Strange VNAV descent behaviour

[Turlad 0]

I'm curious to know if the VNAV logic will be updated with SP1. Currently it sometimes demonstrates some odd and illogical behaviour in the descent. Firstly the descent path / profile is often out by 2000 - 3000ft or more in some situations. Today I was flying into OMDB, and the vertical profile insisted I was 4000ft high - there was only one 2000A constraint between the A/C and RW and simple track mile vs height calc, tells me I was exactly on profile. VNAV didn't agree. (Yes, the leg was updated and valid for present position.)

 

In other situations VNAV has been indicating 2000ft high on 10 miles final, when in actual fact I'm at approx 3000ft on glideslope exactly in the groove. Not a disaster on an ILS (although in the real world alarm bells would start ringing of false G/S), but pretty terminal on a Non-Precision if not spotted. I've had a couple of these instances in the sim - alt constaints weren't the problem.

 

Also oddly, modifying the DES speed produces results opposite to what I would expect. Lets say the ECON DES spd is 275 kts. If I modify the DES speed to 320kts, VNAV calculates a new path below the aircraft, leaving me high. I would expect the opposite i.e. I would be low for the new path. At least this is how it works in the 737. I should point out I'm a FO in the 737-800 in real life and do have a vague idea of what I'm talking about. I appreciate even different 737 FMC software versions make VNAV behaviour differ in that one aircraft. The 777 VNAV is different - in fact more basic than the latest and greatest software we have in the NGX, but how exactly?

 

Can anyone explain the differences between the two airplanes and perhaps why we're seeing these results? Or if not, reassure me that SP1 will fix it.

 

Adam Turley

[3-2-1-Now 126]

Always interesting to read comments like this.

 

I found that VNAV in the 777 will have you descend later than the idle des path suggests you should (often being very close to the high drag profile according to OFFPATH DES). I never understood this.

 

If I fly the descent as depicted by the OFFPATH DES circles and using DES NOW to descend earlier than planned, the whole descent goes more smoothly, and I rarely need to use any additional drag at all. If I fly it as depicted, I will end up high and/or fast.

 

Best regards,

Robin.

[Morvious 43]

I encounter many "issues" with approaches where ABOVE altitude restrictions are present.

 

If you descent on path (according to the FMC) to a waypoint with 8000A restriction for example. Many times the path indicator will change after the waypoint resulting in a 2000ft to high indication.

 

What I do now is that I delete the ABOVE restrictions before descend and all seems better.

 

Because I am no real pilot I am not sure if this is normale in the B777 but it feels odd. .

[3-2-1-Now 126]

Let's take 3 waypoints, equally spaced, with constant descent angle.

 

* 20000 ft
* 15000 ft
* 10000 ft

 

If the second waypoint was 10000A and not 15000 then you end up with:

 

* 20000 ft
* 10000A ft
* 10000 ft

 

Now what should happen is that it will *cross* the second at 15000 ft, even though it states 10000A, and you will *still* be on path, because the aircraft is ON TARGET to cross the 3rd AT restriction at 10000 ft, which is correct.

 

That the path deviates, shows that there is a problem with the on-path calculation algorithm, and that is is not looking at the first AT altitude after PPOS as it should be; it's treating the AT OR ABOVE as an AT instead of ignoring it, and then it thinks you're above path after crossing it because it now wants you to be lower than you really are when flying to the next waypoint. It should catch up at the next AT waypoint, but I didn't examine it this closely to see if it does.

 

Best regards,

Robin.

[777simmer 117]

The 777 VNAV is different - in fact more basic than the latest and greatest software we have in the NGX, but how exactly?

 

Can anyone explain the differences between the two airplanes and perhaps why we're seeing these results? Or if not, reassure me that SP1 will fix it.

 

Adam Turley

Hi Adam.

 

The B737 is like 12 years ago for me so I really can't remeber the details.

All in all though, I think it is pretty much the same.

(But anyone, correct me and refreshen my memory if I am wrong.)

Our airline had RNP Rnav approval on the 737 which we do not have on the 777. But other than that....I dont remember much difference in principle.

 

I actually fly less with the PMDG777 than I would like to.

Partly because I am waiting for the FBW fix in SP1, but also because I had to reinstall from scratch in December (Win7 and FSX) and I am still testing the results.

So I have not seen any situations where Vnav did not work right.

 

If you can, give a few examples of STARS that did not work right.

Like I said, I am still testing my system,...might as well test one of those STARS :-)

 

Forgive me for asking, but were the actual winds in FSX close to the forcast winds in the FMC/Descend/Forcast page?

As you know a strong actual tailwind that is not in the forcast page will cause the descend to start too late and the airplane will become too high and leave the Vnav Path (too high).

If I use OPUS FSX real world weather then I get quite a few (abrubt) wind changes (around 20.000ft) when the simulator tries to change/recover from cruise level winds to ground winds. Those are ofcourse not accounted for in the FMC and could throw it off I imagine.

 

As for real world...I can only compare to the B772.

I find the 777 FMC rather sporty.

It happens a lot that I help out Vnav with a little speed brakes.

(Usually I just change to FLCH at around 10.000ft if I feel Vnav will get me there a bit high or fast. If I want to do a Vnav approach later on, then I just re-enage Vnav again once stabilized in for example 3000ft and clean speed.)

What you are describing is the opposite, the FMC telling you are too high when you are not would make it conservative. Which it is not (B772).

 

Changing the descend speed, or getting shortcuts, or deleteing altitude restrictions during descend are not things the 777 FMC "likes".

It is a slow computer that FMC, and takes quite some time before it has corrected for the change.

Those Off Path descend arcs do a much quicker job for some reason and I personally like to use them.

When I get a change as mentioned above, I usually, again, revert to FLCH and use speed and speedbrakes to get back on my (self calculated) vertical path.

 

By the way, the 777 is more aerodynamic (less drag) than the 737.

(or maybe it just has higher descend N1, I have not checked)

Whatever the reason, I found the 777 does not do a 3 degrees glide path with idle thrust (it does 2.5 degrees, which is 16% less steep!) unless you fly 300kt or more in the descend or in strong headwinds.

So (Altitude x 3)/1000 will often get you high on final (even if you add 10nm for decelleration and flap extention).

 

 

With the real world knowledge you have I would say that the problems you are having are caused by the FMC not calculating as it should. So bugs if you want, yes.

Personally I am quite happy with how Vnav works on the PMDG777.

Before the PMDG777 I was using FS2004 and the PSS777.

I can tell you the performance of that plane was rediculous, forget Vnav!

Maybe we are just expecting too much?

Then again, if you say the PMDG737 is spot on.........

 

Anyway, besides me and others trying some of the STARS you flew that did not work, I also suggest you write your experiences to PMDG and see what they say.

[777simmer 117]

Now what should happen is that it will *cross* the second at 15000 ft, even though it states 10000A, and you will *still* be on path, because the aircraft is ON TARGET to cross the 3rd AT restriction at 10000 ft, which is correct.

Robin.

Yes.

 

Easy to recreate and try!

(will do, time permitting)

 

If your theory is correct and if you stay in Vnav, then you will not deviated from your path. You will just end up low (10.000ft over the middle waypoint instead of 15.000ft.)

However If you would descend with FLCH with the Off Path descend circles as a reference, then you will get into the situation where you are following your 2.5 - 3.0 path but the Vnav path is way lower (if your theory is correct that is).

 

One important thing to not forget:

When there are altitude restrictions during descend, then the 777 Vnav path calculation between two altitude restrictions are not based on an idle descend anymore!!

It will do a continuous descend with some thrust to get in a continuous but slow descend from one point to the other. So you could be slowly descending (for example with 500ft/min) where if you use the altitude vs milage rule, you would stay in level flight and then after a few miles start an idle thrust descend.

This calculated continuous descend path remains in the FMC as a virtual line even after you pass the first point with the altitude restriction.

If you are subsequently being radar vectored and cleared by ATC to descend lower than the first altitude restriction, then you will probably change from Vnav to FLCH and by doing so you end up below that virtual path the FMC had calculated.

So even though your distance x 3 rule has you at the correct altitude, your original path is still above you untill you delete the altitude restrictions from the FMC and give it time to recalculate.

 

Vice versa, if for example, you have a STAR with an altitude restriction of "AT 2000ft" on a 12nm final in the FMC.

But ATC radar vectors you at a more normal path of 4000ft at 12nm.

You are now ON your mentally calculated path, but Vnav will still tells you that you are 2000ft too high unless you delete that altitude restriction!

[Turlad 0]

Hi Rob,

 

Honestly I've been finding the 737 VNAV a little 'sporty' as well. I almost always revert to Level Change and Vertical speed below FL100 Most of the time I don't trust VNAV to slow me down correctly for the 10 mile gate, although I've been beginning to give it the benefit of the doubt recently. It works OK as long as the entered forecast winds are correct. I'll tend to leave it in VNAV for a NPA and use liberal speedbrake. 

 

I don't think the problems I've observed in the PMDG 777 are due to wind though - I don't run much weather and my typical tailwind of 10-12 knots is not going to cause the gross errors I've seen. I don't think I've got a specific STAR example, it's general behavior in all descents.

 

There are differences between the 777 and the 737 version I fly. I've noticed for example that if high, the PMDG 777 will revert to idle thrust and VNAV SPD to regain the profile,then go back to VNAV PTH when on again. Whereas if high in the 737, the airplane will just stay in 'Arm' and VNAV PTH mode and dive for the profile and totally ignore the speed, taking you into the clacker if you don't intervene. The 777 is safer, but the PMDG will even do it on VNAV approaches - I'm not sure that 'on-approach logic' is being obeyed here. 

 

I've also noted that when flying through one leg/waypoint to another on approach, say between a typical CI27 to a FI27, the profile will jump maybe 100ft up or down. The NGX doesn't do this (neither needless to say does the real one) - VNAV is very nicely modeled in this and replicates the real one pretty good.

 

Anyway, somethings up and I hope it gets looked at. Otherwise, PMDG products are amazing and I owe a lot to them.

 

Adam Turley 

[3-2-1-Now 126]

Hi Rob,

 

If your theory is correct and if you stay in Vnav, then you will not deviated from your path. You will just end up low (10.000ft over the middle waypoint instead of 15.000ft.)

 

In VNAV, when passing the middle waypont (as per my example) when it is an AT OR ABOVE, the *aircraft* will pass at 15000 ft but the FMC will think you are now too high!. That is the problem.

 

Interesting comments on the rest of the VNAV behavior, and FMC behavior on the real thing. :smile:

 

Best regards,

Robin.

[777simmer 117]

Ok, long post with nothing to report other than that I could not reproduce the reported FMC problems.

Maybe there is some usefull info in this post for some non the less.

 

 

I did three descends into LOWW, R29 this morning.

 

Tip:

when you have altitude restrictions on your STAR, you can easily check if these restriction will cause you to end up on the low or on the high side of an optimal 2.5 - 3.0 degrees path. (and thus if you need speedbrakes afterwards or can take it easy).

Do this by deleting the altitude restriction WITHOUT executing!

The legs page will now show you at what altitude the FMC would like to pass the waypoint if it had no altitude restriction. Then press erase and everything is as before. As long as you dont press execute, you will not cause Vnav to recalculate its path so nothing gets messed up ;-)

 

1)

Ok, The first descend had an At/Above fL60 altitude restriction in the STAR at waypoint Pesat.

By deleting the altitude restriction I checked that normally the FMC would want to pass there at around FL90.

 

I started the descend from FL350 in Vnav Path and everything was perfect.

The aircraft passed Pesat at FL89 and continued a normal on path descend all the way down to 3000ft.

In the end we ended up at GS intercept with clean speed. So too fast and in line with the real thing that is also often sporty like this.

(for a Vnav final approach you better make sure you put something like a flaps 5 or 15 speed into the legs point before the final descend point!)

 

2)

For this descend I put in 3 "AT"altitude restrictions.

One at 10.000ft, one at 15000ft and one at 20.000ft.

I created waypoints (about 40nm, 55nm and 70nm out) so that the restrictions were at positions where the optimal descend profile (2.5 - 3.0 degrees) would also want to pass those altitudes.

Vnav adhered to all AT restriction aNd continued on Vnav Path all the way down.

No too high warnings here either.

The only thing I did notice is that Vnav started to initiate a level off right before those AT restrictions. This resulted in a slight speed loss that the Auto Throttle wanted to compensate for with added thrust. Which is not good ofcourse (waste of energy) because the short addition of thrust was enough to throw me off path (100 - 200ft hight). Not a big deal though, because a little speedbrakes put me right back on path again. The PMDG777 Auto throttle seems to be less smart than the real thing. It seems to hunt for speed rather than anticipate like the real thing.

 

 

3)

I created 2 additional waypoints before NATEX.

NAT02 - FL268 (predicted)

NAT01 - FL205 (predicted)

NATEX - FL124 (predicted)

LOWW

 

Waypoint NATEX had a predicted (unrestricted) crossing altitude of FL124 on the legs page.

Waypoint NAT01 (NATEX/-25) 25nm before NATEX, had a predicted altitude of FL205.

Waypoint NAT02 (NAT01/-20), had a predicted crossing altitude of FL268.

 

I then entered the following altitude restrictions:

NAT01 FL130A. (so, at or above FL130).

 

NAT02 - FL268 (predicted)

NAT01 - FL130A

NATEX - FL124 (predicted)

LOWW

 

Doing so changed nothing at NATEX or NAT02. They were still predicted as FL124 and FL268 respectively. Which is correct and indicated that Vnav will not change its path at all, because it will pass NAT01 at around FL205 anyway which meets the restriction.

 

Then at NATEX I added FL90. So an AT restriction.

Since this restriction is lower than optimal, Vnav now recalculated a new top of descend and now predicts it will pass NAT02 at around FL210.

 

NAT02 - FL210 (predicted)

NAT01 - FL130A

NATEX - FL90

LOWW

 

During the descend NAT02 was passed in FL211. all fine, no strange too high messages.

Later on NAT01 was passed at FL147, which is higher than FL130 (but also ok and according the restriction) and I got no "too high" warning. So all perfect.

The airplane started to slow from 300kt to 240kt at around FL114 and NATEX was passed at FL090 with 240kt. Again no too high or anything.

The descend continued to LOWW nicely following Vnav Path down to 3000ft with a continuous 600ft/min descend and 34% N1(So not an idle descend, but following that precalculated virtual descend path!)

[Alpha Floor 237]

Loving the discussion guys, very informative.

 

I personally don't trust VNAV, I always end up using FLCH or V/S, or flying manually using the green banana. Perhaps this is the remnant in my brain of poor VNAV logics on older aircraft like the PMDG 737 for FS8. In my experience, VNAV  almost always get's you too high, too fast and wastes energy by adding unnecessary thrust while crossing some altitude restrictions and then requesting drag instead of anticipating and planning ahead. It's sort of like VNAV is managing one waypoint at a time, once he is done with that waypoint, he'll see what comes next, instead of integrating it all together on an optimal, no-drag and no-thrust descent.

 

That said, VNAV, which I would rather call EM (for Energy Management, because that's what it is really) is probably what requires the most computational power on the whole FMS system. LNAV is easy and straightforward, but VNAV is another story. I think in real life VNAV doesn't do the job spot on every time, and pilot intervention is often required. Also I can see some real pilots not trusting VNAV and reverting to FLCH or V/S at the slightest deviation ("What the hell is he doing now?")