FWIW, in NAV mode, the KFC225 uses the CP as the heading to initially fly, and then it uses the deviation bar signal to finely adjust that heading, and that is how it does the wind correction. All the “classic” autopilots work that way.
If you are flying an ILS in HDG mode (which is how an ILS is traditionally flown in the final stages, pre-intercept, vectored or not) then the CP doesn’t matter. But if you are in NAV mode then the CP does matter because now the CP acts as the heading bug for the AP.
The DFC 90 will probably be similar – there are only so many ways to play this stuff.
That said, I have never flown an ILS in NAV mode so haven’t tried this.
With EHSI auto-slew (which I have, though obviously it works only with a GPS nav source, which one is highly unlikely to be using when intercepting a LOC) the CP could be pointing somewhere undesired – another reason for flying the whole thing in HDG mode, and then you don’t get the “WTF is it doing now” moments.
in NAV mode, the KFC225 uses the CP as the heading to initially fly, and then it uses the deviation bar signal to finely adjust that heading, and that is how it does the wind correction
I thought so, too but why did it continue going off course while the localizer deviation was already on the left and further travelling out?
That said, I have never flown an ILS in NAV mode so haven’t tried this.
You’ve never flown a coupled ILS?
I thought so, too but why did it continue going off course while the localizer deviation was already on the left and further travelling out?
the autopilot, as Peter mentioned before, needs the ILS inbound track on the CP to calculate the track angle error, e.g. for x-wind correction while approaching and tracking the localizer. I guess with a heading about 40°less than the tuned-in CP value and drifting right off the localizer the autopilot refuses to steer even more to the left or is eventually limited in the correction value at around 40°. Maybe the A/P follows a “gaining” programm after the LOC was centered, thus reducing the available maximum steering signals.
You’ve never flown a coupled ILS?
Of course – the APR mode.
I don’t know the DFC90 which should be working laterally wholly via ARINC429, but again speaking of “classic” GA autopilots, the CP value (supplied to the AP in XYZ format) is the heading the AP turns on to upon LOC intercept, and then it feeds the deviation bar error into a PI (not PID AFAIK) control loop to generate the wind correction. The P (proportional) gain is not too big (for easy stability) and they rely on the I (integral) term to gradually take out the wind drift. Generally this works well and that is how everything in the last few decades has been flying.
My guess is that the DFC90 is the same except that the control loop has an extra input which comes from the ADC (air data computer) code and which will try to achieve a slicker intercept. After all, if you are tracking an ILS you have to be tracking an ILS, not some GPS track, so there aren’t that many options on how to play this.
So the CP has to be correctly set up. If it isn’t correctly set up (and obviously it doesn’t have to be exact otherwise setting 271 instead of 270 would be a problem) you just take longer to take out the wind drift.
Maybe the A/P follows a “gaining” programm after the LOC was centered, thus reducing the available maximum steering signals.
I would think there is a number of optimisations like that in the software. Most control loops in real world applications put a limit on things – especially how much you let the integral term integrate itself up or down, if the error is not going down fast enough.
too but why did it continue going off course while the localizer deviation was already on the left and further travelling out?
Would you be able to post the GPS track, with some text on what you recall happening at different places long it?
Thank you, that makes sense.
Now there is one question left: how does the AP know when it is on the radial corresponding to the course pointer setting? The localizer is not a VOR and does not transmit radials. Or does it treat the course pointer as a heading solely? I was very surprised by the APs lack of response to the deviation bar once it crossed the ILS, with the wrong course pointer it should have looked like a very strong cross wind and I would have expected a strong correction.
Last check ride my examiner said I shouldn’t play with avionics when in IMC and on a procedure and I agreed. Yet once again I have done it. The curse of being a computer guy…
how does the AP know when it is on the radial corresponding to the course pointer setting
When the localiser deviation is 0.
The problem in your case is that the AP is looking at two inputs to its error function: Localiser and Course error. If the course is set wrongly, that error can easily dominate and it will turn toward the course rather than the localiser.
I suspect that localiser error approaching 0 was the trigger for it to leave heading mode, at that point it looked at localiser error (small to the right) and course error (large to the right) and decided to turn right.
As you blew through the localiser, the I term in the controller was rapidly increasing (due to large course error to the right) and the initially small localiser error had little effect.
If you had let it carry on, it would have turned most of the way to 300 degrees.
Eventually, the localiser error would have built up enough to dominate, and it would have swung back left.
I doubt it would manage anything like a stable intercept though, the high errors pulling in opposite directions would tend to destabilise the control loop, the limiting on the I term and max course error would have kept it from flying all over the place, but it would follow some wobbly path between 258 and 300 until it lost the localiser.
I think this is because the autopilot uses two signals to fly the ILS – localiser deviation and course deviation
Having looked at my KFC225 wiring, I think the above is correct, but what is happening is that the comparison between the CP setting and the actual heading is being done further back – in the HSI/EHSI system. So, in effect, the AP still “knows” both of these and will keep turning the aircraft (at the maximum configured roll rate, say 20 degrees) until the difference is minimum.
The problem in your case is that the AP is looking at two inputs to its error function: Localiser and Course error. If the course is set wrongly, that error can easily dominate and it will turn toward the course rather than the localiser.
Some autopilots use the heading bug for this instead of the course setting. But maybe that’s only old stuff like the Piper Autocontrol?
In HDG mode, the heading bug is fed to the AP as a left/right voltage. That mode is very simple.
If you have an external roll steering converter, that generally fakes a heading bug so again this is fed to the AP as above, and the AP can fly anything the GPS tells it to, DME arcs, you name it, all in HDG mode.
I reckon the DFC90 is controlled laterally the whole time via ARINC429, and doesn’t give a damn about anything like the CP or the HSI heading bug, so the AP itself is probably the wrong place to look if it is behaving strangely.
I would look at how Achim’s Aspen EFD1000 is doing all that stuff…
If the course pointer is off in the G1000 it will also not capture and track the LOC correctly. Had this once when it didn’t auto slew to the correct inbound course.
