Turn Prediction

[Alpha Floor 237]

Rob, upload your pictures to imgur.com and post them here copying the URL of the image.

[scandinavian13 5,192]

 

 

I will send my pics to PMDG support.

 

Thanks.

[FlyGuy42 4]

I think PMDG got the turn prediction right with the NGX, as the behviour you describe seems very similar to that aircraft. Hopefully they can change this in the service pack or as a hotfix.

 

Regards.

[francisleung 5]

I finally had a real world leg that allowed me to take a closer look at turn prediction. FL370 and GS537kt

About an 80 degrees course change (left turn) in Chineese airspace.

The ND displayed a beautifull pre calculated magenta radius for the 80 degrees left turn.

It started to turn about 5.7nm before reaching the waypoint.

With 25 degrees of bank we ended up 0.28 right of track after the waypoint.

Yep, that´s right....slightly off track.

 

Conclusion:

PMDG is correct in saying that it is normal for the aircraft to sometimes get slightly oftrack in extrem turns.

 

However, it is not normal for the the magenta track not being pre calculated as a smooth radius.

And it is also not normal for the airplane to get off track by 2nm or more.

 

 

Hi Rob,

 

What wind was there when you do the flight? 0.28 nm right of track is very good overall compared with my case. Thanks that we both have been reprimanded by the Chinese ATC 

 

I believed it is PMDG settings at Avsim that embedded images are not allowed in all user created topics by default. Although it is not fair that their staff can post in-line images  :(

 

BTW, on this topic, I believed the aircraft cannot track very well on departure if there is an immediate turn after takeoff. In the following screenshot, I took off from EGLL 09R, BPK SID, fully loaded but with not much fuel for a short hop across to EHAM, default scenery:

 

1. ready to roll, showing the immediate left turn on ND : https://db.tt/ybh8BbK9

 

2. overshot D197J, unless that is not a mandatory waypoint : https://db.tt/5iSVmAFF

 

3. At EHAM, I did the VOR 27 approach, again default scenery, clear sky with wind 271 @ 8 knots

 

This is the plan with a tear-drop turn : https://db.tt/ATIGdxrw

 

4. VNAV was tracking normal until the left turn : https://db.tt/2RixRNia

 

5. It suddenly showed that I was 682 ft above profile past PAM-4. Rob, can I assume that this is not the case in real life? https://db.tt/JXW57ig3

 

Many thanks for some clarifications.

 

Francis Leung

[FlyGuy42 4]

Rob, do we have a response from PMDG to your ticket yet?

 

Regards.

[Tabs 2,790]

I need to make something very clear here (which someone touched on briefly earlier in the thread) - this is not the real airplane. It's a $90 simulation of a $300 million dollar aircraft operating on an 8 year old simulation platform that dates back in its core engine to nearly 20 years ago. You guys have no idea the level of complexity going on here and how much "FS crap" we're smoothing over and trying to deal with in these autopilot algorithms. It is honestly something of a miracle that it works as well as it does.

 

I absolutely do have video from a real full flight 777 simulator (I was there flying it and filming) that shows turn under and overshoots occurring. Our tech team pilots told us they have seen the real thing do it before as well. A ton of people (including some real pilots) don't realize this but the magenta line is not supposed to be a depiction of the exact flightpath the airplane will follow. It is a prediction the FMC makes. The FMC is essentially running a simulation of the flight and drawing what it thinks the actual autopilot LNAV controller is going to do to follow the path that's being demanded of it. This prediction is not perfect - one major example of this is when airspeed changes during a turn. The FMC draws a constant radius arc based on a standard rate turn radius at the FMC predicted airspeed for the start of the turn. If the airplane's speed changes during the turn, the standard rate radius for the turn changes but the depiction on the ND doesn't. The result is that the actual position of the airplane goes inside or outside of the ND magenta line. The actual AFDS LNAV controller has nothing to do with the magenta line - they are separate processes. A ton of addon developers just link them together and their addons fly the line perfectly as if they're on rails. We could very easily do this (it's a much less complicated mathematical exercise), but that is not what the real airplane does.

Knowing how it actually works - how exactly could the prediction that the FMC makes for the magenta line "know" precisely what the headwind or tailwind is going to be when it reaches X waypoint? We already account for this stuff with the RTE DATA wind entries, but again those are never perfectly accurate. If you get to the waypoint and the actual wind is different then the turn the LNAV controller actually makes is not going to match the prediction.

 

Now - all of that said, we are going to review this part of the code here at the end of the SP1 process and see if there's anything we might be missing that is causing it to get more off than it should during some of these tailwind/headwind situations at cruise. Absolutely no guarantees here - we are not going to spend months rewriting our LNAV algorithms that work in 99% of cases to fix the 1% that get slightly off. The risk of breaking other things and causing even more problems is extremely high - all of these systems interact with each other dynamically and it is just a huge can of worms to throw the balance off in one and hope it doesn't affect others. I am sorry we can't 100% recreate the real airplane to every last detail like some of you seem to expect, but that is never what we've claimed to be doing here. This is a video game. If you want that level of realism, you're going to be paying upwards of $500 an hour at a certified training facility.

[777simmer 117]

Have you seen the pics and comments I sent to support Ryan?

 

 

 

 

 

 

Rob, do we have a response from PMDG to your ticket yet?

 

Regards.

No not yet....I did not have the possibility to upload my pics to PMDG support untill about 9 hrs ago (about 13:00Z on Wednesday the 7th).

[Tabs 2,790]

I'm still going through today's tickets - there were a lot. I'll reply to you when I get to it.

[francisleung 5]

Thanks Ryan, no need to be annoyed and defensive. Again starting from my first post, I'm not complaining but stating the observations I have on the T7. Hope Sp1 can be released soon. 

 

Best wishes,

 

Francis Leung

[777simmer 117]

Hi Rob,

 

What wind was there when you do the flight? 0.28 nm right of track is very good overall compared with my case. Thanks that we both have been reprimanded by the Chinese ATC

 

I believed it is PMDG settings at Avsim that embedded images are not allowed in all user created topics by default. Although it is not fair that their staff can post in-line images :(

 

BTW, on this topic, I believed the aircraft cannot track very well on departure if there is an immediate turn after takeoff. In the following screenshot, I took off from EGLL 09R, BPK SID, fully loaded but with not much fuel for a short hop across to EHAM, default scenery:

 

1. ready to roll, showing the immediate left turn on ND : https://db.tt/ybh8BbK9

 

2. overshot D197J, unless that is not a mandatory waypoint : https://db.tt/5iSVmAFF

 

3. At EHAM, I did the VOR 27 approach, again default scenery, clear sky with wind 271 @ 8 knots

 

This is the plan with a tear-drop turn : https://db.tt/ATIGdxrw

 

4. VNAV was tracking normal until the left turn : https://db.tt/2RixRNia

 

5. It suddenly showed that I was 682 ft above profile past PAM-4. Rob, can I assume that this is not the case in real life? https://db.tt/JXW57ig3

 

Many thanks for some clarifications.

 

Francis Leung

- we had a 60 degrees crosswind from behind, with 90kt.

In the left turn this changed into a pretty much traight up 90kt tailwind.

But it does not realy matter, because tailwind or not, the FMC knows you ground speed and pre-calculates the radius with that.

And yet, as I said, even the real thing ends up off track a little bit sometimes.

 

- The first turn during departure is usually hand flown. And many assumptions the FMC makes for the turn might not be as actually flown. (maybe you must use a higher minimum bank angle, maybe you used speed intervention with a slower speed, etc etc).

 

- That departure you flew, I looked at it, is just impossible to fly pretty.

If you put LON in the fix page with radial 072/10

And BPK with radial 197

Then you see how close D072J is to the 197 radial from BPK that you have to intercept.

To get on that 197 without overshoot you would have to turn sooner, but apperantely that is not what the designers of the SID want.

It does not say that overshoot is not allowed by the way. It sais turn AT and with that, the overshoot follows automatically.

However, the fact that you have cleaned up and are doing 250kt does not help ofcourse.

Better stay at flaps 5 speed untill inbound to BPK!

 

- Vnav in EHAM: I would have to try for myself real life to give you a definate answere. We hardly ever fly full procedures like that....allways radar vectored.

So I dont know for a fact what the real 777 would do, but I would assume that it is not 682ft hight after PAM-4. If it were then I would give it some time and see if the Vnav pointer comes back up (if I calculate roughly, then you should be ok continuing like that). If it does not, then you can use a little speedbrakes untill Vnav Path engaged again automatically.

 

But that would be one of those little snags I guess that Ryan meant that are realy difficult to iron out.

[3-2-1-Now 126]

@Ryan: first of all, thank you reviewing this at the end of SP1 development.

 

The problem with the overshoot is not that it won't follow the magenta line, but the fact the cross-track error at the end of the turn can be in excess of 2.0 nm!!! The basic problem is the aircraft is not starting the turn early enough. The higher the ground speed, the further out it should turn (at 100 kts it would turn at about 1.0 nm, and 500 kts, about 5 nm).

 

Best regards,

Robin.

[francisleung 5]

However, the fact that you have cleaned up and are doing 250kt does not help ofcourse.

Better stay at flaps 5 speed untill inbound to BPK!

Wow, thanks Rob for the details. I have only been a PPL in real life so flying big irons IFR is really new to me. Yes, in my original flight at EGLL. I was fully loaded and got an "unable next alt" error, despite not using derated thrust. The second time, I used flaps 15 for take off and the unable next alt error disappeared. I'll try and stay on flaps 5 as you suggested and put speed restrictions on the FMC for those initial points next time.

 

 

- Vnav in EHAM: I would have to try for myself real life to give you a definate answere. We hardly ever fly full procedures like that....allways radar vectored.

So I dont know for a fact what the real 777 would do, but I would assume that it is not 682ft hight after PAM-4. If it were then I would give it some time and see if the Vnav pointer comes back up (if I calculate roughly, then you should be ok continuing like that). If it does not, then you can use a little speedbrakes untill Vnav Path engaged again automatically.

 

 

 

For argument sake I tried that VOR 27 approach to see how well the T7 track while the ILS was out of service  What I have observed is I was on VNAV path until the next waypoint switch to PAM-4 then simultaneously the VNAV indicator said I'm 682 ft above. I assume that won't happen in real life. Again, may I stress that I don't treat this as a PMDG bug. It may have been at their programming limits. Anyway, what happened next was at the > 180 degree teardrop turn, the aircraft do a 1500 ft. descend to 'catch up' back on VNAV path.

 

BTW, I noticed that on idle descent from TOD, if I extend the speedbrake, the rate can increase up to 5500 ft/min. Is that what the aircraft do in real life? 

 

Many thanks and have a great weekend. (Friday morning Down Under here)

 

Francis

P.S. I mean 1500 ft/min descend.

[kevin911010 0]

I think I should share my experience on the pmdg 777 here.

 

I had a missed approach and went into a holding pattern. After some minutes I selected a new approach to the same runway again, the route in its simplicity is just a right-hand traffic pattern started on the downwind leg. I can't recall the numbers exactly but on the base to final turn it did overshoot somewhere between 0.5 to 1 mile, staying in LNAV all the time. Not a problem in my case because I had a 8 mile final to correct it, and if I don't like the turn I can always override the A/P and hand fly to the FAF.

 

So in short, out of the many features provided by pmdg, this is a really minor issue. It's a little bit annoying but does not affect how much I like the pmdg 777 at all.  WE are flying the plane, not the A/P. So can't we just treat that as a minor imperfection of a second officer under training in the right hand seat?  I mean, come on guys, real world stuff are imperfect as well. So this is the imperfection we get when we spent $89 on a B777 simulation that we can run in our home machines. I'm not complaining. It's just what you get (-: