Landing the FENIX 320

[Jeff Nielsen 1,385]

11 minutes ago, Aamir said:

Welllllll not really. You're basically positing that Asobo's recreation of the world of aerodynamics is perfect. 

Actually, barley even half.  Not only that, the environmental/WX (wind, etc..), is only about half done.  It all ties in together.  They're making good progress, but along with adding more detail, comes a lot more time.  🙂 The price we pay for Jorge and Seb's vision of what they want the sim to be.

Edited October 16, 2022 by Jeff Nielsen

[mac44 15]

5 hours ago, vin747 said:

Having trouble manually landing the fenix 320 (latest fenix build, SU10).. Although FENIX team has acknowledged that the flight model needs tweaking (earlier builds were easier to fly but latest build is more accurate on the numbers, but lost some of the ease of flying), I still wanna get people's opinions here at AVSIM on the latest flight model.. here's my bug report/chat transcript from the discord:

Did my 1st flight on Fenix today. Everything went great except the manual landing.. Flight model was too sluggish and had lots of inertia compared to fbw a32nx and pmdg 737. Will have to tweak control curves and do a pattern to test it again tonight. But how do others feel about the inertia esp during flare? (edited)

Full disclosure - I was using honeycomb yoke during the first flight.. Will try with joystick and tweaked control curves tonight and report back..

Yoke had been usually great for hand flying the a320 neo.. So I assumed it shouldn't matter..

ok, just tried with linear settings/control curves.. response is a tad better, but the current version is unflyable..

it is a bit easier to hand fly now with linear settings, but real issue is during touchdown.. the floating effect is the plane launching into orbit near the ground because the ATHR is trying too hard to maintain VAPP.. normally in any other airbus, i leave ATHR ON until RETARD.. but if i do that here, there is a power surge during those last few moments.. just launches into orbit right around flare unless I cut to IDLE well before RETARD.. or I fly manually like a boeing without ATHR in which case i have better control over power management..

manual landings are impossible now with the current implementation unless you tune your flying skills to the quirks of this build.. in which case, you have to unlearn years of flying TOLISS, FSL, FlightFactor, FBW A320 etc.. and i don't wanna do that..

 

also having big issues hand flying particularly during landing - behaves very differently from the 2 other sim a320's I am familiar with FBW (experimental version) and FFA320 - seems to want to sink irrespective of airspeed and config (making it almost unflyable) for me..............will try with linear curves to see if that helps

[vin747 316]

linear curves wont help.. they will make the hand flying more responsive in other areas of flight, but landing is still a mess.. it is impossible to do a decent manual landing the airbus way (with ATHR ON) with the current build.. lets wait for the next build..

[ha5mvo 238]

11 hours ago, Aamir said:

 You're basically positing that Asobo's recreation of the world of aerodynamics is perfect. 

Well, reading the various quotes and links I was led to believe it's nothing short thereof!

Seriously though, if it flies by the numbers, say accurate pitch/power per weight/altitude and can replicate "unreliable airspeed" procedure to a degree of fidelity then I don't see how it can provide a bad "feel".

Besides, 'feel" the way I perceive it - in its literal sense, has a lot to do with subjective parameters like controller hardware and so on. 

Alternatively, if what you're saying is that the platform doesn't allow for reliable aerodynamic effects of the environment (which IMHO is the case) then this has more to do with MSFS shortcomings. Perhaps you and I interpret "feel" in a different manner.

[vin747 316]

But other planes still fly and land smoothly in this imperfect MSFS atmosphere. So it can be done whether you fly by the numbers or not. 

[ha5mvo 238]

1 minute ago, vin747 said:

But other planes still fly and land smoothly in this imperfect MSFS atmosphere. So it can be done whether you fly by the numbers or not. 

That's what he's saying, if I read it correctly. Namely, if you want a "smooth landing" you'll need the airplane to fly incorrectly....

[Max Kraus 338]

I've gone off the Fenix in recent weeks. The FBW 320 to me is perfect and much better performing, for some reason that I don't care about. The gap between the two is extremely small now.

[Bobsk8 12,570]

One of the things I enjoy most about flying both in real life and flight sims, is a good landing. I have gone up with instructors IRL, and have prided myself in being able to land where you can barely tell when the touchdown point occurs. Two instructors I had said that I landed the plane as well as they did..I have always tried to do the same thing in the flight sim, be it a GA aircraft or a tubeliner. This is something that I am no longer getting from the Fenix. Every landing seems to be a gamble, no matter what I try. If I managed to eek out a semi smooth landing, it takes half the runway before the wheels touchdown. So this is not my inability to know how to land a plane.  I have decided to go back to the GA aircraft, and have been having a great time the last couple of days flying the C 310 and the Kodiak. Going to wait and see what develops with the Fenix. 

[Aamir 2,907]

I don't think you guys are understanding what I'm saying. Instead of words, I guess I'll take you guys down the debug journey for this issue and let you take a look for yourself.

Here are the issues:

1) Apparent wing stalls in flap full config when landing at the MCDU's computed VApp, usually with offside wind, or you're doing BIG bank angles in flap full config. Video: https://drive.google.com/file/d/1Q0-0Yed4eryUNH_PDRKRKRwAT3L-Iv6b/view?usp=sharing

You can see the right wing basically just give up.

2) Complete sluggishness of the aircraft's V/S in relation to pitch change. It's important to note here that the pitch change itself is satisfactory, i.e - the you pull back on the stick and the pitch does change. Your V/S does not. This leads to landings where you flare the aircraft and it feels like nothing is happening - leading to hard landings. Video: https://drive.google.com/file/d/1TfTPzUdUyXNhGX7GIsjGZZxvScTnjsWm/view?usp=sharing

You can see for a given pitch change of nearly 5 degrees, the V/S doesn't contract at all, or maybe by 200-300fpm. This is inherently incorrect. I exacerbate the issue in that video by coming in high and hot, and attempting to change the pitch quite strongly, just to observe what the aircraft would do re it's inertia. Nothing, is the answer.

--

The first thing to do is understand the mechanisms at play when you pull back on the stick. If you pull back, you increase AOA. This increases lift, which increases FPA. As a result, V/S will increase. Variation of AOA is equal to variation of FPA at a given speed, so V/S is geometrically tied to FPA/AOA variation rate and TAS. So when we report that "pitch change" (which is equal to AOA and FPA change in the short time) doesn't change V/S - we are immediately confused. But this says to me, let's go back and check our "by the numbers" claim re lift.

So we checked 2 extreme gross weight variations (both ends of the spectrum, heavy and light) for its AoA in sim, and we were matching the real values in each flap config. So CL vs AOA slope and positioning seems correct. CL vs AOA is linear, so two ends of the spectrum are "good enough" for testing.

We then checked the stall AOA according to the flight manual and we were also correct.

So now the confusing part. Seems to be that lift is technically fine "by the numbers", so what on earth is going on?

You have to then recognise that in Asobo's engine, certain characteristics like drag are cumulative between flap positions, whereas others are not - and the internal engine does what it wants from the value you set with no transparent logic. Now, there is logic behind it, but we are completely blind on how it works. 

In the deployed FM, we have verified correct CL vs AOA slope at each flap position and correct stall AOA as above. Drag has also been painfully tuned to get correct Flap 2 drag, which was really quite wrong in .104 (old version). 

Now, drag is a hidden computation via the lift and camber entries so anything related to "tweaking" the drag is not easy - you can often find when you are happy with CL vs AOA, that the drag is completely off because of the camber input needed to achieve this "perfect" CL vs AOA... It's a bit of a nightmare to get it all in line with official performance figures and tuned correctly - but we're there based on the results from testing. So, that's great. 

Then why is the airplane non-responsive to pitch change? This is a very confusing result if one assumes the physics engine is a "perfect" recreation. Let's go back to a real world video to observe what an A320 does. 

Super helpful video - and it confirmed how reactive the V/S is to a pitch change in a relatively low energy state. Much much snappier. But then, if we’re “by the numbers” - why isn’t ours doing that? Instead, your stick inputs seem to be going into a bowl of porridge... 

Weird question, and here we were stumped for a while. Then I reinstalled .104 with the incorrect flap drag/lift values. Felt responsive! Just like the video. A quick comparison was run and .104 has, basically, slightly higher than “by the numbers” lift on flap full.
So, I requested a small “lift bump” on flap full in our deployed version. Just to assess what it felt like. Less than the wholly incorrect .104, but within a sensible realm - a couple of %, if that. 

Here’s what it looked like: https://drive.google.com/file/d/1kZnsHfk4ZAm6osV1UqR96lq_7lTm5fFi/view?usp=sharing

Delightful. And what you’d expect. V/S is nice and reactive, not overly so. Looks about right compared to the video vs. what you’d expect in “stick feel”. But not, absolutely, utterly, infallibly “on the numbers” - however you can see, visually, the difference in V/S. It’s not a small difference. And the only change is a small bump in flap full lift when slower.

Your end result: 



AOA vs. CAS in flap full.

Blue is the real world. Red is Fenix. The result of this small lift bump is that you’re now 0.3 degrees off in AoA at 175knts. You're probably around 0.1-0.2 degrees off in the phase we're discussing. The ones you don’t see a Red marker for? That’s because it’s absolutely and directly nailed to the Blue engineering/IRL data.

Yes, those are the margins we’re talking about when it comes to “on the numbers” vs. “off the numbers”. 0.3 degrees at max. That’s what we talk about at Fenix because that’s the standard we’re looking for. We’re not hiding it under a margin of error or something, even though we probably could - and would get away with it. 

Admittedly, it does bother me somewhat on a personal level that most assume our product is simply inferior due to our open communications policy about what is and isn’t perfect. We’re happy to admit when something isn’t 100% correct. That doesn’t mean it’s 100% incorrect. You will find these sorts of margins or frankly likely worse in most developers claiming to be “on the numbers”, because I can tell you this - even a Level D full fidelity simulator with a multi million dollar datapack isn’t 100% perfect. Pilots will absolutely tell you it “feels” different. It may, however, hit the numbers. Hell of a lot closer than the rest of us, sure, but still not perfect. And that’s probably because the computational power to recreate absolute, infallible, real world physics for flight in real time doesn’t exist. Much less on your home desktop, on a game engine. So you make little nips, tucks, and adjustments inside of a small margin to get the desired behavior. 

We will continue to push and change things around so that we move toward that "perfect window", I'm sure there's more for us to learn with this new and evolving aero engine. Tbh, we could have probably left it alone from .104 and just “accepted it”, but that's not why any of us spent 2 years reworking and building upon the ProSim base. Certainly would have been the easier and more fiscally “sound” route - as time is money, and we’ve committed lots of time to chasing 0.3 degrees of AOA. But the promise we make to you all  is that we’re here to do the best we can. Sometimes things that should be a forward step, aren’t. I think that’s OK. I think that’s the price of moving forward and trying to push boundaries with respect to “degrees of accuracy”. 

For anyone that has questions about our approach, we try to answer any questions with transparency and fullness over on our Discord (I know most here don't like it, but we now have a forum area for verified customers that's much more conventional!).

Edited October 17, 2022 by Aamir

[Jeeeno 314]

This is the most interesting thread in Avsim forums. 

[Paladin2005 422]

Thank you very much for that insights, Aamir.

Avsim is in many cases a toxic place. Too many old f*rts having no clue of software development, including its balances to be taken.

Take inputs for failure correction, but dont let Avsim disturb your optimism and strive for perfection.

THanks for that great Airliner!

Edited October 17, 2022 by Paladin2005

[F737MAX 5,544]

2 hours ago, Aamir said:

Certainly would have been the easier and more fiscally “sound” route - as time is money, and we’ve committed lots of time to chasing 0.3 degrees of AOA. But the promise we make to you all  is that we’re here to do the best we can.

There needs to be a balance between these two competing objectives.
Getting the A320 to flare and land similar to how it was in v1.0.2.104, reducing the fuel burn-off rate differences to aid better flight planning, release of sharklets and the IAE variant engines are probably more interesting to the majority of people.

As pointed out, Level D sims aren't precisely 100% faithful representations of the real aicraft they are simulating. While all Fenix owners would love a 1:1 recreation, sometimes getting it to feel close enough is good enough, IMO.

[vin747 316]

In my engineering company, there's a saying: "All models are wrong. Some are useful."

Any simulation only tries to mimic reality.. so, taking a balanced approach to get decent accuracy (+/- 0.3 deg of AoA in this case) and then hitting the other major targets that impact 99% of the simmers (IAE, Sharklets, smooth landings) will be the best way forward..

[lwt1971 7,347]

4 hours ago, Aamir said:

:
:
You have to then recognise that in Asobo's engine, certain characteristics like drag are cumulative between flap positions, whereas others are not - and the internal engine does what it wants from the value you set with no transparent logic. Now, there is logic behind it, but we are completely blind on how it works. 
:
:
Your end result: 



:

Blue is the real world. Red is Fenix. The result of this small lift bump is that you’re now 0.3 degrees off in AoA at 175knts. You're probably around 0.1-0.2 degrees off in the phase we're discussing. The ones you don’t see a Red marker for? That’s because it’s absolutely and directly nailed to the Blue engineering/IRL data.

Yes, those are the margins we’re talking about when it comes to “on the numbers” vs. “off the numbers”. 0.3 degrees at max. That’s what we talk about at Fenix because that’s the standard we’re looking for. We’re not hiding it under a margin of error or something, even though we probably could - and would get away with it. 

Admittedly, it does bother me somewhat on a personal level that most assume our product is simply inferior due to our open communications policy about what is and isn’t perfect. We’re happy to admit when something isn’t 100% correct. That doesn’t mean it’s 100% incorrect. You will find these sorts of margins or frankly likely worse in most developers claiming to be “on the numbers”, because I can tell you this - even a Level D full fidelity simulator with a multi million dollar datapack isn’t 100% perfect. Pilots will absolutely tell you it “feels” different. It may, however, hit the numbers. Hell of a lot closer than the rest of us, sure, but still not perfect. And that’s probably because the computational power to recreate absolute, infallible, real world physics for flight in real time doesn’t exist. Much less on your home desktop, on a game engine. So you make little nips, tucks, and adjustments inside of a small margin to get the desired behavior.

Fascinating read and thanks so much @Aamir for this detailed behind-the-scenes look at the development and debugging process! Absolutely true that no sim including Level-D ones can achieve physics/aerodynamics perfection yet on current hardware. But it's mighty impressive how the latest fixes in the works there for the Fenix come this close with such small variances or are bang-on. Assume you might be already communicating to Asobo to put out more behind-the-scenes info and transparency of their own, such as in parts of their aerodynamics engine you mention above.. the more they do the better for devs like you and users like us.
 

Edited October 17, 2022 by lwt1971

[mobiel 67]

Beautiful support, thanks Aamir for this great explanation