This article is from the Flight Simulators FAQ
http://stason.org/TULARC/sports/flig...ors/index.html
What do you really mean by "realism" in a flight simulator?
There are two major issues to consider: realism and fidelity.
- Realism is how "real" a system feels.
- Fidelity relates to the actual models used.
Realism is a highly subjective issue: a simulator might model each blade of grass on the approach end of a runway, but if the user's flying overhead at 37,000', that won't affect his perception of *realism* at that point. Similarly, a graphics system might provide a high- resolution database, but if it only uses an orthographic projection, it won't win over many pilots!
Simulation is, therefore, the *art* of providing the expected cues and response characteristics for a specified mission.
Most military simulators are so specialized that they're optimized for certain missions or flight regimes; airline simulators tend to be much more flexible (all regime).
For each regime, appropriate feedback must be maintained. Real-world systems models are usually (but not always) the cornerstone of high-fidelity simulation; final "realism", even in airline simulators, is obtained only after an exhaustive survey and fine-tuning process. The acceptance process for even a production-run simulator can take up to a year.
A third issue is perception, and the intent of the game as an entertainment product. For example,
pilots realize that airplanes are essentially very easy to fly and land: non-pilots may expect them to be horrifyingly complex to fly, given a lot of the mystique surrounding aviation, a lot of
which has been enthusiastically promoted by pilots themselves. 
All retail flight simulators are just games, and, to some degree, help shape and feed off the perceptions of their users.
So if the users expect an F-16 to be almost impossible to fly, an F-16 simulator that IS almost impossible to fly wouldn't disappoint anyone except real pilots. Conversely, a simulator that is actually easy to fly might disappoint game-players as too easy, or "arcade-ish," because it IS too realistic.
In discussing "realism," one should really pay attention to three factors:
Code:
1. The flight dynamics and flight instrumentation. (flight simulator)
2. The visual system.
3. The systems support. (systems simulator)
The basis for such discussions in this forum should be from the pilot, not entertainment, perspective.
How does the flight model influence fidelity?
How does the flight model influence fidelity?
True fidelity in a flight simulator comes from the flight equations used in the flight model. In general, the more complex the flight model, the better performance you are going to get, though there are instances where even a *good* flight model can lead to poor flight simulation (more on that in a minute).
In general, most of the low-end, low-cost simulators on the market use what is known as a "3 Degree of Freedom", or 3 DOF, flight model. This means that the equations of motion only determine x, y, and z displacements of the aircraft in space, and then use this information to determine the flight attitude. The actual characteristics are based on the so-called "performance" equations, which themselves are usually only defined for steady-state situations. Various other characteristics, such as roll rate, must be fudged by the author.
Some simulations don't even pay any attention to angle of attack, using stick input or airplane pitch as the final determining characteristic.
Most high-end simulators use a 6 DOF model, described below, and a lot of PC-based simulators tend to ignore these kinds of models completely, and rely on a "point-space" performance model instead.
The equations of motion do not make the flight model, however; they merely set the limit on what is and is not possible. In order to support these equations, you must also have good models for finding the lift-curve slope, drag coefficients, stability derivatives, and other parameters.
In addition, you have to decide how you want to calculate these parameters. Should you calculate your lift on each wing independently, or just the lift on the whole wing surface area? The latter method would be faster and easier, but the former would allow you to model such flight dynamics as the "Dutch roll" modes, stall-spin conditions, and other common effects. How about downwash effects, which alter the effective angle of attack of the tail? There are several issues that need to be considered.
Finally, after you have your flight model together, you need to find the parameters that fit your aircraft, so that your *plane* flies just as realistically as your flight model does. For example, you could have a high-end, 6 DOF flight model, but if your Cessna 172 has the wrong wing area modeled, it won't *fly* like one.