Mirror Glideslope

RotorAndWing

Final Approach
Joined
Sep 5, 2008
Messages
8,496
Location
Other side of the world
Display Name

Display name:
Rotor&Wing
Report: Air France A343 at Paris on Mar 13th 2012, intercepted mirror glide slope, large pitch oscillations and approach to stall



An Air France Airbus A340-300, registration F-GLZU performing flight AF-3093 from Bamako (Mali) to Paris Charles de Gaulle (France), was descending towards Paris for an ILS Category III approach to runway 08R but went significantly above the approach profile despite descending with speed brakes deployed. The approach was continued. When the aircraft was 9nm before the runway threshold, the aircraft had been established on the localizer however was at 4950 feet MSL about 1750 feet above glideslope, target speed selected at the FCU was 183 knots, the flaps were selected to position 1. The crew slowed the aircraft and selected flaps to position 2, at 4nm before touchdown the aircraft was at 3700 feet MSL and 2100 feet above glideslope, the glideslope indications in the cockpit had already reverted to the side band of the glideslope transmitter (mirror glideslope at 9 degrees), about 30 seconds later the crew selected the gear down. At about 2nm out, the aircraft descended through 2850 feet MSL 1600 feet above glide, the vertical channel of the autopilot mode change glideslope capture and autothrust changed to speed mode. The aircraft pitched up from about 1 degrees nose up to 26 degrees nose up, the vertical speed changed from -1600 fpm to +3300 fpm, the airspeed dropped from 163 KIAS to 130 KIAS. When the aircraft rotated through 26 degrees nose up, the crew disconnected the autopilot and pushed the side stick forward to near the mechanical stop, the pitch attitude and the rate of climb reduces, the aircraft began to accelerate to 143 KIAS again, the autothrust system disconnected. About 30 seconds later the crew re-engages autopilot 1 and autothrust in climb thrust with the intention to perform an automatic go-around, the autopilot's lateral channel engages in LOC, the vertical channel in G/S, the autothrust in SPEED, the aircraft crossed the runway threshold at 2700 feet MSL (2370 feet AGL), the pitch attitude reduces to 5 degrees nose down and the aircraft rapidly descended. The pilot monitoring recognized the modes announciated were not suitable, disengaged the autopilot, placed the thrust levers into the TOGA detent and the aircraft entered a normal go-around at 2000 feet MSL. The crew subsequently positioned the aircraft for a second approach and landed without further incident.

The French BEA released their final report in French (English version released on Sep 13th 2013) concluding the probable causes of the serious incident were:

- the inadequate monitoring of the aircraft's trajectory by both air traffic control and the crew while conducting a CATIII approach under radar guidance

- the decision of the crew to continue the approach although the aircraft had crossed the final approach point above vertical profile

Contributing factors were:

- the absence of visual cues on the radar screen of air traffic control when aircraft intercept the vertical profile below 5000 feet MSL

- the inappropriate methode of intercepting the glideslope from above used by the crew

- the capture of a side band/mirror glideslope by the autopilot which resulted in an excessive pitch increase

- fatigue by both controllers and crew which may have contributed to the serious incident


The BEA complained that the recordings of the cockpit voice recorder had already been overwritten when the BEA was notified of the incident. The investigation such had to rely on data off the quick access data recorder, radar data, the ATC recordings and testimonies by controllers and crew.

The aircraft was maintaining FL090 about 30nm before the runway threshold with the autopilot's lateral channel in HDG mode, the vertical mode in ALT and autothrust in speed mode maintaining 250 KIAS, the crew contacted approach at that time and was cleared to intercept the localizer 08R followed by the instruction to descend to FL080. At about that time the aircraft crossed the vertical profile from below to above the profile. The aircraft was subsequently cleared to descend to FL060, the autopilot's vertical channel and autothrust engage in OP DES mode, the lateral channel captures and tracks the localizer. The aircraft was at 7220 feet MSL about 1275 feet above vertical profile at 17.5nm before the runway threshold, at that point the controller instructed to maintain a speed of 200 IAS or greater, the aircraft is still doing 250 KIAS. The crew inquires for further descent, the controller apologizes for forgetting and clears the aircraft to descend to 3000 feet and the ILS approach runway 08R. The crew selected 220 KIAS and 3000 feet into the flight control unit (FCU), the autoflight systems maintain the open descent mode. The speed as well as the rate of descent reduce (tracking selecting speed has priority) resulting in an ever increased deviation above the vertical profile, the crew applied speed brakes. After the aircraft had slowed to 220 KIAS the rate of descent increased to 1840 feet per minute reducing the deviation (to follow the 3 degrees glidepath a rate of 1100 feet per minute would be needed).

When the aircraft was 10nm out descending through 5500 feet the controller instructed to maintain a speed of 160 KIAS or greater and handed the aircraft off to tower, but did not inform tower that the aircraft was above the vertical profile. The crew selected 183 KIAS target speed and selected the flaps to position 1, once again the rate of descent is reduced by the autopilot and the deviation from profile increased again resulting in the sequence of events already described in the initial paragraph.

The BEA reported depending on the altitude assigned by ATC to intercept the ILS the final approach point is 14nm out at 5000 feet MSL, 11nm at 4000 feet MSL, 8nm at 3000 feet and 5nm out at 2000 feet MSL. The radar screen of the controllers provide a visual cue only for intercepting the final approach at 5000 feet although controllers are permitted to assign other altitudes to permit separation of simultaneous parallel approaches. The ATC procedures require, with low visibility procedures in use, that the aircraft should reach a speed of 180 KIAS or less at 15nm before touchdown and a speed of 160 KIAS or less before intercepting the glidepath.

When the autopilot activated the capture glideslope mode, the aircraft would already have needed to descend at an angle of 10 degrees to reach the runway and was consequently already above the 9 degrees side band of the glideslope transmitter, in which the deflections would be reversed by design, above 9 degrees angle the indications would demand an upward correction causing the autopilot to command the large pitch increase until the crew disconnected the autopilot.

5 safety recommendations were released as result of the investigation.
 
I find it odd that the autopilot is not listed as a contributing factor. Perhaps the Airbus Flight Control Laws need an update or something. Given this and AF 447. Seems either the crew are clearly not setting the ap correctly or they are and the Airbus flight control laws aren't working. Certainly seems to be the case here.
 
I find it odd that the autopilot is not listed as a contributing factor. Perhaps the Airbus Flight Control Laws need an update or something. Given this and AF 447. Seems either the crew are clearly not setting the ap correctly or they are and the Airbus flight control laws aren't working. Certainly seems to be the case here.

The airplane did what the crew commanded in both instances.
 
Better than being below the GS and not realizing it, I guess....
 
I wonder if they were aware the mirror glide slope existed, I wasn't.
 
The airplane did what the crew commanded in both instances.


The autopilot manual (Altimatic IIIc) for the 'kota makes it clear that glide slope interception must be from below because of the problem of multiple glideslopes. The autopilot will not accept an intercept from above. Too bad Airbus doesn't use similar logic. Too bad the crew training didn't include that glideslope intercept must be from below. I had that training and I was my CFII's first student.
 
Some autopilots will intercept from above and some won't. I was aware false glideslopes existed and saw it for myself in KJAC.
 
The airplane did what the crew commanded in both instances.

It seems like the biggest failure here is to brief the approach and then fly it according to the brief. They should have realized something was amiss if they just read the frucking plate.

I wonder if they were aware the mirror glide slope existed, I wasn't.

I see it all the time in PHX. I get slam dunked into there (VFR), often on a 3 mile final at 3000 ft AGL.
 
I wonder if they were aware the mirror glide slope existed, I wasn't.

I was taught that these mirrors were why I should intercept an ILS glide slope from below while staying above whatever minimum altitude I should fly above. I know someone will come up with exceptions, but the plane is usually below the glide slope during the intermediate approach segment, in my very limited experience.

R&W- thanks for the reminder about these false glide slopes.
 
Last edited:
I was taught that these mirrors were why I should intercept an ILS glide slope from below while staying above whatever minimum altitude I should fly above. I know someone will come up with exceptions, but the plane is usually below the glide slope during the intermediate approach segment, in my very limited experience.

R&W- thanks for the reminder about these false glide slopes.

I was probably taught about them 20 years ago and have since forgotten, I didn't forget about intercepting from below. Getting old is hell lol.
 
The airplane did what the crew commanded in both instances.

Aren't the flight control laws supposed to prevent the AP from stalling the airplane even if is commanded to do that? In this case the crew initiated the AP & AT were initiated and caused a stall.
 
Aren't the flight control laws supposed to prevent the AP from stalling the airplane even if is commanded to do that? In this case the crew initiated the AP & AT were initiated and caused a stall.

Where is there anything in the description that the aircraft stalled?
 
It seems like the biggest failure here is to brief the approach and then fly it according to the brief. They should have realized something was amiss if they just read the frucking plate.



I see it all the time in PHX. I get slam dunked into there (VFR), often on a 3 mile final at 3000 ft AGL.

*At 4,500* "Cessna 12345 proceed direct numbers runway 25L, keep your speed up Boeing 737 to follow you. Contact tower"
Yep, I've gotten that one there before too :D But I've never gone in IFR, I don't know if you'd see a "mirror" glideslope from that high above or not. I've never experienced one.
 
Not trying to be a Monday Morning QB, but it looks like there was some complacency/lack of monitoring and too much trust in George. To me, you have to be asleep to let the nose get 26 degrees high on an approach before overriding.
 
So, I'm assuming that these artifact glide slopes still come in to the antenna creating a steeper glide slope. Given that, would the CatIII system still have been able to execute a successful landing arresting the higher sink rate at the bottom?
 
So, I'm assuming that these artifact glide slopes still come in to the antenna creating a steeper glide slope. Given that, would the CatIII system still have been able to execute a successful landing arresting the higher sink rate at the bottom?

Remember the reverse sensing-the signal indicating one is too low on the glide path is often above the signal that indicates one is above the glide slope. Tat would probably mess up a program.
 
*At 4,500* "Cessna 12345 proceed direct numbers runway 25L, keep your speed up Boeing 737 to follow you. Contact tower"
Yep, I've gotten that one there before too :D But I've never gone in IFR, I don't know if you'd see a "mirror" glideslope from that high above or not. I've never experienced one.

It's a bit harder in a PA-31 than a Cessna, but still doable. I plug the ILS in regardless just for backup, VFR.
 
Remember the reverse sensing-the signal indicating one is too low on the glide path is often above the signal that indicates one is above the glide slope. Tat would probably mess up a program.

Well, as I read the description, it was closing the gap above glide slope as it was coming down.:dunno:
 
An approach to stall and stall are not the same thing.

I should have taken your avatar more seriously when responding to you. The point remains that the airbus flight control laws are created to prevent such a situation as an "approach to stall.":rolleyes2:
 
I should have taken your avatar more seriously when responding to you. The point remains that the airbus flight control laws are created to prevent such a situation as an "approach to stall.":rolleyes2:

The software did what it is supposed to do. The alpha floor protection didn't allow the aircraft to stall. I'm not sure what you're getting at, there is nothing wrong with the software, but the people inputting the commands messed up in this one.
 
Couldn't the software be set to ignore the gs above a certain altitude and prevent that?
 
Well, as I read the description, it was closing the gap above glide slope as it was coming down.:dunno:

Well, it's almost closing the gap. There's also the reflection off the something that is creating the false glide path. I'm guessing, but not sure, that it is the runway, since the ILS is on the far side of the pavement you are landing on. That moves the false glide path to the wrong place too- towards you on the approach. If it was a side band off the antenna, it should still lead to the correct place, but steeper as you noted.

I also understand that the glide slope shouldn't be trusted below decision height, but one should be going back up beyond that point anyway if they still need to rely on the ILS to keep oriented for landing.
 
Well, it's almost closing the gap. There's also the reflection off the something that is creating the false glide path. I'm guessing, but not sure, that it is the runway, since the ILS is on the far side of the pavement you are landing on. That moves the false glide path to the wrong place too- towards you on the approach. If it was a side band off the antenna, it should still lead to the correct place, but steeper as you noted.

I also understand that the glide slope shouldn't be trusted below decision height, but one should be going back up beyond that point anyway if they still need to rely on the ILS to keep oriented for landing.

What is decision height on a CATIII approach, and what takes the plane to the runway from there?
 
What is decision height on a CATIII approach, and what takes the plane to the runway from there?

Sorry, I have no idea.

Honestly, I'm VFR only. I stopped IFR training when I moved out here. I couldn't stay IFR current and didn't like what I could encounter in the clouds- icing or embedded CZ. That last one is a lot less of an excuse now with XM and ADSB WX now.

Edit: I believe CAT III approaches have a very low DH, right? There's probably more to it than the simple ILS.
 
Last edited:
Well, CATIII is also what is known as Autoland that takes the plane all the way to the runway. I'm IR, but I fly almost exclusively VFR because I don't like what lives in the clouds around here either.
 
This aircraft was grossly out of position to capture a glide slope normally given the altitudes they were at and the distances indicated from the runway. The aircraft seems to however grabbed a false path that are known to exist in places where properly flown aircraft just aren't supposed to be on an ILS approach. Put it this way, if you have flown thought the OM, crossing radial, DME fix or just where the glide slope pointer centers and for some reason got more than two dots high of fly down indication, you could very possibly encounter a false path and if your AP is capable enough, it could grab ahold of it. All glideslopes regardless of type exhibit false paths in this condition regardless of type. The real wild card is whether the pilot (auto or otherwise) hooks up on the path that's displayed and goes for Mr. Toads Wild Ride. Noteworthy other stuff....glideslopes don't use runways as the reflecting plane, ever. Glideslopes are also good only to about 50 feet which is generally your threshold crossing height. If your taking an aircraft on a CAT III auto land deal, it's all radar altimetry, and multiple autopilots that have to agree completely on the course that you used to be on cause that glideslope is long gone at touchdown. The localized..well...that's another story all together as it plays all the way to the antenna.
 
It's a bit harder in a PA-31 than a Cessna, but still doable. I plug the ILS in regardless just for backup, VFR.

I have little time in PA-31's (Hour or so, VERY little) but I imagine it's quite a bit different especially where I'm n/a and you're worried a lot more about shock cooling those turbos. I rarely have time to plug in the ILS considering I usually come from KCHD. If it was IFR it'd definitely be a different story!
 
Well, it's almost closing the gap. There's also the reflection off the something that is creating the false glide path. I'm guessing, but not sure, that it is the runway, since the ILS is on the far side of the pavement you are landing on. That moves the false glide path to the wrong place too- towards you on the approach. If it was a side band off the antenna, it should still lead to the correct place, but steeper as you noted.

I also understand that the glide slope shouldn't be trusted below decision height, but one should be going back up beyond that point anyway if they still need to rely on the ILS to keep oriented for landing.

How so? The localizer antenna is at the fare end of the runway, but the glideslope antenna is beside the runway near the touchdown point.
 
How so? The localizer antenna is at the fare end of the runway, but the glideslope antenna is beside the runway near the touchdown point.

I stand corrected about the location, if not the cause of possibly reflection off the ground or side bands creating the false glide slope.

Edit: I've been referring to side bands- I mean lobes from the antenna other than the main beam we should follow down the glide slope.
 
Last edited:
I should have taken your avatar more seriously when responding to you. The point remains that the airbus flight control laws are created to prevent such a situation as an "approach to stall.":rolleyes2:

Really? Under Normal Law how does that work? Please explain.
 
The autopilot manual (Altimatic IIIc) for the 'kota makes it clear that glide slope interception must be from below because of the problem of multiple glideslopes. The autopilot will not accept an intercept from above. Too bad Airbus doesn't use similar logic. Too bad the crew training didn't include that glideslope intercept must be from below. I had that training and I was my CFII's first student.



Some autopilots will intercept from above and some won't. I was aware false glideslopes existed and saw it for myself in KJAC.

The Airbus autopilots will intercept from above or below. Some airports you will routinely get intercepts from above. RNWY 24 Manila (RPLL) is one that does it routinely as well as a few others.
 
It seems like the biggest failure here is to brief the approach and then fly it according to the brief. They should have realized something was amiss if they just read the frucking plate.
.

Correct.


Our procedure where I fly is the PM (pilot monitoring) to call out the FAF and altitude crossing. Also during the approach briefing the PF (pilot flying) should brief the altitudes on the approach and cross check them on the FMS.
 
Correct.


Our procedure where I fly is the PM (pilot monitoring) to call out the FAF and altitude crossing. Also during the approach briefing the PF (pilot flying) should brief the altitudes on the approach and cross check them on the FMS.

Likewise where I work (minus the PM/PF stuff, it's all the same guy). And the FMS.
 
Maybe this has already been explained and I missed it but - how do you capture a glideslope from above? Wouldn't it be behind you? :dunno:
 
Back
Top