Sign in to follow this  
3-2-1-Now

Turn Prediction

Recommended Posts

Hi,

 

Turn prediction needs tweaking. At the moment it overshoots turns, and if it has a tailwind, the error increases.

 

Best regards,

Robin.

Share this post


Link to post
Share on other sites
Help AVSIM continue to serve you!
Please donate today!

 

 


Turn prediction needs tweaking. At the moment it overshoots turns, and if it has a tailwind, the error increases.

 

If you believe it's truly an error, you should know by now that the official avenue to report it is via the support portal.  Chances of things being seen here are slim, especially with the team constantly working on finishing the SP up...

 

I will say, though, that no aircraft is perfect, and I'm not quite sure this is actually a bug.

Share this post


Link to post
Share on other sites

Good point. I'll raise a support ticket so they're aware.

 

Regarding the turn prediction - flying S turns is not what the real aircraft would be allowed to do.

 

Example: turn is a 60 degree right turn. Aircraft will turn right, overshoot, so it has to sustain the turn back towards the track, then it will have a small overshoot during the left turn back on course, then it will turn right to correct and then be on course.

 

I also had such a 60 degree right turn on the LAM3A to RNAV RW27R at Heathrow. Even at 137 kts it overshot the turn and had to turn to correct. This is in no way how the real thing would fly.

 

Try a flight from Heathrow to Schipol. DVR5F REFA1A. Make sure the aircraft has a tailwind of 50 kts. For me it overshoots such turns, and spends a rather long time turning to correct itself, when all it had to do was start the turn a mile earlier.

 

Best regards,

Robin.

Share this post


Link to post
Share on other sites

 

 


Example: turn is a 60 degree right turn. Aircraft will turn right, overshoot, so it has to sustain the turn back towards the track, then it will have a small overshoot during the left turn back on course, then it will turn right to correct and then be on course.

 

It's definitely worth a look, but there are tolerances to everything.  There's a reason controllers are allowed a limit of 30 degrees (20 in certain cases) for a final vector for the LOC.  Part of it is so that the aircraft/pilot can establish themselves in a stable state within a reasonable time before landing.

 

With angles like 60 degrees, I'm not at all surprised it would cross the course.  You may be right (so definitely get that ticket in), but I wouldn't be surprised if this is accurate to the aircraft's behavior.

Share this post


Link to post
Share on other sites

Good point. I'll raise a support ticket so they're aware.

 

Regarding the turn prediction - flying S turns is not what the real aircraft would be allowed to do.

 

Example: turn is a 60 degree right turn. Aircraft will turn right, overshoot, so it has to sustain the turn back towards the track, then it will have a small overshoot during the left turn back on course, then it will turn right to correct and then be on course.

 

I also had such a 60 degree right turn on the LAM3A to RNAV RW27R at Heathrow. Even at 137 kts it overshot the turn and had to turn to correct. This is in no way how the real thing would fly.

 

Try a flight from Heathrow to Schipol. DVR5F REFA1A. Make sure the aircraft has a tailwind of 50 kts. For me it overshoots such turns, and spends a rather long time turning to correct itself, when all it had to do was start the turn a mile earlier.

 

Best regards,

Robin.

I noticed that for a long time, and the overshoot will get worse if the next waypoint is very close ( under 20 nm). Perhaps a 777 pilot can comment on whether this is the case in the real world.

 

Francis Leung

Share this post


Link to post
Share on other sites

Hi,

 

I raised a ticket, and the response back was that they have numerous videos of the Level D simulator showing over- and under-shoots in similar situations.

 

Best regards,

Robin.

Share this post


Link to post
Share on other sites

In the interests of "as real as it gets", does the 777 overshoots in the real world then? If not, then I would say it is desirable if PMDG can improve that in the next updates. I won't call that a bug though, not that serious.

 

Francis Leung

Share this post


Link to post
Share on other sites

I must say that if it over-shoots the way the (FS) sim does, then I'm definitely raising an eyebrow at it. I thought it would be (much) cleaner in turns than it is.

 

A340-300 is a similar vintage, and my friend recently jumpseated on it (actual aircraft) and watched it flawlessly capture the ILS, with a 90 degree intercept at 250 kts, with no overshoot, and roll out on course. Good luck doing that in the sim...

 

Best regards,

Robin.

Share this post


Link to post
Share on other sites

I might be wrong, but 60 or 90 degrees turns to intercept the localizer is quite extreme. Intercept angles of 20 to 40° are more realisitc and as Kyle said, most ATC will give your angles of 30°. When a 60 to 90° turn is needed to intercept, I usually do it via the MCP heading selector and when established, might go back to LNAV or APP or manual if visibility is good. I'm not sure if a RW pilot trusts LNAV for a 60° intercept angle during approach.

Share this post


Link to post
Share on other sites

The PMDG 777 can do a 90 degree intercept.

 

If that 90 degree intercept is part of an RNAV STAR, such as LIZZI7A Arrival to Runway 16 at YMML Melbourne.

 

http://www.airservicesaustralia.com/aip/current/dap/MMLSR05-135.pdf

http://www.airservicesaustralia.com/aip/current/dap/MMLII01-137.pdf

 

For this, the LNAV/VNAV would be engaged and the APP mode armed after the 352°radial is passed.

 

I don't see a 90° intercept working off "HDG SEL" mode with APP mode armed.

Share this post


Link to post
Share on other sites

 

 


I don't see a 90° intercept working off "HDG SEL" mode with APP mode armed

 

No, I manually intercept using HDG SEL if a 90° turn is required. When on the LOC, I arm APP. Not saying this is the correct procedure, but for steep intercept turns, it puts me on the glide nicer and smoother than LNAV/VNAV. Just my experience.

When visibility is fine, I might just turn the AP off before the turn and continue hand flying.

Share this post


Link to post
Share on other sites

I might be wrong, but 60 or 90 degrees turns to intercept the localizer is quite extreme.

 

Let me quote what I wrote in post #8:

 

 

 

my friend recently jumpseated on it (actual aircraft) and watched it flawlessly capture the ILS, with a 90 degree intercept at 250 kts, with no overshoot, and roll out on course.

 

You may *think* it is extreme, but it isn't. Aircraft are more than capable of flying these intercepts. Concorde could make a 179 degree intercept! It is written in the flight manuals.

 

People saying these things are "sim limitations" really need to get a clue about what actual systems can do and how they work. Where it would be a real sim limitation is when using the FSX localizer signal to determine when to turn, if it is incorrectly modelled.

 

In short: the greater the turn angle, the earlier the turn needs to be made. If you'd turn at Standrd Rate 0.5 nm before intercept for a 20 degree change of heading, then you'd turn 2.5 nm before for a 90 degree intercept. It is all down to turn radius, and is easy to calculate for a given ground speed.

 

http://en.wikipedia.org/wiki/Standard_rate_turn

 

43aaffce2652dd4feabb052a16c7a55f.png

 

Best regards,

Robin.

Share this post


Link to post
Share on other sites

 

 


Concorde could make a 179 degree intercept! It is written in the flight manuals.

 

"Can" and "flawlessly" are two very different terms.

 

Knowing the type of navigation Concorde was using, I can nearly guarantee a few S-Turns as it tried to capture whatever signal it was trying to follow.  Of course, closer to the source of this signal, it would be tougher.

 

 

 


People saying these things are "sim limitations" really need to get a clue about what actual systems can do and how they work. Where it would be a real sim limitation is when using the FSX localizer signal to determine when to turn, if it is incorrectly modelled.

 

You're assuming "correctly modeled" is a perfect system, though.  The GS antenna of an ILS creates false glideslopes.  This, of course, is not "perfect."  FSX's depiction of ILS in the sim is not perfect because it's perfect in the sim. GS is available and perfect no matter where you are on approach.  False glideslopes don't exist.

 

So, modeling something to be perfect isn't necessarily modeling it to be correct in all cases.  Sure, I bet someone could easily write code for turn prediction to be more accurate, but is that going to be accurate to the real thing?  I doubt it.  There's a reason we still have RNP 0.3 as a max tolerance for "very precise" approaches.  Technology isn't always perfect, and there's still error in it, even when it's "more perfect than not."

Share this post


Link to post
Share on other sites
There's a reason we still have RNP 0.3 as a max tolerance for "very precise" approaches

 

As you should know, that is related to the tolerance of position of where the aircraft is vs. where it thinks it is, and nothing to do with how tight it can turn or anything else. The aircraft can have an ANP of 0.29 and still fly an RNP 0.3 approach, though it wll put it right at the edge of the protected corridor.

 

http://www.smartcockpit.com/download.php?path=docs/&file=Getting_To_Grips_With_RNP-AR.pdf

 

Best regards,

Robin.

Share this post


Link to post
Share on other sites

 

 


As you should know, that is related to the tolerance of position of where the aircraft is vs. where it thinks it is, and nothing to do with how tight it can turn or anything else. The aircraft can have an ANP of 0.29 and still fly an RNP 0.3 approach, though it wll put it right at the edge of the protected corridor.

 

Yes, I should (and do) know.  I helped design them...

 

You're correct in that it's a tolerance of calculated position versus actual position.  My point wasn't that it is the same technology.

 

RNP limits are set because we know technology is imperfect.  That's why we have ANP to begin with - it evaluates how much imperfection is in our assumption of position.  Hit POS on the 777 or the NGX and you'll see how out of sync the different sources of nav data are (especially as flight gets longer).  If technology was perfect, we wouldn't need ANP to measure how imperfect it truly is.

 

Using a different example: TERPS requirements (FAA, of course - though I'm sure ICAO members use similar standards) are really there because of the error present in the systems.  There is error in an ILS (both in LOC and GS systems) signal.  There is also potential error in how the aircraft tracks that signal, even if that signal were perfect.  Both of those limitations are accommodated in the tolerances set forth in the TERPS requirements.  Again, if all systems were perfect, those wouldn't be neccessary.

 

Aircraft can't even be 100% sure of their actual location.

Systems that guide aircraft cannot provide a 100% accurate signal.

 

So how is it that, given all of this error in those other technologies, aircraft tracking would somehow be 100% "on rails?"

Share this post


Link to post
Share on other sites

It'd be funny if PMDG managed to simulate False Localizer/Glideslope captures. Can you imagine all the forum complaints! lol!

Share this post


Link to post
Share on other sites

In Lnav with AP engaged, the real 777 does not overshoot!

 

Never?

 

I dont know, but in 12+ years Of normal operation I cant remember having ever seen that.

 

Of course you can create situations where the airplane just cant do what you want.

And maybe that departure out of Heathrow is one of them situations (have not looked at it).

For example, create two waypoints is sequence, 1nm apart and at a 90 degrees angle!

No way it can do that at 300kt and you WILL see an overshoot!

That's why I said "normal ops"!

 

The system knows your ground speed and anticipates for strong winds!

 

In a hold for example, if the tail winds are too strong (too high ground speed), it will even tell you "unable HOLD airspace".

Because it knows exactly what the turn radius is going to be!

 

Now capturing a localizer is a different story.

Lnav can not anticipate the reception of the localizer!

Once it starts receiving it the INS/GPS system will still calculate when to turn so that you dont overshoot the loc (based on loc closure rate I think) . But if you fly to the loc with 300kt at 90degrees, than it should have turned before it even received the inbound course. Which it does not, cant, and thus will cause an overshoot.

 

 

A level D simulator, by the way, is still only a simulator!

I have seen many level D screw ups, the real 777 does not do.

 

 

And as for the PMDG777, I really does not do too bad I think.

I have not noticed those overshoots.

I am sure the logic can be improved, but for an FSX simulator....not too bad at all?

  • Upvote 3

Share this post


Link to post
Share on other sites

EDIT:I meant "But for an FSX simulator.....not to bad at all ! " ofcourse (?=!)

 

 

The most annoying part of AVSIM....that you can not edit your posts!!!

Share this post


Link to post
Share on other sites

In Lnav with AP engaged, the real 777 does not overshoot!

 

Never?

 

I dont know, but in 12+ years Of normal operation I cant remember having ever seen that.

 

 

That's nice to know what is the T7 doing in the real world thanks.

 

 

And as for the PMDG777, I really does not do too bad I think.

I have not noticed those overshoots.

 

 

Well I can show you how the PMDG T7 flies the famous 78 degree turn north of Beijing China (ZBAA) via the following screenshots. This is a common real world route from Europe to Shanghai/Korea/Japan via north China, airways W80 connecting to B339. You can reproduce it yourself easily:

 

 

1. approaching VOR HUR: tracking correctly with a 25 kt tailwind which is FSX default and commonly encountered in the real world:

 

https://db.tt/LrngoWYY

 

2. T7 will start the turn at about 5.3 nm to HUR, it will overshoot:

 

 

https://db.tt/kYInXVFB

 

3. at about half the turn, already overshot 0.4nm:

 

 

https://db.tt/gpMF0hnG

 

4. maximum overshoot 1.8 nm to the left of track, note the distance to the next waypoint LADIX:

 

 

https://db.tt/N4kUyZZ6

 

5. finally it took 29 nm from HUR to fully correct the track:

 

 

https://db.tt/pVPdY4bI

 

Again, don't get me wrong I'm critisizing PMDG. I'm just stating the facts. The T7 is great, if it can track like the real one, magnificant: a wish list.

 

 

Francis Leung

Share this post


Link to post
Share on other sites

Yeah...I must say that is a pretty extreme turn with 500kt ground speed.

I have not been in that area, but I am going to stick to my earlier statement that it should not overshoot.

 

It knows your GS, it knows that it uses 25degrees of bank. Those two factors results in a calculatable turn radius!

 

The only thing that would cause overshoot would be a "fly over" waypoint versus "fly by" waypoint.

But your pics clearly show HUR as fly by.....thus it should not overshoot.

 

There is an oh....dont know....., something like 45 degrees turn, after Afghanistan into Pakistan (Pavlo to Jhang).....that causes absolutely no overshoot even at 600+ ground speed!

 

 

EDIT: Say.....stupid question.....you guys are not flying in accelerated mode right?

  • Upvote 1

Share this post


Link to post
Share on other sites

In Lnav with AP engaged, the real 777 does not overshoot!

 

Never?

 

I dont know, but in 12+ years Of normal operation I cant remember having ever seen that.

 

Of course you can create situations where the airplane just cant do what you want.

And maybe that departure out of Heathrow is one of them situations (have not looked at it).

For example, create two waypoints is sequence, 1nm apart and at a 90 degrees angle!

No way it can do that at 300kt and you WILL see an overshoot!

That's why I said "normal ops"!

 

The system knows your ground speed and anticipates for strong winds!

 

In a hold for example, if the tail winds are too strong (too high ground speed), it will even tell you "unable HOLD airspace".

Because it knows exactly what the turn radius is going to be!

 

Now capturing a localizer is a different story.

Lnav can not anticipate the reception of the localizer!

Once it starts receiving it the INS/GPS system will still calculate when to turn so that you dont overshoot the loc (based on loc closure rate I think) . But if you fly to the loc with 300kt at 90degrees, than it should have turned before it even received the inbound course. Which it does not, cant, and thus will cause an overshoot.

 

 

A level D simulator, by the way, is still only a simulator!

I have seen many level D screw ups, the real 777 does not do.

 

 

And as for the PMDG777, I really does not do too bad I think.

I have not noticed those overshoots.

I am sure the logic can be improved, but for an FSX simulator....not too bad at all?

 

Thanks for this post. When I was told in the support ticket that the aircraft could not compensate for winds, I was extremely skeptical.

 

The 777 FMS uses the Motorola 68010 processor, and it is a processor I know well. It is not beyond its capabilities to factor ground speed into a turn prediction algorithm, thus alowing for ground speed in turn prediction (and in fact, as it is following a ground track, it would HAVE to consider ground speed). The processor is slow, but not THAT slow. Heck - it is discussed as if a turn prediction algorithm is complex, which it is not (see my post above for the calculation of turn radius which would form the basis of turn prediction).

 

I agree that LNAV could not *detect* LOC intercept, but it does pre-empt it by starting the turn even before LOC is actually captured by the system, based upon FMS position. It helps "feed" the aircraft onto the localizer.

 

I also know that LNAV and ILS are totally seperate systems, but the ILS example (and it was ILS intercept I was talking about) was merely an example of what a real aircraft is capable of doing, which is something many simmers dismiss as "impossible" or "not accurate".

 

Best regards,

Robin.

Share this post


Link to post
Share on other sites

If this is true, and the overshoot in the sim is exaggerated, perhaps Rob should raise a ticket quoting his real world experience? It may get the issue noticed and addressed,

 

Regards.

Share this post


Link to post
Share on other sites

We all know  that Pmdg  do a good job in what they do in simulating,  but surely we cant expect what the real world t7  to do  exactly what it  does on our Sims?  If it  does 95 percent or even 98   do we have to worry about  the other 2  percent  that it  cant do exactly. Guess we are expecting  to much  on what our Sims can  do comparing  it  to the real t7

Share this post


Link to post
Share on other sites

Tks all for the reply and support. Yes of coz the sim was not running at 2x or more, you don't see that announced on the upper right corner of the screen shot right  :biggrin:

 

BTW, the next screenshot is the ASN real world weather on 7/3 last month above ZBAA. The tailwind is basically 82 knots, making my original upload of 25kt quite mild! 

 

 

Francis Leung


Screenshot missing on last post:

 

 

https://db.tt/2Ksuz2FF

 

 

Francis Leung

Share this post


Link to post
Share on other sites

We all know  that Pmdg  do a good job in what they do in simulating,  but surely we cant expect what the real world t7  to do  exactly what it  does on our Sims?  If it  does 95 percent or even 98   do we have to worry about  the other 2  percent  that it  cant do exactly. Guess we are expecting  to much  on what our Sims can  do comparing  it  to the real t7

I agree with that. For me when I see the PMDG plane do something silly, I just correct the mistake and forget about it. I am not saying this is the way to go....it is just that it all does not bother me too much. Things that do bother me and I would like to see improved (FBW system or irratic GoAround thrust behavior) I do (did) report.

If this is true, and the overshoot in the sim is exaggerated, perhaps Rob should raise a ticket quoting his real world experience? It may get the issue noticed and addressed,

 .

Ofcourse I also agree that if things CAN be improved....then you would normally expect from PMDG that they will do so.

 

I will have a look at Lnav turn predication when I have some time....

I have no problem reporting my findings to PMDG.

 

As it is, I have some problems with my system due to Windows Defender that was running during GEX/UTX installation (yes stupid....I thought it was of but it was not).

So I cant test untill maybe in two days or so.

  • Upvote 1

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this