FTX

WHERE HAS TWENTY FIVE YEARS GONE?

The year is 1981, and it's my first introduction to Flight Simulator (A2-FS1 on the Apple II), from the Bruce Artwick Organization (BAO), whose FS engine is later to be aquired by Microsoft. To me, using FS1 was a revelation, and very exciting for an aviation nut whose father's idea of a family excursion was to take his kids to the observation deck of the local international airport!

Prior to discovering FS1 I was writing my own crude flight simulator code on devices as primitive as the HP41C calculator! Try doing approaches on that baby and surviving :)

Fast forward to 2006, and the FS community learns what Microsoft is planning in the latest version of the oldest continual PC game franchise in history:

Ground texture resolution increased from 5m/pixel to 1m/pixel and higher

Improved 3D engine with architectural enhancements and DirectX 10 support

A new SDK with improvement to the autogen and terrain engine

Naturally to me, this was very welcome news! During the creation of VOZ, the one thing which often hampered Matt Tomkins' and my own vision was the very low resolution of the ground terrain textures, which often resulted in a lot of design compromises being made. It was extraordinarily difficult to capture the unique feel of the diverse Australian terrain when the pallete was so limited. However the phenomenal success of VOZ and its adoption by the FS community on a global scale told me that to a large extent Matt and I had been able to create a unique feel for the textures even at the low 5m/pixel resolution of FS2004.

In moving to 1m/pixel textures we have both a solution to the fidelity and the added complexity of working on sixteen times the terrain information per square kilometer. Not only does it literally mean sixteen times the work to produce realistic terrain, it also places additional demands on landclass and the more seamless alignment of textures adjoining each other.

DOES FSX SOLVE THE PROBLEM?

Let me state from the outset that I am a huge fan of Microsoft's vision across many of its software products, and in particular, the tremendous talent and skills of the ACES team in Redmond for both their persistence with and the continual evolution of the Flight Simulator platform. It is commonly known within the FS development community that ACES themselves never intend to create a fully complete all-encompassing product which users just install at each new release and never enhance ever again. On the contrary, many in the ACES team have been quoted as saying their mission is first and foremost to create an extensible platform for the FS development community to build upon.

So this begs the question; did Microsoft raise the bar with the ground terrain in FSX? Of course the answer is going to be subjective, but from my point of view they certainly created the right foundation and gave developers enough of a "sandbox" with which to play with.

That said, the key (but in no way complete) ingredients for creating a believable world to fly over in the simulator are:

Accurate landclass

Region-specific textures

Seamless joining of unique tilesets

FSX out of the box has many improvements, including much improved terrain mesh, higher fidelity aircraft, weather and lighting systems and easier to work with APIs such as SimConnect.

Where it fails in my opinion, and has disappointed many users, is the less than accurate landclass, generic textures and the harshness with which those textures adjoin each other on the ground. In many places in the virtual depiction of the world, including many parts of North America, there is an abundance of "desert" like landscapes and very hostile and yellow-brown vistas await those who load flights in their favourite areas.

This is not entirely the fault of the textures per se, but more to do with the inaccurate landclass being used by default FSX. Of course, the textures themselves also are too generic to attempt to show the diversity of the all the world's unique geographical regions. As Phil Taylor from ACES once said in a TV interview, "We had to squeeze a terabyte of data into three DVDs" - an unenviable task indeed.

So it is clear some compromises had to be made in order to deliver a finished product to the users, and one of the victims of the huge increase in data size was the ground terrain textures. Closer inspection of the default textures in FSX shows source data dating back to FS2002 and extensive re-use of the same source images with only subtle variations between geographical regions and seasons. Simplly increasing the resolution from 5m/pixel to 1m/pixel whilst using original data which in some cases is over five years old will only highlight the flaws in the source data. In many cases the aerial photos have been sourced at 1m or lower resolution and do not have any "depth" of detail to offer a richer experience for the user flying over terrain.

If we now add to this dilemna the less than accurate landclass and harsh texture boundaries which prevail in FSX, the result is a barren and very synthetic looking landscape that in many cases bears little resemblance to the geographical region being depicted. This is further excacerbated by the lack of accurate seasonal regions which often have disastrous results in continents like Australia and also many parts of the world like Europe and South America.

SO HOW DO WE FIX IT?

It's not an impossible task to make things right, just a lot of work, a sensible and balanced delivery plan and the patience to take the time to do it right. And of course, if Phil Taylor had been given the budget to hire an additional sixteen full-time texture artists, carte blanche access to unlimited super high resolution aerial photography, then ACES would surely have delivered what they know is possible with their superb engine. And of course, FSX would then retail for about $1000 at your local store and be ready for release by 2012!

There are two possible approaches to creating an authentic terrain product in FSX. The first is a "full photoreal" product which sources complete aerial photo coverage of an entire city, region, country or even a whole continent and provides that as a single product. Great in theory, but I've never been a fan of large-scale, huge area photoreal products mainly because they are excruciatingly slow to load, consume multiple 100's of GB of hard disk space, and have to make compromised design decisions like no seasonal variations, poor or no autogen and only reasonable texture quality. My tune will change in years to come of course, when 50 terabyte solid state drives and 1 terabyte memory video cards and GPU/CPU power combine to make this approach a lot more feasible.

In the meantime, I prefer to employ the synthetic landclass-based approach, which combines descriptions of what should be on the ground with a collection of aerial photo based texture tiles, repeated in ways which create about an 85% accurate depiction of the real world. The benefits to this approach? That's easy! Infinitely faster loading times, a fraction of the hard disk space, full support for seasons, more detailed and accurate autogen, and less cost to source the data, allowing developers to "invest more in less", if that makes sense.

So how does a small startup company like Orbx do what is seemingly impossible to solve? The answer is simple: we don't tackle a "one product fits the whole world" solution to the problem. We don't believe in texture base packs which simply replace everything for the whole world. Rather, our approach is similar to what was pioneered with VOZ, in that we will create smaller regions within continents or countries, and source very specific aerial photography and landclass for those small regions only. If we create a single regions at a time that produce amazingly accurate depictions of each geographical area, we will hope users will embrace that level of fidelity and patiently wait for us to continue to re-invent the world landscape, one region at a time.

The cost load is then spread over a development cycle which we see spanning some five years or more, and as more region packs are released we can generate enough revenue to invest in more highly detaled aerial photography, landclass and region-specific data. FTX is planned to have a lifespan well into the FS11 release cycle.

So, adding to the list of key ingredients I listed above, here is what Orbx is adding to the recipe to create the Full Terrain Experience (FTX):

Licensed aerial photography sourced at 15cm per pixel

Licensed accurate landclass specific to each continent

Hiring expert texture artists who can craft the source data into FSX format

Adding the "Orbx" touch to boundary meshing, 3D night lighting and autogen

Removing the need for any texture "switcher" completely for the user

Seamless transition of textures as you fly across region boundaries!

Publishing an open SDK to allow developers to add airports and scenery

Plus a few "tricks" up our sleeves :)

IT'S ALL ABOUT THE DATA

There's an old saying in the IT world; "Garbage In, Garbage Out", and the same holds true for source data in the creation of FSX ground terrain. Orbx and its shareholders have gone to great lengths and considerable expense to license actual aerial photography from professional suppliers whose business is attaching high definition camera equipment to the belies of aircraft and mapping specific areas. Often aimed at the real estate, military, town planning, aerospace and other commercial customers, for the first time their data is now being re-purposed so Microsoft Flight Simulator users can see the fruits of having this data converted to FSX.

Here's the key to the increase in realsim and fidelity: if you source data at 15cm per pixel, even if it is resampled to the lower resolution of 1m per pixel, most of the source detail is preserved, simply by using sophisticated resampling software that our data suppliers have access to. This software can cost tens of thousands of dollars, and that's why Orbx is prepared to defer to outside expertise and obtain resampled data which loses little to no information in the downsizing process. We let the experts do the hard work for us, rather than spend endless hours in PhotoShop trying to reinvent the wheel.

Now, if we compare this process to say, grabbing screengrabs from Google Earth and trying to convert those into usable textures for FSX, we introduce a bunch of issues. Firstly, it's not data which is available for developers to repurpose as payware anyway, and secondly, despite some great resolution "pockets" of data that GE provides in certain cities of the world, for the most part it is based on satellite images taken hundreds of miles up in space using lenses and resampling which just cannot compete with a 12MP digital camera designed to be used on aircraft flying at 2500FT AGL.

Get the right data, and you'll get the right result. That's the philosophy behind FTX and I can assure you that once you see the results in FSX you will never, ever go back to default or FS2004 again. The unanimous opinion of the FTX beta testing team is that they rarely, if ever, run FS2004 again once they have flown over FTX landscapes. That's saying something.

THE FINAL PIECES

Let me finish off by adding the last few ingredients to the recipe for FTX, and the proof will be there for all of you to see for yourselves very soon indeed. We begin with a region based approach, add the best landclass we can source, create the highest fidelity textures sourced from actual photos of the region being crafted then finally, add these elements:

Re-engineer the texture joins to eliminate the jagged edges in FSX

Price each region so you don't need to mortgage your house to buy FTX

Provide world-class support and service to our users

A firm belief that we can influence the future of FS!

WELCOME TO THE FULL TERRAIN EXPERIENCE!

The entire FTX team and I are utterly convinced the next few years will be the most exciting time ever in the history of consumer flight simulation, and we have plans for technology and products which will truly change the game forever.

Come along for the ride won't you?