Why is LPV "more stable" than an ILS?

Why only on an LPV approach? The S-turns you describe are presumably only in the lateral plane so roll steering should be available when following any track; an approach or not.

My observation is that the GFC700 uses standard CDI based control mechanisms and does not take advantage of roll steering. My observation also is that it does take advantage of roll steering with the GPS approaches. My guess for the discrepancy is they are probably restricted by TSO compliance on the ILS because there is no technical reason why both inputs could not be integrated into the localizer tracking algorithm. Even if the algorithm used the localizer as the definitive source for tracking, the detection of the drift could be enhanced by using the actual track information anytime it was available to be used, with the algorithm reducing to just the standard rate of CDI deflection when suitable GPS track wasn't available, so the tracking would be no worse than what already exists and be enhanced to the extent that the GPS can give a heads up before there is a measurable CDI deflection to act on.

Peter,

I described the observed tracking difference using a GFC700 which is a highly integrated system. With other legacy autopilots, tracking using either a roll steering adapter is typically thru the heading error signal interface. These autopilots can't track the lateral and vertical (ie approach mode) and be simultaneously in heading mode, so regardless if the approach is an ILS or a LPV, the tracking is the same and the same type of S turning will result. Even with the autopilots that directly support roll steering are used, for example the Stec 55X or the KFC225, the roll steering is mutually exclusive with approach mode and vertical guidance. The Avidyne DFC90 unit is a new generation digital autopilot and can use the digital interface for both vertical and lateral that is available from the WAAS GPS, but there isn't an equivalent for the ILS in the ARINC labels which are still CDI displacements.

How is that different from GPS i.e. I am 1 mile left of the track?

If the CDI signals are the same and they are what is used to control the lateral path, there would be no difference, ILS or GPS. The CDI signals are crude. All the autopilot knows is the HSI course selector has determined the no wind direction that the autopilot assumes it will follow. Assuming that an intercept to the final approach course is performed using a heading input until the CDI is midscale, all the autopilot can use is the rate at which the CDI is moving to judge when to turn to join the approach course. It will use the course pointer to determine the initial course. Only one of three things can happen at this point, the CDI can remain in the center, move to the left or move to the right. If it remains in the center, the autopilot will determine it needs no wind correction. If the CDI drifts off of the center, the autopilot can use the rate of movement to determine how much off the course it is and remember a wind correction factor once the needle is back in the center. However, the autopilot must first make a turn greater than the wind correction to get back on course and once back on center, take out all of the intercept except for the calculated wind correction. Since the localizer is not always steady, it must allow for some drift before it makes the determination, in other words it must filter out small deviations. If this doesn't hold the course, the process is iterated until it does. Then on most autopilots they desensitise the tracking as the course gets more and more sensitive closer to the runway. If the winds change, the drift will occur again and it will get further off before it corrects back due to the lowered sensitivity. Under some conditions, the autopilot is not able to settle down on the appropriate wind correction and can end up behind the situation.

With a more sophisticated autopilot, the GPS offers a different method of tracking than CDI tracking. The GPS knows the GS, track, desired track, and cross track and can detect deviations in the track long before there is any noticeable CDI deflection. The roll steering output is a precise calculated bank angle that the autopilot should use to correct this error. Since any crosswind drift will show up in the track measured by the GPS, it can immediately be corrected and even very small cross track errors can be continuously adjusted as needed. Corrections of 1 or two degrees are not atypical for this method of tracking. Using roll steering to track the GPS path is smooth and almost unnoticeable to the pilot. You will rarely ever see the CDI deviate from the center, much less a full needle width.

When flying an autopilot that can track the ILS and has a HSI that uses the course pointer to establish the no wind heading, I have been able to monitor the track using the GPS and make small course pointer corrections when the GPS is indicating that the CDI will drift off center. This can be done before the actual CDI indications show up. If the HSI is an EHSI with auto slew, this doesn't work.

Incidentally, does a WAAS box have any manual control over the FSD?

Yes, but I have not found a reason to use it other than to increase the tracking sensitivity enroute with an autopilot installation that doesn't have roll steering integrated or an add on adapter. I would much more likely just use the heading mode of the autopilot and make any minor adjustments based on the GPS DTK, TRK, and XTK, as it is usually much smoother.

one can set the FSD to 5.0 1.0 or 0.3nm, and this is useful in other contexts e.g. using the OBS mode to fly an NDB or VOR approach for which there is no GPS overlay

With the GNS and later units, almost all the NPA and ILS procedures are in the database. In the US, a part 91 operator may not use the non overlay VOR/NDB approaches to conduct these procedures unless using the ground navigation equipment that provides the procedure final approach lateral guidance. This is almost never a real limitation as almost every runway has a RNAV straight in. We can substitute for ADF and DME using the GPS, but not for the primary course guidance, unless the procedure has GPS in the name. The GPS is only available for situational awareness and the pilot has to acknowledge this if they load such a procedure (ILS, VOR, NDB, ...). I call this the lawyer message. The sequencing to the FAF uses Terminal CDI FSD (+/- 1.0 NM) and when the FAF is active, it switches to +/- .3 NM. There is no prohibition if you monitor the ADF, VOR, or localizer on the appropriate indicator type and as long as the indications are within bounds, you can use the GPS for situational awareness. I don't have a ADF installed in my airplane, so I can't do the NDB approaches, but I do have dual VOR and would use my HSI and GNS530W coupled to my autopilot while my number 2 CDI would be displaying the VOR CDI.

I have no idea if this is true for the GNS boxes but the KLN94 database is bare for "overlays" in Europe relative to the USA.

The GNS boxes have all approaches for Europe, at least I've never found one missing. They show the same warning that you're not supposed to fly those overlays as NCYankee mentioned.

Also, I've never had a case where a published SID/STAR or an IAF was missing from the database.

What I did notice is that my GNS430W would not give me those LNAV+V approaches (the artificial glidepath) that I should get for almost all airports with non precision approaches. However, that was with SW 3.0 whereas I now have the update to 5.0. Maybe that has changed, I haven't had the chance to try it (when I wanted to do this at EDTY, the GNS offered LPV which was of course for more interesting to try out).

I have no idea if this is true for the GNS boxes but the KLN94 database is bare for "overlays" in Europe relative to the USA.

I just checked the GNS/GTN/G1000 database for UK and it seems it includes ILS, NDB, and VOR procedures that don't use GPS. For instance, EGKA shows RNAV 2/20 and NDB/DME 2/20.

The GNS boxes have all approaches for Europe, at least I've never found one missing.

Here in the US, there are a few weird approaches that are not in the GNS/GTN/G1000 database. There is a DME arc at Martin State that the arc goes all the way to the MAP. It isn't included because it can't be coded with ARINC 424. Any procedure that includes a RF leg is not included. All RNAV (RNP) procedures are not included (note the GNS/GTN/G1000 AFMS does not support these approach types) and only authorized users are permitted to fly them. There may still be a few procedures that are different based on the class of aircraft (piston verses Turbojet) that only the Turbojet version is coded. That still leaves the vast majority of procedures in the database.

I don't think current LOC tracking uses any integral term, but I really have no idea. I am sure others have been here before...

My Century AP from ~1971 does have an integral term, assuming there's still some electrolyte left in that capacitor...