February 28, 2005

Summary

Feasiblity Study for Open Source Glass Cockpits for Experimental Aircraft, by Phil Cobbin

This report summarizes research to date on feasability of construction of glass cockpit for experimental aircraft using off the shelf hardware and open source software components. The central focus of the study has been to investigate the feasibility of using the open source “Flightgear” flight simulation package to generate realtime synthetic vision for a head down display (HDD). The first part of this document reviews the sofware being investigated, followed by a summary of some of the hardware under consideration. This report is not meant to be an exhaustive review of the relevent hardware and software issues, but merely a summary of results obtained so far, for what is a speculative feasibility study.

This study is the partial result of ongoing efforts in construction of a Van's RV-7A experimental aircraft and the author's interest in the subject matter after discussion of synthetic flight display research with NASA officials at Oshkosh 2003. It being the author's understanding that some of NASA's flight testing of synthetic vision for IFR flight was conducted using a Linux box running a realtime operating system (RTOS), most likely a variant of RTLinux.

A pet interest of the author in the study being the prospect of blurring the distinction between simulation software and control system display software. Conceptually, if you can using the flight simulation to display digitally captured flight data then it one further step to integration of such a system with digital auto-pilot.

Flightgear

The author first identified Flightgear a few years back when the program was in it's infancy. The program has matured into a sophisticated open source flight sim which supports multiple flight dynamic models. Flightgear uses XML files to configure HUD displays and so forth which minimizes the need to write or re-write the extensive inventory of C++ code. In terms of architecture, flightgear is a simulation executive built on top of/using a set of other software libraries for such things as realtime rendering of the synthetic out the window (OTW) view using the SIMGEAR, PLIB and other libraries. Flighgear's databases are derived from government sources for both terrain and airports/navaids. At this juncture the author is still reviewing the subsidiary programs used to convert the raw government terrain data into flightgear's terrain file formats.

Flightgear is currently running on the author's Fedora Core Three Linux box which has a 1.1 ghz Athlon cpu. and NVIDIA graphics acceleration board. It is important to note the scene rendering supported by flight gear makes graphics acceleration mandatory. A substantial amount of time was required to get the NVIDIA module to run properly under Fedora Core Three requiring one to boot into run level 3 under Linux, manually insmod the driver then change to run level 5 to get the X window system running with the acceleration working. This seems to be a Fedora Core issue as it appears to use graphics features of the hardware early in the boot process such that a solution to the module loading may require it to be scripted early in the boot process. At this juncture the solution is in the “first you make it work...then you make it elegant” genre.

With a few minor tweaks to the source code a HDD was constructed to evaluate the feasibility of real time display of both the synthetic vision view and useful flight data integrated in a HUD type format.



The synthetic vision displayed is being rendered at a frame rate of 15 frames per second. A second instance of Flightgear was running at the time to feed data to the application. The second application of Flightgear was rendering a scene with visibility synthetically degraded to represent severe IFR conditions.



The HDD rendering displays airspeed, VSI, altitute (MSL/AGL), heading, longitude and lattitude. The AGL altitude is computed on the fly, based on the state of the data model. Obviously, there are issues of errors between what is displayed and reality when real hardware and conditions are in the loop.

A core developer of Flightgear is associated with the University of Minnesota and has applied the scene rendering technology of Flightgear to synthetic vision display in a test environment were a state police car was driven via synthetic vision only around a calibrated course.

KFLOG

A second area of glass panel display software is the real-time generation of moving map displays to show general situation awareness and specifics as to airports and navaids in the vicinity of an aircraft. The open source KFLOG program from Germany was developed for use by glider pilots to evaluate their flight performance for competitions where a GPS device is used to track the flight and KFLOG is used to evaluate flight performance.

Part of KFLOG's capabilities includes a map generation system which displays topo maps using color coding to designate altitude. KFLOG's mapping abilitity includes a mouse over feature which tool tips the airspace specifications of surrounding airports. The derivative CUMULUS project is a modified version of KFLOG for use on hand helds which does moving maps.

I am currently working through some compiling errors on CUMULUS. KFLOG, on the other hand, has been compiled successfully from source on the author's system. After some burrowing around the source code a test modification to the software was made to test out the feasibility of generating topo maps with terrain avoidance built into the display.



As the above sample illustrates, KFLOG with a little tweaking can be made to display a terrain topo's with selected regions above a set altitude set to red=dead.

At this juncture the author is considering merging the CUMULUS and KFLOG source code materials to integrate CUMULUS as a part of a modified KFLOG. CUMULUS, in addition to having the i/o built in to read NMEA GPS data includes features for modifying the depiction of airports in the map as to whether or not they are considered to be within the aircraft (glider's) glide range. A potentially useful safety application.

Hardware

To make a glass cockpit display requires a range of hardware capable of performing in an aviation environment. Starting with the display, the display in a cockpit environment like that of the Van's series of aircraft requires that the monitor be of a the sunlight readable monitor (SRM) classification . These monitors are typically 5 times brighter that other TFT LCD displays and significantly more expensive. Oddly, the primary obstacle (other than cost) for these displays seems to be finding one in the 10 inch range now as they are typically 12 inches and larger. A commercial grade rugged SRM easily can cost $1400.

On the CPU front, in car computing has been driving small form factor computer costs down and ruggedness up. A 1.1GHZ Mini-ITX motherboard can be acquired in the $150 range. If one can get by without a SRM display, 10 inch VGA displays for in car use can be had for a couple hundred dollars. The primary drawback seems to be the operating temperature range of the displays.

The computing power and display means for a glass display are, relatively speaking, the easy part, the complex and expensive part of the system is solid state gyro's required for the attitude reference system. NMEA compliant GPS transmitters can be obtained for the $100 +/- range with one major caveat. GPS transmitters typically broadcast on one hertz. The attitude reference system hardware is in the 20+ hertz range.

The core hardware required for a synthetic flight display is the attitude reference sensor. These devices typically integrated roll pitch and yaw position along with the relevant access acceleration rates. Among the producers of these devices are:

KVH: A maker of tactical grade inertial navigation devices featuring the companies patented fiber optic gyroscopic and compass sensors. KVH looks to be supplier of sensors of military grade and I found them from reviewing Armidillo Aerospace materials for the groups rocket project which use KVH sensors.

Crossbow Technology: Crossbow manufactures the device used by Chilton and the company exhibited their three axis sensor at Oshkosh last summer. A strapdown attitude indicator starts at about $7500.

PCFlightSystems manufactures the eGYRO-XP sensor which is intended as a portable strap down sensor. According to the companie's literature the device is not intended for permanent installation. The three access sensor is priced under $1000 and outputs at greater than 20 hz. This sensor has some interesting limitations and one considering using the sensor should review the technical specs. Oddly, the sensor will not apparently work properly if used in automotive platform or manipulated by hand. Conversly, the crossbow sensor excelled on the dynamic platform used to illustrate that sensor at Oshkosh.

BGI: Systron Donner Inertial Division manufactures a sensor for high end applcations. The device(s) are quoted in the $24k range.

Watson Industries, located in Eau Claire, Wisconsin: Watson industries manufactures AHRS specifically for aircraft. The following is quote from a recent email from the company president, "Analog Devices does not have any plans to make a decent gyro, just a cheap gyro. Their gyro has about 1/100 the performance of our gyro and is not able to meet an FAA level of performance. Only Murata and Tokin have worse gyros. The next best are Silicone Sensing Systems and Systron Donner which are about 10 time better than Analog Devices, but still only with the most extreme of efforts can they be marginally acceptable. Currently, all the AHRS manufacturers use Systron Donner except Crossbow and Watson Industries. Crossbow uses Silicone Sensing Systems for their low cost products and a fiber optic gyro for their higher priced AHRS."

The Watson AHRS system is about $10k.

Conclusion

This report summarizes some results to date on a feasibility study for construction of an open source based glass cockpit to display real time moving maps for navigation assistance and a synthetic out the window (OTW) display in a HUD like format, sometimes referred to as a Head Down Display or HDD. The focus of the study has been to investigate the feasibility of using flight simulation software as a real time display engine.

February 22, 2005

Moose Tail








Back in the Fall we did the initial drilling on the vertical stabilizer attachment plate. I can't believe it's taken me this long to get around to finish drilling the rest of the holes (plus a few odd hard to get at rivets near the hinge brackets that awaited a new yoke from Oshkosh for the squeezer, back in August). Slow but sure.

I thought fabricating and fitting the petot tube lines in the left wing would be a bear, but after doing the fuel vent lines in the fusalage it was a piece of cake. Do the 90 degree bend and fitting first, then fish the tube through the ribs. I figured while I was at it I should do some platenut riveting (access plates) and platenut drill details that are left over from the fall on the wings.

I have a few more punch list items to attend to now that I took the horizontal and vertical stabilizer out of storage and clecoed them to the fusalage. At this point there isn't to much left to primer and debur until the finish kit arrives. Famous last words...

With the tail feathers back on the fusalage I figured I'd move the moose to the tail.

Meanwhile, the top of chimney at home plugged up with creosote, causing some interesting stove venting... NOT! So I had to do a little impromptu ice climbing on the roof and get out a pair of crampons. There's going to be a payback for front pointing up to the crown of the roof. The hat vent was so plugged I had to use the pick on an old walking ice axe to chop the crud out!

February 21, 2005

Doubler Plate






Came up with a simple doubler plate for the bulkhead to facilitate drilling the final holes for rivets to to tie the stiffener to the bulkhead. The original drillings were now in the wrong place. These two plates minimized the number of new holes by using one of the holes for the earlier design on the left plate and two on the right. When you consider how little shear load those side flanges are likely to contribute in a load to failure scenario, the bulkhead connection is now in the brick outhouse category for over engineering... Hopefully.

February 20, 2005

Stiffener










Now that we've had time to figure out how to install the new design for the roll bar stiffener, we were finally able to get the last riveting done on the roll bar itself. We were also able to get the bulk of the drilling and fitting of the new stiffener with our high point awaiting fabrication of two doubler plates to drill the last two pairs of stiffener attachment holes.

The doublers were suggested by Van's tech support to get around a misalignment with the holes already drilled in the bulkhead for the older, and now obsolete, stiffener design. After hearing lots of horror stories of how hard it is to rivet the roll bar together, ours came out pretty good, I thought, with all the joints being tight. I think the jig work and fiddling we did up front to fit the parts together paid off, particularly the extensive use of welding clamps in the drilling process.

Hopefully in about a month or so the finish kit will be arriving. Meanwhile, I have quite a punchlist of details all over the airframe to go back and finish.

February 18, 2005

Reworking



The revised parts for the roll bar came yesterday from Van's. I had hoped the new hole layout on the flanges of the roll bar stiffener would line up with the old holes I had taken care to NC drill to get everything in the right place.

No such luck. As the jpeg shows there is a 1/8" difference in hole locations between the designs. I considered chopping off the flange ears and using my original brackets. E-mailed Van's customer service on the matter. Gus Funnell at Van's got back to me right away and suggested considering an easier route to fix the issue: Install a doubler plate. That's probably what I'll wind up doing.

Meanwhile, we've been a bit under the weather with a nasty respitory bug/flu that clipped Cathy starting last week so now it's my turn... Coughing so hard I thought I was going to crack a rib.

On the software hardware fronts I have a sample of a 300 degree per second rated MEMS gyro from analog devices to play around with, mainly to try to understand the technology. I suspect by this next Oshkosh we will see more vendors providing a wide range of solid state gyro options.

I'm still looking at this as a VFR exercise to see what kind of terrain avoidance and synthetic vision options can be built using off the shelf components. You have to tip your hat the sophistication and capabilities of much of the open source software that is available to work with. I am currently burrowing into the databases and software required for scene generation in synthetic vision... Kind of like wandering through the bowels of an aircraft carrier with only a penlight.

February 09, 2005

Open Source


This is a screen shot of the open-source flightgear running on a 1.1 gHz Linux box. I hacked and chopped the XML setup files for Flightgear's A4 aircraft to see what one could put together for a synthetic vision head down display, i.e. HUD minus the projector. The simulation is running over 10 frames a second on the display generation, so the theory is, you strip out the simulation executive and drive the model with data input from a flight computer (running real time Linux) which samples the sensors. You may be able to generate the display for the cockpit using the growing list of vehicle survivable TFT/LCD display screens in the 6-inch to 10-inch variety.

Flightgear's terrain database is built from government published digital terrain data... Hint: I've ordered the DVDs for the world, which chew up about 12 gigs but what the heck, disk drives are cheap. We'll see how the terrain looks around Lebanon and compare that to the kflog maps... and over course the sectionals.

At this juncture I would guess that two small form factor Linux single board computers would do the trick for data collection (2x redundancy) and two more to drive two 6, 8, or 10-inch displays with each computer/display being able to pop the synthetic vision panel as well as an engine status panel. Next step, I have to get the 0.98 source of flight gear to compile and run on my Linux box (over the binary 0.96 I'm now running) ...then it's the black hole on time: Burrowing into the source codes.

Funny Story: The Flightgear Guru hails out of the University of Minnesota. They had a project where they did detailed digital maps of Brainard Interational Speedway and as part of a safe plowing concept study for working in a whiteout, outfitted a state police cruiser with GPS and a horde of sensors to drive a synthetic vision display... Then let real state police run the race course with only the monitor for the primary display of what's out front. The author, knowing the limitation of the system, drove at 35. The state cops, having infinite faith in technology, literally put the pedal to the metal and cranked along at 90, with only the monitor running Flightgear's synthetic vision for an indication of what's really going on out front.

Wonder what you can do if you only put it together for VFR reference and flight planning. Hmm...

February 08, 2005

Whoops!


When I last primed a slew of parts for the roll bar my paint gun plugged, so it's been a while 'til I got around to priming the last of the small parts so I can finish the roll bar assembly... or so I thought.

A few days after finishing drilling and fitting of the roll bar, Van's issued a mod to the aircraft in response to someone folding a nose gear and doing the nose-over-end routine which evidenced a weakness in the roll bar design. I've emailed Van's for the new parts to make the mods.

The tie-in of the top stiffener from the roll bar to the bulkhead looks to have a weak spot on load transmission of the stiffener to the fusalage structure. It's been a bit warmer (for a bit) here so I thought I'd burn some wood in the garage to heat the place so I could finish priming parts before the monthly EAA chapter meeting.