Board index FlightGear Development Aircraft Systems

Carousel experiments

Modeling aircraft systems, like electrical stuff, hydraulics, pneumatics? Feel free to ask support.

Carousel experiments

Postby zakalawe » Fri Feb 05, 2021 10:33 am

I'm making some experiments to create a re-usable simulation of a Delco Carousel (IV, if you're counting). Intention would be a 3D model, and a pop-out 2D window for ease of use (copying Stuart's method for the FG1000). This is motivated by Alant's work on creating a VC-10, since no 1960s airliner should be without a Carousel or three....

I've checked existing aircraft and the Concorde has a similar INS simulation, but the obvious candidate aircraft such as the 707 and 727 don't; Tristar and DC-10 would be the other ones I can guess. (Maybe the Trident as well?) If anyone knows of any existing prior work, please do share it here.

Unlike the Concorde solution, I'm hoping to build this on top of the route-manager / GPS code, so that you can load in waypoints from the GUI (of course with manual entry also available for the purists). That's going to mean a slight modification to the Carousel to allow more than ten waypoints; I'm still deciding how to handle that visually. Again with a purist mode with the exact original appearance, but of course then you're limited to maximum of ten waypoints in your flight-plan.

Probably this takes some time, I'll aim to get the simulation side (commands, property API) defined and then worry about the visuals after. Don't hold your breath :)
zakalawe
 
Posts: 1259
Joined: Sat Jul 19, 2008 5:48 pm
Location: Edinburgh, Scotland
Callsign: G-ZKLW
Version: next
OS: Mac

Re: Carousel experiments

Postby stuart » Fri Feb 05, 2021 1:24 pm

Hi James,

I recall a discussion (years ago probably) about modelling INS where there was a suggestion that the drift etc. would be modelled as a layer in the core code, so the aircraft implementation could simply use the route-manager/GPS interface. Is that what you're thinking of implementing here?

-Stuart
G-MWLX
User avatar
stuart
Moderator
 
Posts: 1629
Joined: Wed Nov 29, 2006 10:56 am
Location: Edinburgh
Callsign: G-MWLX

Re: Carousel experiments

Postby Alant » Fri Feb 05, 2021 3:03 pm

James. There are detailed instructions on the use of the VC10 INS in the BOAC Operations Flying Manual. Do you want a pdf copy of this?

The RAF VC10 Pilots Notes does not mention INS. It does have a dead reckoning system (Ground Position Indicator) that uses Doppler radar to measure distance travelled. Loran and Omega were also fitted.
The VC10 navigators station has plenty of panel space for all of these systems.

I made my own nasal based route manager for the TSR2´s INS. TSR2 navigation is not directly from waypoint to waypoint. The planned route passes each waypoint with a planned offset to port or starboard. The waypoints are chosen so that they are identifiable on the sideways looking radar, or visually if needed. The navigator uses these fixes to apply corrections to the INS. The waypoints were just the lat lon positions of the fix points. The navigator was supplied with a book of photos which he could match against the radar image and measure the error.
I did retain Flightgear´s basic data layout so there is some compatibility with Flightgear´s own route plans.
My TSR2 INS does not currently estimate errors. James will be pleased to hear that it was a Feranti unit.
The aircraft was not fitted with any radio aids - i.e. no VOR, ADF TACAN, Decca, Loran etc. These could be turned off or jammed during a military conflict.
Stuart´s FG1000 pop-out sounds interesting. I did make an autopilot controller as a menu item. It has the same layout as the aircraft AFCS panel. My plans were to make an AFCS panel using Phi that could be accessed from a notepad or smartphone on my knee, but never succeeded in making it interactive so the buttons and switches worked. A Phi based navigation panel is also on the to-do list.
The panel is cumbersome to operate with a mouse. Phi and a touch screen would be much easier.
Alant
 
Posts: 1219
Joined: Wed Jun 23, 2010 6:58 am
Location: Portugal
Callsign: Tarnish99
Version: latest Git
OS: Windows 10/11

Re: Carousel experiments

Postby Johan G » Fri Feb 05, 2021 3:11 pm

Some references:


Panel dimensions

Panel dimensions seem to be equal with mil-spec panels, nominally making the mounting panel 5.75 in wide and 4.5 in high [1][2] and the plastic lighting plate (in essence the top portion of the panel, with engraved back lit text) 0.046 in smaller in width and breadth than the mounting panel. [3]
____
[1] See Carousel Control Display Unit, Commercial Airline, Smithsonian National Air and Space Museum.
[2] MS25212C, Military Standard: Control Panel, Console Type, Aircraft Equipment, Basic Dimensions (25 AUG 1960)
[3] MIL-P-7788E, Military Specification: Panels, Information, Integrally Illuminated (15-MAY-1977), section 3.4.1.
Low-level flying — It's all fun and games till someone looses an engine. (Paraphrased from a YouTube video)
Improving the Dassault Mirage F1 (Wiki, Forum, GitLab. Work in slow progress)
Some YouTube videos
Johan G
Moderator
 
Posts: 6629
Joined: Fri Aug 06, 2010 6:33 pm
Location: Sweden
Callsign: SE-JG
IRC name: Johan_G
Version: 2020.3.4
OS: Windows 10, 64 bit

Re: Carousel experiments

Postby zakalawe » Fri Feb 05, 2021 3:51 pm

I already purchased the Delco Operators Manual from one of the 'buy old manuals in PDF form' : but I'll definitely take a copy of the BOAC manual as well.

Regarding exporting panels, I'd recommend Canvas for this: FGQCanvas runs nicely on iOS and should run easily on Android also. And no need to create a special UI, what we use inside FlightGear just works.

Stuart, as you guessed this is about the fact that the 'GPS' is really an arbitrary navigation system which defaults to use the real position; my plan for the Carousel is to add noise to the position proportional to acceleration & time-since-alignment : I previously tried to simulate the INS from first principles but that was numerically unstable; using a noise-filter I think we can establish a plausible drift rate, at least enough to make for interesting flying.
zakalawe
 
Posts: 1259
Joined: Sat Jul 19, 2008 5:48 pm
Location: Edinburgh, Scotland
Callsign: G-ZKLW
Version: next
OS: Mac

Re: Carousel experiments

Postby V12 » Fri Feb 05, 2021 11:13 pm

FG Concorde has only fake INS with zero position errors.

Somewhere I have manual for freeware FSX addon CIVA INS used on FSLabs Concorde X.
Fly high, fly fast - fly Concorde !
V12
 
Posts: 2757
Joined: Thu Jan 12, 2017 5:27 pm
Location: LZIB
Callsign: BAWV12

Re: Carousel experiments

Postby zakalawe » Sat Feb 06, 2021 6:11 pm

V12 wrote in Fri Feb 05, 2021 11:13 pm:FG Concorde has only fake INS with zero position errors.
The position is 'perfec't but apart from that it seems a pretty decent little INS to me. It's tided to the Concorde which is a problem, and it doesn't use the FG route-manager data which means manually entering waypoints (boooooo), but aside from that I'd say it's nicely executed.
zakalawe
 
Posts: 1259
Joined: Sat Jul 19, 2008 5:48 pm
Location: Edinburgh, Scotland
Callsign: G-ZKLW
Version: next
OS: Mac

Re: Carousel experiments

Postby V12 » Sun Feb 07, 2021 12:06 pm

Realistic INS simulation must calculate plane position based on the acceleration, velocity and heading. This method is not implemented in FG Concorde if I remember correctly. There is not simulated align augmentation by receiving 2 VOR - DME signals.
Fly high, fly fast - fly Concorde !
V12
 
Posts: 2757
Joined: Thu Jan 12, 2017 5:27 pm
Location: LZIB
Callsign: BAWV12

Re: Carousel experiments

Postby V12 » Tue Feb 23, 2021 8:10 am

Fly high, fly fast - fly Concorde !
V12
 
Posts: 2757
Joined: Thu Jan 12, 2017 5:27 pm
Location: LZIB
Callsign: BAWV12

Re: Carousel experiments

Postby Thorsten » Tue Feb 23, 2021 8:14 am

I previously tried to simulate the INS from first principles but that was numerically unstable; using a noise-filter I think we can establish a plausible drift rate, at least enough to make for interesting flying.


Shuttle area navigation is based on a Kalman filter incorporating TACAN, air data, manual state updates or MLS and GPS into an inertially-propagated state vector - and we have that running, so if you're looking for a working example of IMU drift...
Thorsten
 
Posts: 12490
Joined: Mon Nov 02, 2009 9:33 am

Re: Carousel experiments

Postby Alant » Tue Feb 23, 2021 10:01 am

The first job is to simulate a working INS panel. That will be sufficient for the casual user.
Simulating INS drift and how to apply corrections can then follow.
Remember that most Flightgear aircraft do not even simulate gyro compass drift and alignment with magnetic compass system correctly.
Alan
Alant
 
Posts: 1219
Joined: Wed Jun 23, 2010 6:58 am
Location: Portugal
Callsign: Tarnish99
Version: latest Git
OS: Windows 10/11

Re: Carousel experiments

Postby zakalawe » Tue Feb 23, 2021 11:01 am

Alant wrote in Tue Feb 23, 2021 10:01 am:The first job is to simulate a working INS panel. That will be sufficient for the casual user.
Simulating INS drift and how to apply corrections can then follow.


Yes that was my feeling also. It would be good to give tee GPS instrument code some options to specific the error/drift however, under property control, since it deliberately derives all its route-related data from the internal position, not the 'real' position. I.e once we add some error to the position, everything else will fall out more or less for free.
zakalawe
 
Posts: 1259
Joined: Sat Jul 19, 2008 5:48 pm
Location: Edinburgh, Scotland
Callsign: G-ZKLW
Version: next
OS: Mac

Re: Carousel experiments

Postby zakalawe » Tue Feb 23, 2021 11:04 am

Thorsten wrote in Tue Feb 23, 2021 8:14 am:Shuttle area navigation is based on a Kalman filter incorporating TACAN, air data, manual state updates or MLS and GPS into an inertially-propagated state vector - and we have that running, so if you're looking for a working example of IMU drift...


I guess the drift is coming from the 'inertial propagation' part? My original attempt was to directly integrate the measured acceleration values in /acceleration and therefore derive inertial velocity and position. Unfortunately even at 120Hz this blew up after only a few seconds. Probably needed to use a smarter integrator :)
zakalawe
 
Posts: 1259
Joined: Sat Jul 19, 2008 5:48 pm
Location: Edinburgh, Scotland
Callsign: G-ZKLW
Version: next
OS: Mac

Re: Carousel experiments

Postby Thorsten » Tue Feb 23, 2021 12:19 pm

I guess the drift is coming from the 'inertial propagation' part?


Well, we know the potential drift rates of the original IMU devices, so we simply did a random-noise model on the accelerometer errors to match that and simulate attitude, velocity and position drift from there.

An occasionally running housekeeping loop can keep track of the drift quite well - you can simulate any rates you like and set everything to perfect at user request, incorporate manual corrections...

The complete scheme is then about incorporating data - TACAN for instance has ranging and angular errors, so the filter generates the most likely state vector given the IMU sensings and the TACAN signal. I'm not sure many aircraft would use such a scheme though (and GPS is kind of the killer application anyway, with cm-scale precision, who needs inertial navigation any more...)

My original attempt was to directly integrate the measured acceleration values in /acceleration and therefore derive inertial velocity and position.


Conceptually interesting, but you never get to the original specifications that way... And it's terribly error-sennsitive, so I 'm not surprised that didn't work.
Thorsten
 
Posts: 12490
Joined: Mon Nov 02, 2009 9:33 am

Re: Carousel experiments

Postby zakalawe » Tue Feb 23, 2021 12:30 pm

Thorsten wrote in Tue Feb 23, 2021 12:19 pm:we simply did a random-noise model on the accelerometer errors to match that and simulate attitude, velocity and position drift from there.


Great, that sounds pretty much what I had in mind, I'm hoping the new noise filters in the XML can support it even, so I can feed the 'drifty' position into the 'GPS'.

The 'incorporate multiple data sources and find the computed position based on their quality' isn't needed for the Carousel but the Boeing FMCs do that: triangulating DMEs, and then computing FMC position based on radio position, GPS potions and the INSs. (I think there's eve some dynamic weighting) That needs a module for sure, outside the GPS code; it's on my ToDo list for the 737 FMC work, when I get back to it.
zakalawe
 
Posts: 1259
Joined: Sat Jul 19, 2008 5:48 pm
Location: Edinburgh, Scotland
Callsign: G-ZKLW
Version: next
OS: Mac


Return to Systems

Who is online

Users browsing this forum: No registered users and 1 guest