GX data
Just an update on where we are with this GX-to-Trio issue.
1. The documentation for the GX series says that the "G" sentence in the MovingMap serial output stream gives "distance off course, hundredths of nm". We've recorded the serial stream and confirmed that this is not true. In fact, the distance is in units of anything between .01 nm and .0006 nm, depending on the current CDI scaling the GX is using. However, what CDI scaling the GX is using is not represented in the serial stream. This presents a problem for any autopilot that depends on this "G" sentence to track a GPS flight plan; the control gain effectively can change by a factor of 15 without the autopilot knowing it.
2. We've looked at some workarounds for this. For example, by looking at the distance to waypoint, bearing to waypoint, and desired track, it should be possible for the autopilot itself to compute cross-track error. This will be a low-precision estimate (especially when close to a waypoint, but even in general, in the GX stream, BTW and DTK are only in whole degrees -- though the documentation says in tenths of a degree, the tenths are always zero), but the idea is that it could be used, together with the "G" sentence, to estimate the GX's current CDI scaling. Unfortunately, this idea breaks down badly when the GX is doing turn anticipation (because then its reported DTK and the XTK in the "G" sentence are relative to a tangent to the anticipation curve connecting to the next waypoint, while BTW and distance to waypoint are relative to the current waypoint).
3. GX's with recent software have the option of Moving Map with Extended Data format, which includes information about the state of the GX's external annunciators. The annunciators encode whether the GX is in approach enabled or approach active mode, and these are the modes that trigger CDI scaling changes. However, the rate of scaling change with mode change is not documented; not all GX's have the Extended Data option; and the user can change the scaling manually, which wouldn't show up in the annunciator state.
4. Garmin AT isn't interested in fixing this problem. Ideally, they should then open the source, and publish the toolchain permitting modifying the software of discontinued avionics, if only for the experimental community. But that's another topic...
5. Recent software version GX's provide a GPSS data stream. The roll commands in this stream are sent at only a 1Hz rate, so it remains to be seen how well it can be made to work. But it seems at this point that it is probably the best option to get good flightplan tracking performance from the approach-capable GX's. Trio wants to do development of this on their new Pro Pilot platform, and then backport to the EZ Pilot if possible. I'll be working with them on that, and I'll let you all know how it goes.
--Paul