NG FMC: Examples of RF legs

[cavaricooper]

It seems the AP is not calculating a proper constant angle turn to maintain the flightpatch ...

Developing a perfect control system, including wind influence and performance variations, is ... almost "rocket science". The Boeing engineers are better than me.

|-|ardy

... not for long, a few NG versions down the road, all the Boeing 747-8 programmers will be here, taking notes

C

PS- I would love some insight from RW operators who piloted the 744 through the NG retrofits- their experiences, whether there was an "adjustment" period, and the use of the MCP through these types of approaches.  I'm a firm believer in the necessity of interaction with the automatics.  This might be due to my single digit IQ,  but it tends to keep me more in the loop.  I'd like to know if that's just an abundance of caution on my part, or the advocated methodology IRL?

[Alec246]

Hardy,

What you have done alone, on a complete 747 simulator, is an achievement no multiteam addon developer for other sims has come close to reach. Kudos for that.

I can't even imagine all the complex simulation that goes on with a Fixed Radius Leg, taking into account so many variables.

I tried it again, Wind Calm, Flaps 30 stabilized before the final continuous turn to Madeira, and here's what I got:

https://imgur.com/ZUE4j0R

The Flightpath Trend on the ND is showing the overshoot, and the AP continue the turn with same bank until it's too late and it has to go the other way around. If I was handflying it, I would keep that trend above the Magenta line, and everything would turn out fine, if I get it's behavior correctly?

[Hardy Heinlin]

Alexis,

it looks easy for a human being. It's not as easy for a robot. A human can easily grab a fresh chicken egg without destroying it; the finger pressure is just as high to keep it in the hand, and just as low to keep the eggshell intact. To develop such a control system for a robot involves lots of sensor inputs, and sensor evaluation algorithms (acceleration calculations in multiple dimensions), then control reactions (also in multiple dimensions), reaction evaluations, mechanical motion calculations, then functions to supress control oscillations etc. At the end, the human observer just sits there and wonders: "Why is this robot so stupid and cracks that eggshell? Why doesn't it just relieve the finger pressure?"

Anyway, I got an idea yesterday for a new attempt. Maybe it will improve it. Or maybe not. No guarantee. And no release date.


Regards,

|-|ardy

[Jeroen Hoppenbrouwers]

Anyway, I got an idea yesterday for a new attempt. Maybe it will improve it. Or maybe not. No guarantee. And no release date.

Hardy is getting closer and closer to real avionics development.   :-P

[cavaricooper]

Anyway, I got an idea yesterday for a new attempt. Maybe it will improve it. Or maybe not. No guarantee. And no release date.

Regards,

|-|ardy

Seattle is already humming with activity.  Black helicopters are spooling up.  There are lots of fingers poised on notepads throughout the Sound.  It will eventually be as the rest of the simulation is- unrivaled. 

Shortly thereafter, the Boeing 744 NG FMCs will get a software revision

C

[Britjet]

'The Boeing engineers are better than me."
(|-|ardy)

Hahahahahahahahahahahahahahahahaha!

[Alec246]

LOL

That would be true if the machines were made by common enginners, like weak MIT ones. If you Hardy made the machine, we would end up with a Truffle Omelete, with Scrapes of Parmigiano Reggiano on top, and a Glass of Wine to go with.

Looking forward to the update that's going to make Boeing Enginners loose their jobs

[mark744]

PSX NGFMC is already more reliable than the real one 

[Hardy Heinlin]

The LNAV guidance for RF leg transitions has been fine-tuned in PSX 10.100:

http://aerowinx.com/board/index.php?topic=4191.0


|-|ardy

[beat578]

Anyway, I got an idea yesterday for a new attempt. Maybe it will improve it. Or maybe not. No guarantee. And no release date.

|-|ardy

Wow. I don't even understand half of the topic and BAM, release is 10 days later... I wish I had a fragment of your brain and thinking speed (and programming skills). Great work! Thanks for all the effort that goes into that simulator!

[cavaricooper]

The LNAV guidance for RF leg transitions has been fine-tuned in PSX 10.100:

http://aerowinx.com/board/index.php?topic=4191.0


|-|ardy

Looking forward to fully exploring and appreciating this. For some reason there are several people in Seattle that are more excited than I.... a RW fleet wide FMC revision is no doubt forthcoming;)

Ta- C

[cavaricooper]

Hardy-

I just spent some time with 10.100- Bravo!  The overall LNAV track is SO MUCH better.  There is now only a very slight momentary hesitation away from the course line before BUTSE and TAGUY.  Almost like the smoothing addition runs out right before/at the waypoint change.

Is it possible to further disallow any corrective turn away from the general cardinal course direction, leaving all correction to a decrease or increase in bank towards the new quadrant/cardinal course i.e.- only steeper or less steep turns in the direction of the change?

I realize my 30 minute session pales in comparison to the many hours of work these few lines request, but I firmly believe no one wants PSX more accurate than you do, and so I report my observations here.  I welcome more useful input from RW 744 operators, past and present.

Obliged as ever- C

[Hardy Heinlin]

Not possible.

[Alec246]

While I cannot say I have tested very extensively, I can tell the Madeira RNP Approach is much better flown now, with better stability and less deviation. A step in the right direction for sure!

[David Palmer]

...
Anyway, I got an idea yesterday for a new attempt. Maybe it will improve it. Or maybe not. No guarantee. And no release date.

Regards,

|-|ardy

Did this idea happen to implement a 'Kalman Filter' by any chance?

Regards,
David.

[Hardy Heinlin]

No Kalman filter. I fine-tuned some parameters of my existing algorithm (some months ago).


Regards,

|-|ardy