This might not be an actual answer to the question but I cannot resist to write something anyway!
I am part of a project (my part is very small but never the less) that builds an AJ
-37 Viggen simulator. This includes both cockpit and software. We have the ambition to make a 100% accurate simulation, from the cockpit to the flight characteristics.
I won't go into detail about the cockpit but we have the interior of an actual aircraft so it's pretty safe to say it's accurate. Software wise then... We use Flight Simulator X (migrating from 2004) as a base for the software. Why invent the wheel once again...
Now, the delicate part is to try to make the Viggen inside MFS to behave like the real thing. First of all, the project owner/leader (that also builds the cockpit and all associated hardware) is a pilot and flight instructor. This enables him to have a quite good guess about the basic handling of the aircraft.
The magic part however, is when all known (and unknown) real life parameters are entered into the configuration files of MFS. The guy that does this has done a tremendous work, reading documentation and asking people and in the end calculated and entered numbers into MFS. Part from documentation, we do have quite some very useful connections with people actually working (well, worked) with Viggen. Of course this is crucial in order to save effort and increase accuracy.
To this point everything can be considered to be qualified guesses though. What we do in the end to verify what we have implemented, is what most people do (or wanna do I guess): get the people that work with the real thing to fly the thing as much as possible, trying out every aspect of the simulation. In other words, we need a test pilot with real life experience.
And luckily, we've had several Viggen pilots flying our simulator and their feedback has been invaluable. They do not only provide us with feedback on specific details of the simulation, we also get a "general" impression from them. As one of them said after his first ride: "The first impression of the Viggen was that it really smells like a Viggen!". We then know we're pretty close. And when some other Viggen pilot says "This is as close you can get" we're really happy!
How do we perform a test flight? We're not professional aircraft manufacturers but we try to hang around the simulator, noting (mentally often) every thing the pilot observes. It can be anything from "that gauge should not accelerate in that way when I do this" to "the aircraft feels a little bit fast when applying air brakes in this speed range".
Iterating this process as much as possible hopefully (we think so anyway) has taken us so far so that it comes down to opinions rather than facts. Of course, there are so many things to simulate that we cannot get everything perfect or even near perfect. But the ambition is there anyway. And we're not done by any means (probably never will be) - there's always more functionality around the corner to implement!
The ironic part of it is, every now and then we do something that forces us to go back a few steps and do it all over again. For instance, going from 2004 to FSX forces us to revise all flight data and aircraft behavior since the latest iteration of MFS behaves a little different than the previous one.
In the end: how do we know what *is* correct and what *seems* correct? We don't make a difference. Since this is for amusement rides only (not education) we don't need to verify all the way. A real life pilots words that it feels accurate is enough for us.
To make a real simulator accurate is another story I guess. But the basic idea is the same, even though there's probably more formal methods of doing it.
Proud to part of www.novelair.com.