software function which does a flare, so the landing gear doesn’t get smashed up. There is no optimisation to achieve a perfect touchdown of the sort a good human pilot tries to do, and can do on a good day. Cat 3 landings are therefore often very rough.
A good GA autopilot should track the LOC well enough to the runway (mine certainly does; I have tested it) but it won’t do a flare, partly because there is no function to do it and partly because the GS signal is disappearing. Such a landing would be an emergency procedure when you have run out of options and you are on your last ILS and there is no more fuel to try elsewhere. The hope is that the conditions will not be exactly zero-zero and that you will see the runway before you hit it. If you get visual at 100ft then you can land manually. In real zero-zero you will almost certainly smash up the landing gear, but if you are out of options, that’s better than killing yourself.
This is why fog is a really bad thing, and worse because nobody can forecast it. It tends to be forecast after it has formed at the airport just upwind along the coast I have diverted a few times and all cases were due to basically IMC all the way down to the ground (call it what you like) and all of them were unforecast.
Of course we are all talking theoretically but the there is no such limitation with LPV.
I don’t think there is currently any system out there which provides an autoland (Cat IIIa or IIIB) function for GA aircraft.
I think our best bet would be a HUD/EVS system allowing for CAT II approaches. Don’t think there is a future for more advanced autopilots in our spamcans…
There is no way that you can get sensible G/P info below about 35-40ft
Well, sure, because at that point you’re buzzing by the glideslope antenna! The GS antenna is somewhere between 7 and 18m high, that is 23 to 60ft. But at 40ft, you’re only 800ft away from touchdown, and since the center of the touchdown zone is 1000ft from runway end, you’re already above the runway.
Now assumed you’ve slowed down to 90kts, assuming no wind, you’ll be making 455fpm descent rate, that is 2.3m/s, which is below the design vertical rate for impact of 10ft/s (3m/s).
So once you loose the GS, you can just continue with your vertical rate until impact, and the aircraft should be able to withstand the landing.
