On Moday of this week I had a long conversation with Matt Shaw, lead Designer for Fighter Duel 2.0. The original Fighter Duel is a high-performance dogfight simulator where you can fly one of sixteen classic World War II warbirds against up to eight enemy. It is connectable, which is its design goal, to allow you to fly against friends via network or kali. The head to head multiplayer was unusually stable in the NetDuel version, with very little warping and very few connection problems. Not only that, but frame rate was amazing given the hardware most of us ran in those days.

Not long ago we ran an update to our original interview with more info on the flight modelling, physics and ballistics. The point of that update was to demonstrate that these guys are passionate about what they do, and if you are a prop fan this sim should have you on the edge of your chair.

Then a couple of weeks back I got some stick time and an hour or so to personally observe what Matt and Ted and crew have been up to in development of this sim. Was I impressed? Do ducks fly? =8-D Dang! This thing is looking great! Better still, you will see it in Q4 of this year. For that complete hands on experience see our briefing: FD2.0 Intel.

Click for a larger shot...

When I saw Fighter Duel 2.0 it was running on a generic P1 233 MMX under the ancient Voodoo 1 technology at 800x600. Flight modelling and physics and ballistics were pretty much complete. I witnessed full articulation on the gear and flaps, at a level of physical detailing I haven't seen anywhere else. Three flaps on each side of the Corsair all take their proper place. Each is modelled separately. Each contributes appropriately to lift and drag. Each can be damaged.

I'm rehearsing some of this detail because early this week I called up Matt Shaw to ask him some questions about what I had seen and had not seen in Las Vegas a few weeks ago. Some of his answers surprised me! I knew these guys were obsessive about detail, but they are downright crazy!! So, without further ado, lets get into it!

Matt picked up his phone in his office. His creaky chair punctuated the eerie pauses. I could picture the room.. gloomy, with smoke hanging in the air from the Havana cigar that trailed from Matts lips...

So, you wanna know more about FD2 huh?

Er, yeah... is that okay?

Jest make sure ya tell it like it is, or I'll have ta send the boys up there, unnerstan?

Sure, Matt, no problem. So... what are you guys really doing up there? Got yerselves a Cray er what?

Cray, shmay.. we do most of the calcs in our heads and then input the optimized code for the program..."

Well, ok, thats not EXACTLY the conversation.. check it out for yourself:

CSIM: Tell us a bit about the training format.

Matt: Training is optional. YOu have the option while in any campaign or combat mission to stop and select a different aircraft. Let me explain the different modes in Fighter Duel 2.0. We have the training mode, the duel mode, combat and campaign.

The duel modes are very similar to FD original where you have AI controlled enemies or one on one online. You can setup any scenario you like in the duel mode against AI aircraft, choosing one on one, or specifying the type and number of enemy you will fly against, as well as their skill level and the starting orientation of the engagement.

CSIM: What about wingman AI? This is completely new to Fighter Duel since we didn't have AI wingmen and combat missions in FD original.

Matt: You are the flight leader for a squadron. YOu control your wingman by a series of commands, the same general list you see elsewhere: cover me, break, attack target etc. You also have command over the other flight leader, in terms of setting general objectives.

The basic system if you choose individual combat missions is that these are missions from the campaign, but now you have complete control over how you fly them. The whole conception of Fighter Duel is "what if?" As you progress in the campaign your selection of aircraft changes, in part according to your performance. And this also affects the enemy and what they can field against you. Now for the separate combat missions you choose what you want, and this time you have the full selection of aircraft to fly.

CSIM: So there is actual resource management integrated into this, pilot skill levels and all kinds of things?

Matt: Yes. Basically the resources you are trying to manage are reasonably that of squadron from the standpoint of the flight leader. What will happen is if you do very well in the campaign your squadron pilots will increase in knowledge and skill and will continue to grow in skill while they live. If they die, and if your losses are heavy, you will be drawing from a pool of beginners who will be fairly green. So we've taken strategy from WWII role playing games and integrated it. It requires little effort on your part to manage other than succeeding in your war efforts and keeping your pilots alive.

CSIM: Are there pilot records?

Matt: Not as such. We may or may not name pilots.

CSIM: How will we know if the pilots in our squad are green or intermediate or...?

Matt: You will have some indication how good they are. Like in real life, its kind of general.. you will know their overall strengths. We have factors for skill in general, and skill in a particular aircraft. As in real life, throughout the war you will shift aircraft. In the campaign you have some control over this with the pool of aircraft. For example... if you start out in an Allied squadron and have chosen to fly as an American you will have very early aircraft like a Wildcat or a Buffalo. If you like, you can fly that Buffalo through the entire campaign. But your pool will increase with more advanced aircraft as it actually did.

CSIM: Tell us more about comms. Will you model a radio? Will we communicate with the tower or ground?

Matt: Communication is between you and others in your squadron. We were considering it, but it doesn't seem that worthwhile.

CSIM: What sources are you using for flight models and avionics data, are they different than what you used for the original Fighter Duel?

Matt: Not really, its more cumulative. We've got a vast library of reference material including flight reports from the various aircraft, including rhetorical evidence. We have to take those with a grain of salt because sometimes we have reports that are quite different, or a memory can become embellished. Beyond that the biggest change is that the flight model now has hundreds of actual parameters in its algorithms.

CSIM: So now you can model lift along different parts of the airframe?

Matt: Yes, we actually can do that now. Before we had to take generalized numbers, but with an object based model we can calculate for each part.

CSIM: So this must apply to damage modelling also... this part is damaged, so its no longer producing lift or is producing turbulence or drag...

Matt: Absolutely. We went this way on purpose because we wanted to model different pieces and their functions and represent the whole aircraft in this way. In Duel Two we have taken an object oriented design for the whole program. This makes it extendable.. we can add features to it without bringing the program to its knees or making it unstable.

CSIM: Sounds like a natural for an add on for home aircraft designers!

Matt: We have considered that, but our aircraft are REALLY complex. There are a lot of moving parts and its not something the average person could do. The biggest issue is that something like flight model adjustments, the numbers wouldn't mean much to the average person. And what happens in multiplayer?

CSIM: Tell us about AI. Will AI pilots have the same g force limitations as the virtual pilot? Do you model losing sight if the enemy pilot is pulling high g's and blacking out?

Matt: Yes, all of that. We've had to make major changes to AI in order to deal with wingman formations. Frankly, its much more complex than a simple attack AI. One of the reasons that people flew formations of the same aircraft is that this made it easier to keep up with the leader.

So we've also had to work in visibility and limitations. We block their views when we should. One of the differences in place here is that we are always modelling the virtual dogfight cockpit, and these tones of visibility are available to the AI algorithmically.

One of the parameters for increased skill in the AI pilot is visual acuity. This is a real factor among human pilots. As a pilot gains experience he becomes better at spotting aircraft further out.

The same with fatigue. Black outs and red outs have a cumulative effect. As an AI pilot grows tired he will be less efficient not only in flying but in his visual acuity, and he will be able to manage fewer g's and onset of g effects will be quicker. This is modelled for the virtual pilot also.

CSIM: WOW, this sounds pretty amazing. Must have been a lot of work...

Matt: Not really. Its an advantage of an object oriented simulation. Well.. in some ways its good and in some ways its bad. We have the capability to plug in so many features that its hard to know when to stop!

CSIM: Lets get specific on one particular airframe and consider the modelling. The P38 Lightning... how many areas on that aircraft can take damage?

Matt: Well for example, there are separate damage areas for the nacelles and the engines. WIngs are separate, the fuselage is separate and in the case of the P38 there are actually two. We can resolve this even further down for collision detection and we haven't decided how far we want to go. There is an issue for performance when you are doing polygon to polygon detection.

For example, since we are modelling individual bullets in the stream we can determine if you actually hit the left aileron. This can become very CPU intensive so we are working on some algorithms now.

And once you have this kind of ability, then you have to decide if you are going to do volume calculations to see if the bullet passed through the plane or not. We didn't want to do one of these "hit bubble" models, although we do use that as an initial check. So we say, "ok, here is the position of the bullet, its pretty close to this aircraft, did it intersect the path of this object?" If yes, then we go into some more detailed checks.

CSIM: I know that ballistics themselves have become very sophisticated. Does this even apply to altitude?

Matt: First, we don't have a generic gun system. Each gun is modelled with its own energy and muzzle velocity, and if you fire a gun for too long then the trajectory will be affected because we model some effects of gun overheating.

Each bullet is modelled, and air density is modelled though its not a large factor. The largest factors that are modelled that affect trajectory are drag and gravity. The next factor is that at higher altitudes the air temperature drops and the speed of sound will drop. When the bullet passes the sound barrier it effects the bullet. But then this affects an aircraft much more than a bullet because of compressibility.

CSIM:Is this same level of detailing present in aircraft systems?

Matt: One of the big differences is that in 2.0 we restrict your use of emergency power or WEP. We monitor engine temperature so that if someone overheats an engine they are going to have problems, at the least they will have reduced power. Also, to go beyond full military power you will have to hit a key. Engine efficiency is also modelled, and you may get into your aircraft one day and your engine won't perform exactly the same way it did yesterday.

Of course, if your radiator takes a hit you are going to overheat your engine. One of the advantages of a radial engine is that its air cooled and a lot more resistant to that kind of problem.

By the way, this efficiency factor also extends to the AI. In a case where the Corsair shouldn't turn and fight but should try to extend, a green pilot might turn and fight. If you pounce on a guy out of the sun he might panic and run. Or if a pilot is out of bullets he's NOT going to engage you, whereas in Duel One once they were engaged they would not give up. CSIM: Lets talk a bit about the campaign. What connects the missions?

Matt: We have levels. As you progress through the campaign, in each level, there are four to six missions. Depending on factors of randomness and success or failure will determine which missions you get when you advance to the next level.

For example, if you do very poorly in an offensive mission your next mission might be defensive. If you do well there your rating will improve and your crews overall ability will increase. Part of this is random also.

Now there is resource management here too. You're hitting the enemy and they are hitting you. Resources are lost somewhere, so if you destroy your primaries and some secondaries, the enemy will have less ability to a. field good pilots, b. field aircraft. The enemy is also advancing in their aircraft, and they have limited numbers of aircraft. If you are kicking ass all over the place missions will actuall y get easier for you. UNLESS.. the enemy rallies and throws everything they have at you to try to take you out...

CSIM: So there is an overall strategic AI happening too...

Matt. Yes. Its not incredibly apparent to you, as it wouldn't be to an average squadron leader. You don't control the whole war. Things happen, you 're told what to do, and you don't know exactly how tough the resistance will be today.

CSIM: I assume we are taking out hard targets, aircraft on the ground etc

Matt: thats part of your mission. Primary objective might be a strike to take out bunkers, secondary could be an airfield. If you can take out planes on the ground you may encounter less resistance next time. Or take out the AAA batteries..

CSIM: What about the naval aspect, how is that integrated?

Matt: We have ships, and aircraft carriers, and you will be tasked to engage them. They will also be defending themselves.

CSIM: And are they a resource factor?

Matt: Definitely. You may find, coming to base after an offensive mission, that your base is under attack from carrier based aircraft. If you had taken that ship out in the last mission your base wouldn't be under attack.

CSIM: Tell us about wind, turbulence, ground effect?

Matt: Yes, we have ground effect, and of course its different for different aircraft. We had to add the secondary landing flaps on the Me163 Komet because the plane has an incredible amount of ground effect. Turbulence is slated, its pretty simple to put in. We model altitude, compressibility so the ingredients are there. On a hot day these factors change.

CSIM: What about clouds?

Matt: We have fog, and we're working on ground patch fogging. We intend to have clouds, and we want to do them really nice. One of the problems is that we've really bumped up the graphics level of the program. We want them to match.

CSIM: Have you considered a mission builder? Might we see it later?

Matt: We considered it, but don't have enough time. So yes, you might see it in a later add-on along with some additional aircraft. Its not entirely up to us, our publisher will have some input.

CSIM: What about sound? Will we hear the difference between a P51 and a Messerschmidt?

Matt: We do have a lot of sounds sampled already and we'll try to get more. We do have 3d positional sound. Of course in a real war when you have these huge engines its pretty unlikely that you would hear the enemy go by you, but there is some of that, including doppler shift. In fact we have two settings available to you for 3d sound: you can select speakers or headphones, which require two different algorithms. Stereo separation for headphones is hard to beat!

CSIM: What are we looking at for a minimum spec machine?

Matt: We are still on this balance beam of how many features we can include. Generally the biggest hit after rendering is the polygon matrix. We're looking at this point at a 166 MMX and for best performance you will want a 3d board with it. Of course if you run this at 1600x1200 you'll need a lot more than a 166 MMX. We allow whatever resolution your machine will do.

CSIM: WOW... what does this look like under SLI mode with Voodoo2?

Matt: The new version of our rendering engine has very, very good improvements under Voodoo2. SLI mode is great, and the most striking change is how this runs under NT with dual processors. On this platform the Voodoo2 board is fed fully and reaches its full potential. Even NT with a single processor gives a good performance gain since 2.0 is fully threaded.

CSIM: What multiplayer options will be in place and how many players can we expect to see online with TCP/IP?

Matt: There are a number of issues here. First is that to ensure that all the missions would work online would take us years of testing. So the first set of missions that will release for multiplayer will be the duel missions. These will also work over a LAN. The other issue is whose machine will host this? We're not trying to create WARBIRDS here with server machines hosting the work. SO we're planning to allow eight or nine people to play, cooperatively or whatever won't matter. Special combat missions will be available, whether from the main list or not, that are multiplayer certified. Will we allow people to try the other missions even though they aren't certified? Maybe, or we can make other missions available via small patches.

CSIM: When will 2.0 hit beta?

Matt: It won't be June. As we've headed into the mission work we've found we have a lot to do so we are aiming at the end of the year.

CSIM: Thanks Matt, all the best with this and I look forward to seeing the latest at E3.