Let's consider some facts.
Starting first with the characteristics of a PC joystick:
Fact 1.1. For all PC joysticks the forward and rearward travel are the same.
Fact 1.2. The absolute amount of PC joystick handle travel in millimeters or inches is significantly less than that found in a real aircraft, since the PC joystick is much shorter than real control column - (with the exception of a few home brews, or when compared to certain aircraft that use joystick-like control sticks, the F-16 for example).
Fact 1.3. PC joysticks without Force Feedback use a spring, so with no force applied to them by the user they will always return to the neutral/centered stick position. Only FFB sticks allow the developers to control the zero-force stick position in the software, as well as the absolute amounts of stick forces. The FFB joysticks without a spring (like MS FFB2) are more precise than the ones that, along with FFB, use a spring or rubber neck (Saitek Cyborg Evo Force or Logitech Force 3D Pro).
Fact 1.4. Joystick spring (or FFB motors) are capable of creating only a small fraction of the amount of force that a pilot would have to contend with in a real flight.
1.jpg 27.01KB 7919 downloads
Now on to the facts related to airplanes:
Fact 2.1. For the majority of airplanes, the upward deflection of the elevator control surface is greater than the downward deflection. As a result, the control column/stick in the cockpit has more travel aft than it has forward travel (from aerodynamic center).
Fact 2.2. For the most part, the airflow around horizontal stabilizers is complex, resulting in asymmetrical pressures on the upper surface of the elevators compared with the lower surface of the elevators. That's why, if the control column is let free or control lines are severed, the elevators will practically never end up with zero deflection, albeit close to it. The resulting elevator deflection depends on a large number of factors, related to the aerodynamics of the flow around tail surfaces of an airplane (which itself is related to many factors).
Fact 2.3. It follows from the fact 2.2 that for airplanes with a simple mechanical control linkage that the zero-force control column position rarely coincides with the geometrically neutral column position, albeit close to it. Given fact 2.1 it follows that the zero-force control column. Position is rarely going to be half-way between full-aft and full-forward control column positions, and on top of that the zero-force control column position constantly changes in flight depending on the airflow around the airplane tail surfaces (which changes with airspeed, propeller RPM, Angle of Attack, side sleep, etc).
2.jpg 44.4KB 8052 downloads
Fact 2.4. For each value of elevator deflection (and corresponding control column position) in flight, there is a particular value of Angle of Attack (AoA) that the airplane will settle to. This correspondence changes very little for a wide range of speeds, so we can consider it constant for any practical purpose (we are not considering supersonic flight ). Engine RPM is also a factor here - due to prop wash effect on the airplane surfaces - but this effect does not change anything in principle - the airplane AoA is defined by elevators deflection (and control column position).
Fact 2.5. The greater the airspeed the greater the force of lift (and G load) given the same value of AoA of the airplane.
Fact 2.6. It follows from fact 2.5, that in order to maintain level flight at constant altitude, with an increase in airspeed the AoA will have to be reduced to prevent climbing. Conversely, AoA must increase when airspeed is reduced to avoid descending. So in order to maintain constant altitude, level flight, for low airspeeds the airplane must have a high AoA, and for high speeds a low AoA.
Fact 2.7. It follows from 2.4 and 2.6 that in level flight for different airspeeds that the elevator deflection will be different in order to provide the required AoA. While some modern aircraft might do this automatically, for airplanes with only basic flight controls (like those of WWI vintage), the pilot will be required to move the control column forward as airspeed increases, and conversely, to move the control column aft as airspeed decreases. So unless the aircraft is equipped with a cockpit adjustable trim system to relieve these forces the pilot will be holding constant pressure on the controls.
Now let's analyze the above facts.
If you consider fact 1.1 and fact 2.1, it is evident that we can not make each position of the PC joystick handle correspond with the position of the control column in the real cockpit exactly, without any geometrical distortions. There are 3 simple and obvious solutions to this problem:
- Max aft deflection of the real aircraft control column corresponds to max aft PC joystick position.
- Max forward deflection of the real aircraft control column corresponds to the maximum forward PC joystick position.
- Between these two extremes there is direct linear dependency between the real aircraft control column and the position of the PC joystick.
This means that the neutral control column position does not coincide with the neutral PC joystick handle position, and the elevator control surfaces may be deflected somewhat when the PC joystick is centered.
3.jpg 65.7KB 7996 downloads
- Max aft deflection of the control column corresponds to max aft PC joystick deflection.
- Max forward deflection of the control column corresponds to the maximum forward PC joystick position.
- Neutral control column position corresponds to neutral PC joystick handle position.
This means that dependency between the control column position in the cockpit and PC joystick handle position is no longer linear, with a discontinuity (bend) at the neutral point.
4.jpg 63.82KB 7958 downloads
- Max aft deflection of the control column corresponds to max aft PC joystick deflection.
- Neutral control column position corresponds to the neutral PC joystick position.
- There is direct linear dependency between the control column position in the cockpit and the PC joystick position.
This means that Max forward control column position corresponds to a fraction of the full forward deflection of the PC joystick handle, and there is an unused sector of the PC joystick travel when commanding large forward deflections (deadzone at near the travel limits).
5.jpg 76.4KB 7963 downloads
Each of the above solutions has its own pros and cons. Since we have to pick "the least evil" solution out of those available, I will only go over the disadvantages of each.
Disadvantages for Solution A:
In this case, a PC joystick in the neutral position will correspond to some value aft of the deflection of the control column in the cockpit (for example, maximum elevator deflection angles for Fokker D.VII are 30 deg. up and 18 deg. down - so the neutral PC joystick position will correspond to 6 degrees up elevator deflection). This is different from Solutions B and C, where the neutral PC joystick position corresponds to the neutral control column position in the cockpit of the real aircraft (and therefore elevator position), and in accordance with fact 2.4, the solution A will cause airplane to settle down to the higher value of AoA compared to solutions B and C. And in the light of fact 2.6, the airplane will be trimmed to a slower airspeed than that of solutions B and C. In light of fact 2.7, with solution A the virtual airplane will require more forward PC joystick deflection for level compared to solutions B and C.
Disadvantages of solutions B and C:
In the process of developing an airplane a very important characteristic is being carefully observed: the amount of control column deflection that provides 1 degree increase in Angle of Attack (or achieving 1 G load at a given airspeed), as well as the incremental amount of stick force(in pounds or kilograms) to achieve that difference. This control characteristic should meet certain established standards, to ensure controllability and comfort when flying the airplane.
If, for example, the control column forces needed were excessive, flying such an airplane would be very difficult, pilot fatigue would develop quickly, perhaps resulting in complete failure of the pilot to control the aircraft.
If the control column forces were too light, a pilot will not be able to develop fine and precise feel of the control column (because the pilot mostly "feels" the plane through control column forces, not through control column deflection angles). In such a case, the pilot's input could cause divergent oscillations in angle of attack, or in g loads - which is dangerous.
By the way, this is exactly why, (and remember fact 1.4 as well), pilots commonly perceive that PC flight simulator controls are "excessively sensitive", as compared to a real airplane. That's why my advise is to always try to pick a joystick with stiffer springs, or strong FFB mechanism, it is not only closer to the real life, but also will allow you more precise control of an airplane in a simulator.
As with stick forces, the situation is similar with control column deflection. If, in order to achieve desired change in AoA, the pilot has to deflect the control column by a large value, then the control of such a plane will be uncomfortable and sluggish - the control column travel required between aft and forward limits will end up being too large. But if that full travel is too short - the pilot will have to control the plane with minute stick movements, which will reduce the precision of control and will cause errors and oscillations (overshoots and corrections). Remembering fact 1.2, this is exactly what we observe in flight simulators.
Unfortunately, "stretching" the curve of joystick deflection vs. increase in AoA, trying to bring it closer to real control column curve is not appropriate since in this case the full deflection of joystick handle will only cover the narrow deflection sector of the real control column - and we would then be depriving the virtual pilot of the ability to command maximum G loads and perform maneuvers which the real airplane is capable of.
Hopefully it is very clear that a virtual pilot, unlike a real one, in accordance with facts 1.2 and 1.4, has a much less precise device to control the virtual airplane - a short and light joystick. This is one of the main reasons (along with g-force feel, binocular vision and other important factors) for frequently observed problems with the precise control of virtual airplanes in computer games. This is about the same as trying to drive a car with a steering wheel, the size of a coin.
This is exactly why, to achieve acceptable precision of control of a virtual airplane in the game, it is important to utilize full travel (from stop to stop) of an already short joystick, and to have the slope of Angle of Attack vs joystick handle deflection curve be as gentle as possible.
What consequences do we find in the case of Solution B?
The obvious consequence is a "discontinuity" of the curve of AoA vs. joystick deflection (in accordance with fact 2.4). The relation curve is less steep for forward stick deflections, and more steep for aft deflections (positive AoA zone). Consequently, in the range of AoA and G-loads where the most of the flying (and aiming in dogfight) is done, the airplane controls become more "touchy" or "too sensitive" as compared to Solution A. As a result, precise flying of the airplane is more difficult, and the virtual pilot will be more likely to encounter divergent oscillations. Also, a discrepancy in airplane reactions to equal deflections of a joystick will manifest, responses are "quick" when moving the PC joystick aft, and "sluggish" when pushing it forward. If you make a recording of joystick deflection vs time during a typical combat sortie it will become apparent that in case of Solution B (as opposed to A) a smaller range of joystick deflection angles was utilized (predominantly aft), and forward deflections were practically unused.
In the case of Solution C, it carries all disadvantages of Solution B, with the added problem of excessively quick control responses at negative AoA (forward stick deflections), and there is a totally unused zone of forward joystick deflections towards the limits of the PC joystick travel. This further adds to the problems that a PC joystick has already, namely the limited travel of the joystick (fact 1.2), and this solution would further reduce the useful travel of the stick.
The only advantage of Solutions B and C vs. solution A is that it is not necessary to deflect the joystick forward in a level flight at high speed, and also that when the joystick is let go, the virtual airplane's elevators will settle to the neutral position more close to the position of the real airplane elevators. However, considering facts 1.3 and 2.3, the elevators deflection with "free" control column will practically never be the same as elevators of a real plane (with control column "free"), if we are talking about a regular spring-loaded joystick. Such is the flaw of a typical joystick, which, unfortunately, was never designed to imitate the finer points of control stick forces of a real airplane control column. Exceptions to this are the force feedback joysticks (FFB), which allow developers to control the zero-force deflection angle, in accordance with control column zero-force deflection angles in the real airplane cockpit, depending on flight conditions. We do that in RoF. And the lucky owners of FFB joysticks can check it out for themselves. However, even that does not free them from problems with the zero-force stick position and the tendency of airplanes to pitch up, which the real WWI pilots had had to contend with as well. While real Fokker D.VII pilots would tie their control column to the instrument panel with a rubber band, so as to reduce the requirement of having to apply forward pressure on the stick all the time, you can try to do the same using your joystick and your monitor.
For the owners of regular spring loaded joysticks, due to their inherent design limitation, (central spring loading and equal travel forward and aft), it makes no sense at all to talk about solutions B and C to "fix" the supposedly incorrect zero-force deflection - it will always be incorrect, be it solution A, B or C. What's left is only complaints about less "comfortable" level flight and increased airplane tendency to pitch up.
Having analyzed all these issues, I made a decision to go with Solution A.
Below you can see the graph "AoA vs Joystick angle", and also graphs of joystick angle required for level flight at different airspeeds, for all 3 solutions, for Fokker D.VII in RoF.
6.jpg 64.97KB 7994 downloads
Nonetheless, I do understand well that there are disadvantages of Solution A, so in order to increase the gameplay comfort, I have some ideas for the future, and some that have already been implemented.
1) Level Autopilot - will allow you not to hold the joystick forward in a prolonged flight. Clearly this did not exist during WWI, so take it as our aim to help to the players in the game. To engage the Level Autopilot, it is not necessary to be in perfect level flight - small degree of bank or climb/descent is acceptable. Press
2) Easy controls mode (turned on by selecting "Easy piloting" in the game difficulty settings.) In this case, you will have a "virtual flight instructor" helping you to control the airplane, with a fine control stick and pedals inputs, preventing gross pilot mistakes, augmenting the flight stability, establishing horizontal level flight or coordinated turns. His actions can be observed by watching control column and pedals movements in the cockpit in response to your inputs. If you let the joystick free, the airplane will tend to continue straight flight on the current direction, if it is in physical condition to do so (depending on airspeed, engine RPM, damage received etc.).
3) Trim tabs. Strictly speaking, only a few planes of WWI had trim – (only S.E.5a had this of the airplanes included to RoF yet). However we do have plans to make trim available on all airplanes, and enable/disable it through difficulty settings menu. The trimming operation for owners of FFB joysticks will be no different from that of a real airplane - it will allow users to "trim out" stick forces in flight. For spring-loaded joystick owners, trimming will adjust the relation between elevators deflection angle and the joystick neutral (centered) position, i.e. something in-between solutions A and B. Here the partisans of Solution B (I am sure there will be ones) will win back their "lost money" .
By the way, I am planning in the nearest future to have a poll with voting - which of the above solutions A, B or C the majority of users will want to have as a default game setting. Possibly, my choice of solution A will not turn out to be the favorite one, but now, after my explanation, at least, you will understand what are you voting for.
4) Joystick curves settings in the game. Don't ask me "when?" This work is in our plans for the future, but (like trim) not in the most immediate ones. However, I will have to disillusion right now those who believe that joystick curve settings will help to cure the problems discussed here. My deep conviction is that a "dead zone" and non-linear joystick curve will just make pitch control of an airplane worse. I came to this conclusion while working on Advanced Flight Model for Su-25 of "Lock On: Flaming Cliffs". The problem is that the "dead zone" (which can be useful for roll or yaw control, if joystick has a centering problem or noisy pots/sensors) in the pitch channel just creates a disconnect between angle of Attack (AoA) and joystick deflection angle (remember fact 2.4?), and does it for a particular value of AoA at that. And as a consequence, for totally different values of G load (fact 2.5). This has nothing to do with the desire of a virtual pilot to have a level flight when joystick is let free. Simply, when controlling aircraft G loads, you will find an uncomfortable "step" in a most unexpected place (which depends on airspeed), where the airplane will stop responding with G load change to joystick deflection. This might happen during a turn, for example. Also there will be a disconnect between the required joystick deflection and achieving level flight at different airspeeds (fact 2.7).
7.jpg 65.4KB 7906 downloads
Non-linear joystick curves (response) also distort the relationship between Angle of Attack and joystick deflection in such a way that at moderate angles of attack the response is "soft" or "slow", but when commanding high or low AoA it becomes "abrupt" or "fast". But the moderate AoA values correspond to a particular airspeed in a level flight (for example for Fokker D.VII it is about 100km/h), and at those speeds it would be easier to maintain level flight indeed. But at speeds in excess of 150km/h the precision of control with the joystick will, to the contrary, decrease, as the plane will be "too touchy" or too sensitive to controls. And this is on top of the fact that at higher airspeeds the airplane is already more sensitive to control input even without any non-linear curves. It looks like we are improving in one area and making it worse in another. And this in no way corresponds to virtual pilot desire to correct "abrupt" reaction of the plane to the joystick inputs. All this is shown on the following figures, using Fokker D.VII as an example.
8.jpg 70.91KB 7898 downloads
The problem of emulation of real aircraft controls in a computer game with a joystick is not new. The more exact the physics model in a flight simulator, the more the fine details and discrepancies show themselves. These details, for the most part, are known and understood by real pilots (not always though), but will surprise inexperienced virtual pilots. Games like Retaliator or F-19 had none of those problems at all, don't you remember? Unfortunately, ordinary spring-loaded joysticks are not adequate for this task. Force Feedback joysticks (FFB) do have an advantage here, but not without shortcomings, and not every virtual pilot has them. That's why, for the time being, we have one choice: the best out of all choices possible. I made my choice while working on Advanced Flight Model of Su-25 for "Lock On: Flaming Cliffs". This is Solution A. To help RoF users, the autopilot was made available, as well as "easy" control mode. The future measures which I presented here being trim and joystick response curves, albeit I voiced my opinion on the problems with curves (and again, it is difficult to talk about the time frame or promise anything). Also I do plan to run a poll with the question "do you prefer Solution B (or C) over Solution A". It is possible that based on the results, the preferred solution will be reconsidered. And finally, I will be glad to hear your ideas - what would be the best way given the circumstances, and what else can be thought of to make airplane control in RoF realistic and comfortable at the same time. Feel free to post your ideas in this thread. I will read them and comment when I have a free moment. If I don’t respond please don’t think I am ignoring anyone. I will certainly read everyone’s ideas and post when I can. ROF coding never ends my friends.
P.S. Great thanks to TX-broWright for translation this post to English!
Neoqb Lead Engineer.