One of the main objectives of a research organization such as the NACA is to come up with new ideas that contribute in some way to progress in aeronautics. It is surprising, however, how few engineers working at such a center ever conceive a new idea in their entire careers. Scientists noted for advancements in their technical fields have stated (and I believe it to be true) that to produce a worthwhile new idea, the researcher must first have a complete knowledge of all advancements that have been made previously. In addition, knowledge of related fields is often helpful because the ability to recognize interrelations between disciplines will reveal new concepts. A second important factor, based on my own experience, is to be involved in experimental work. Such work often requires new techniques, and the old adage "necessity is the mother of invention" certainly applies in this case. In addition, experimental work often produces unexpected results that stimulate thinking on the causes of these results.
Of course, there are some engineers who suggest new ideas that are not worthwhile advancements. They may be ideas that have already been tried or are contrary to accepted physical principles. Such ideas may result from a lack of sufficient background in the field The possibility of introducing bad ideas is often the downfall of brainstorming sessions in which all members of a group are invited to suggest new ideas in hope of coming up with something useful. Poor ideas are worse than useless, particularly if the proposer is a person with enthusiasm and persistence who is able to convince supervisors that his idea should be developed, often at considerable cost in man-hours and facilities.
In my position, I was fortunate in having available the tools for experimental research that required development of new ideas. Some of these ideas were of limited scope and were applicable to the problem at hand or soon became obsolete with the rapid progress of aeronautics. Others had broader application. This chapter presents some examples of new ideas that I developed in my research work.
The First Piloted Simulator Used for Research at Langley
One subject investigated in flying qualities tests was the ability of the pilot to control the short-period yawing oscillations of the airplane. This mode of motion, called the Dutch roll, was usually poorly damped, and with the advent of swept-wing fighter airplanes flying at high altitude, the damping tended to become still less. On actual airplanes, however, it was at that time difficult to vary the....
....damping to investigate the limiting conditions under which the pilot's control would become marginal. To study a wider range of conditions, a simulator called the yaw chair was built (figure 8.1). This device consisted of a chair mounted in a frame free to pivot on a vertical axis. The frame was restrained by springs and was controlled by other springs operated by the rudder pedals. A hydraulic servomechanism was also installed that could apply moments to the chair in phase with its angular velocity, thereby allowing the damping to be varied from stable to unstable. The availability of facilities in the flight research hangar to build devices of angle iron and welded steel tubing was very useful in constructing a simple device of this kind.
The studies made with the yaw chair showed that the pilot could not damp out the oscillations if the period was much less than one second. Effects of the degree of damping or instability were also studied (ref. 8.1). One interesting result was that the pilot could damp out the oscillations just as well with his eyes closed as with them open. This result showed the importance of so-called kinesthetic cues, which are obtained from the sensory mechanisms in the inner ear, in the control of airplanes by human pilots.
Though the yaw chair was a simple mechanical device capable of studying only one type of control problem, it foreshadowed the development of much more complex general-purpose simulators that became possible with the development of computer and display technology. Operation of such simulators later became a major research activity at the Langley Research Center. Before such developments, however, a simulator called the NAP chair (for normal acceleration and pitch) was built that provided the pilot with a simulation of the short-period vertical motion of an airplane applicable to tasks such as formation flight and in-flight refueling. In the 1950's, a simulator was developed  to impart three-axis rotary motion to a cockpit. This simulator had the objective of studying pilots' ability to control lateral oscillations involving large rolling motions. At this time, however, the space age started and the gimbal system ended up in applications to study atmospheric entry or docking. The rotary motion simulator was never used for its intended purpose.
Though these simulators had limited application, they had the advantage of providing very precise reproduction of the motion of the actual vehicle together with visual cues that came directly from the real visual scene. Later simulators frequently used more complex mechanisms to produce the motion of a very heavy cockpit, but because of the mass of the moving system and friction in the drive mechanism they were unable to reproduce the actual motion precisely. Likewise, visual scenes that were produced by television or computer techniques frequently lacked detail of the actual scene or involved lags inconsistent with the motion of the vehicle. Such simulators were often found to be unsuitable for research on the pilot's response characteristics and his interaction with the vehicle dynamics. Simulator engineers were slow to recognize these deficiencies, and when they did, much of the research on the simulators was devoted to improving the simulators rather than on solving real airplane control problems. In recent years, the importance of poorly damped high-frequency airplane motions has decreased because of the widespread use of automatic stability augmentation devices to avoid such characteristics. As a result, many simulator studies deal with problems such as development of instrument displays and training of pilots in specific vehicles. Some modern simulators are used for development of flight control systems. For example, the ability of the human pilot to track a moving target with various types of control systems has been studied. In such applications, consideration must always be given to the sensitivity and speed of response of the human pilot's sensory system as compared to the sensitivity and dynamic characteristics of control force and visual and motion cues of the simulator. A human pilot can track a moving target with an accuracy of better than 0.2 degree. Obviously this problem cannot be studied with a simulator that projects the target with an accuracy of 0.5 degree. In fact, the pointing accuracy of the simulator projection system should be better than 0.02 degree, or one-tenth the pilot's tracking error.
Roll coupling means the effect of rapid rolling of an airplane on the motion in other directions, that is, about the pitch and yaw axes. At the time I went to college, such coupling was not a subject discussed in the stability and control courses. In general, airplanes flying up to that time had not experienced problems from this effect in normal flight. It was recognized that in stalls or spins, the interaction of motions about the various axes had to be considered, but analyses of these phenomena had rarely been attempted because of the complex and unknown aerodynamic effects caused by the stalled flow on the airplane.
I would not have tried to analyze the effect of rolling if it had not been for some experimental data that required explanation. During WW II, as mentioned previously, airplanes had started to encounter serious compressibility effects in dives at transonic speeds. Wind tunnels were unable to investigate these problems because of a phenomenon called choking. In a closed-throat tunnel, as the airspeed is raised to the point at which shock waves start to form on the model, a small additional increase in speed will result in the shock waves spreading across the entire test section. Further attempts to increase the Mach number in the test section by applying more power to the tunnel simply result in greater pressure drop across the shock waves with no increase in Mach number.
 Gilruth, then head of the Stability and Control Branch, was well aware of the compressibility problems because of the tests being made on high-speed fighter airplanes. He considered how the airplanes being flown in the Flight Research Division might be used to obtain data at transonic speeds. He proposed two new test methods, one called the wing-flow method and the other the free-fall method. The wing-flow method will be discussed subsequently. F. J. Bailey, head of the Performance Branch, was also responsible for many of the ideas required to implement the free-fall method. The free-fall method consists of dropping heavily weighted test models from an altitude of about 30,000 feet. These models would reach Mach numbers of 1.0 to 1.2 before impacting the ground. New developments in telemetering and radar during the war made it possible to obtain useful aerodynamic data from these models during the course of the drop.
The free-fall method was rapidly put into operation, using first a B-29 bomber and later an F-82 fighter, to carry the test models. Actually this was a joint project with the British, in which the British were to develop recoverable models that could be reused to obtain more data, whereas the Americans under Gilruth were to develop expendable models. After about two years of operation, Gilruth's group had successfully tested about 20 models, whereas the British had still to make a successful recovery of a model. The validity of Gilruth's direct approach was thus demonstrated.
Most of the free-fall testing was under the direction of Charles W. Mathews and Jim Rogers Thompson of the Flight Research Division. The models tested were symmetrical bomb-shaped streamline bodies with cruciform tail fins and, in some cases, with cruciform wings. These models fell in a zero-lift trajectory, with measurements confined to drag of various components.
About that time the Bell XS-1 (later called the X-1) rocket airplane was being developed to attempt to reach supersonic speeds.
A question arose as to the effectiveness of the elevator control in the transonic speed range. To study this problem, as well as to investigate the ability of the free-fall method to obtain data in lifting conditions, it was decided to test a model of the XS-1 airplane. A photograph of the model is shown in figure 8.2. The model had a wing span of 7 feet and weighed 1350 pounds. To maintain the desired lifting condition in flight, the model was equipped with a very simple type of autopilot. An accelerometer sensed when the normal acceleration (that is, the acceleration normal to the plane of the wings and fuselage) went outside the range 0.4 to 0.8 g. If the acceleration went outside these limits, the elevator was moved at a slow rate to bring the acceleration back into this range. The model was intended to roll slowly during the drop to maintain a predictable trajectory.
When I fly model airplanes, I customarily look at the model from the rear to check whether the wing is twisted. As a matter of habit, I did this on the XS-1 model. Though the wing was supposed to be machined to an accuracy of one thousandth of an inch, I could readily see that the wing was twisted about one-eighth of an inch at the trailing edge of each tip. Calculations showed that this amount of twist would have caused the model to roll at a rate of about 250 degrees per second during the high-speed portion of the dive. This rate was considered excessive (though we had no way to know, at that time, what rate was really excessive). Therefore two small wedges, like ailerons, were placed on the trailing edge to oppose the built-in twist.
The model was dropped successfully. The exact date of the drop is not known, but a memorandum by C. W. Mathews for H. A. Soulé that gave the results of the drop is dated September 3, 1947. Data from the drop are shown in figure 8.3. The longitudinal control worked as intended. At low Mach numbers, in low-density air at high altitude, the full-up deflection of the elevator of 10 degrees is not enough to maintain 0.4 g.
 As the speed increases, the model is held in the desired limits of acceleration except for a period of rather wild oscillations at a Mach number around 0.75. At the highest Mach number, full-down deflection of the elevator of 4 degrees is not enough to hold 0.8g, and the acceleration builds up to about 2g.
The model was not equipped with any instrument to measure rolling velocity, but when it was dropped, it was observed in the optical tracker to be rolling by the flashing of light from the wings. Near the end of the flight, however, the model stopped rolling and performed a pull out. It flew far from the designated drop area on a former bombing range called Plum Tree Island and crashed about a mile away in the swamps around the town of Poquoson.
Of course, dropping a heavy bomb in a populated area is not considered very good practice, but evidently no one was aware of the incident. As a result, nothing was ever said about it, and the model, to this day, is probably buried 15-feet deep in the mud somewhere around Poquoson.
In examining the records further, the oscillation mentioned previously was found to represent a violent pitching in angle of attack between the positive and negative stall. The motion below the stall was actually a divergence, but when the stall angle was reached the resulting large increase in restoring moment pushed the model back in the other direction. At this point in the trajectory, the model was well below the Mach number for any serious transonic effects. Knowing the heavy weight of the model and its observed rolling motion, I was led to analyze the problem as some kind of gyroscopic effect. In making the analysis, I used a set of equations called the Euler dynamical equations, which had been taught at MIT by Professor Rauscher as well as in the graduate course Introduction to Theoretical Physics. These equations are nonlinear, but because the methods of solution that I had learned for the airplane stability equations dealt with linear equations, I attempted to linearize the equations by considering small displacements in pitch and yaw and by assuming the rolling velocity to be constant. I had considerable difficulty in deciding on the signs of the terms involving the coupling effect of the rolling velocity. Finally, I worked out the equations both ways and arrived at the correct sign by considering the simplified limiting case of a rotating rod, the solution for which I could visualize. The resulting equations were also expressed in nondimensional form. Restoring moments were expressed in terms of the natural frequencies in pitch and yaw of the nonrolling airplane, a technique taught for oscillating systems by Professor Draper. In addition, the frequencies were nondimensionalized in terms of the rolling frequency.
When the characteristics of the XS-1 model were substituted in the equations, the results clearly showed the possibility of a divergent motion even at relatively low values of rolling velocity. The instability was likely to occur when the values of longitudinal stability and directional stability were markedly different and when a large amount of the weight was distributed along the fuselage. In the case of the XS-1 model, the directional stability was high, but the longitudinal stability was low, particularly at high subsonic speeds.
This analysis was published in a NACA Technical Note entitled Effect of Steady Rolling on Longitudinal and Directional Stability (ref. 8.2). This type of instability had never been encountered on full-scale airplanes, and indeed, it would not occur on the full-scale XS-I. The report predicted, however, that some of the new fighter airplanes then on the drawing board could encounter this problem. In the case of the fighter airplanes, the conditions favorable to instability involved long fuselages, short wings, and high values of longitudinal stability together with low values of directional stability, both of which occur at low supersonic speeds. Later, the report received quite a lot of attention when the problem was actually encountered on airplanes like the X-3 and the F-100. The....
...existence of my theoretical analysis showed the design changes required to correct the problem and may have saved some pilot's lives by allowing quick action to correct the designs.
Later, the phenomenon was called "roll coupling" or "inertial coupling," and many technical reports were written presenting more detailed analysis of this problem. The more complete analyses took into account the nonlinear nature of the equations and required solutions on analog or digital computers. As a result, the solutions referred to individual cases. I was not sufficiently familiar with these techniques at the time, and the necessary computing equipment was not then available at Langley. My approach was to use the methods that I had learned for the classical airplane stability equations, which were linear. This method revealed the basic physical principles that caused the instability, a result that could not have been obtained by study of more exact solutions for individual cases. I could see that my uncertainty about the sign of the coupling terms could have been avoided if I had started with the complete equations and then linearized them. When others had been able to analyze the complete equations, I felt that my knowledge of this subject was somewhat deficient and should have been treated more thoroughly in the courses at MIT. I realize now, however, that the simpler approach gave a more basic understanding of the problem.
The XS-1 drop test showed that the elevator control was adequate, but in addition revealed a new type of instability that proved important in the design of future airplanes. This experience shows that the important research results of a test are often quite different from those intended when the tests are planned.
I have received more recognition from the report on roll coupling than from any other work done in my career at Langley. I remember giving a talk on the results at one of the Langley Department Meetings, seminars that were held after working hours to allow different divisions to present their latest research results. I used some crude devices constructed of dowel rods coupled together with either a piece of wire or a flat sheet of celluloid to illustrate the importance of the difference in stability in the two planes. These devices are sketched in figure 8.4. When the rods were spun between the hands, the rod connected with wire would remain stable, whereas the rod connected with the celluloid would fly out to a large deflection. The talk received a round of applause, something quite unusual in these meetings.
This work is also the only research that I conducted that received recognition in the American Institute of Aeronautics and Astronautics magazine, Astronautics and Aeronautics, now called Aerospace America, which contains a monthly column entitled "Out of the Past," presenting historical items occurring 25 and 50 years previously. Twenty-five years after my report appeared, the magazine in June 1973 presented the following item:
 NASA research scientist William H. Phillips of the Langley Memorial Aeronautics Laboratory Flight Research Division publishes Technical Note TN- 1627 which contains a theoretical prediction of the problem of inertial coupling. The phenomenon later plagues aircraft having long fuselages and short wings, where the mass of the load is spread along the fuselage with little spanwise distribution of load. In particular, it forces a slowdown in the operational introduction of the North American F-1OOA Super Sabre, and first manifests itself in flights of the Douglas X-3 research plane. Increasing the area of the vertical tail and the span of the wings solves it.
NACA Research Memorandum H55A13 by the NACA High-Speed Flight Station, 4 Feb. 1955: NASA, Aeronautics & Astronautics 1915196O, pp. 60-61.
A Method for Determining the Moments and Products of Inertia of Full-Scale Airplanes
To compare flight measurements of airplane motion with theoretical predictions, it is necessary to know the weight, center-of-gravity location, and the moments and products of inertia. In the case of lateral motion, the effect of product of inertia is more frequently expressed in terms of the tilt of the principal longitudinal axis of inertia from the reference axis. At the present time, airplanes, at least those built by large airplane companies, have known weight and inertia characteristics because the companies have a weight control section that keeps up-to-date computer records of the weight and exact location of every item in the airplane. During WW II, however, the weights were less closely controlled, and in many cases the final drawings were not available until long after the airplane was in production. Even today, in airplanes that are fitted with special research equipment or airplanes produced by small manufacturers, weight and inertia characteristics may not be accurately known. For accurate correlation of flight data with theory, therefore, measurements of these characteristics are often required.
Determining weight and center-of-gravity location might appear to be a simple physical measurement if suitable scales are available. Scales for weighing airplanes had been built into the hangar floor, but in years after about 195O, accurate strain gauge load cells were usually more convenient to use. Despite the basic nature of the measurement, new engineers, when assigned this task, often made errors. As a result, a single engineer, William C. Gracey, was assigned the job of overseeing the weighing of each new airplane as it was received. Gracey had also written a report on measurement of the moments of inertia of airplanes by an oscillation technique. This method was occasionally used, but to get moment of inertia in pitch, it was usually more convenient to mount the airplane on jacks and oscillate it while it was restrained by springs of known stiffness attached under the tail.
These methods still failed to supply data on the tilt of the principal longitudinal axis of inertia. A report by Sternfield had shown that the tilt of this axis was of critical importance in determining the damping of the lateral oscillations (ref. 8.3). At the Wright Air Development Center at Wright-Patterson Field, a proposal was once under consideration to build a very large facility that was capable of holding large airplanes and oscillating them to determine their inertia characteristics. So far as I know, this facility was never built. I devised a technique, however, that was at least capable of making accurate measurements of the moment of inertia in yaw and the tilt of the principal axis of airplanes as large as fighters with very inexpensive equipment (ref. 8.4).
In this method, the airplane is suspended by a single cable from a crane located in a hangar. Most fighter airplanes are equipped with a hoisting sling to allow them to be moved by cranes or helicopters, so this equipment is already available. The airplane is restrained in yaw by a pair of springs attached to brackets ahead of and behind the suspension cable. From the period of oscillation in yaw, the moment of inertia in yaw can be determined. A photograph of a fighter airplane set up for these measurements is given in figure 8.5.
The pitch angle of a line through the attachment points of the springs can be varied. When the airplane is oscillated in yaw under restraint of the springs, it usually has a rolling oscillation, the amplitude of which can be measured by a position measuring device at the wing tip. For some particular angle of the spring attachment, however, the roll oscillation goes to zero. This point can be found by interpolating between two of the spring settings. The tilt of the principal axis is not equal to this angle at which the roll is zero, but it is related to it by a simple formula
tilt of principal axis from X body reference axis
spring angle for zero roll
lZp, lXr moments of inertia about Z and X body reference axes
I have not seen this formula given in any reference book, though I am sure someone must have worked it out previously. So many authors were involved in preparing the brief report (ref. 8.4) describing this project that I did not include my name on it, but I always thought that originating this technique and deriving this simple formula were worthwhile contributions.
One piece of apparatus required for most flight tests was the trailing bomb or towed  airspeed head. This device was a heavily weighted body containing a static pressure orifice that measured the static pressure sufficiently far from the airplane to avoid errors due to the flow field of the airplane. The pressure was transmitted to the airplane through a tube attached to the towline. This device was used to calibrate the airspeed installation in the airplane, which usually was in error by several miles per hour because of the influence of the flow about the airplane on the pilot-static head mounted on the airplane. These towed airspeed heads had a frequent tendency to become unstable or to break loose and fall to the ground where they could conceivably cause damage or injury. I made a study of the stability of these airspeed heads with the same methods that I had been taught at MIT to study the stability of airplanes. I also studied the previous literature on the subject. A previous study by Herman Glauert, the famous British aerodynamicist, had covered the longitudinal oscillations. I treated the lateral oscillations. My analysis showed that the least stable region was with the body close to the airplane, whereas Glauert showed the instability with the body far from the airplane. I discovered an error in Glauert's analysis in which he had reversed the stable and unstable regions. In view of his reputation, however, I did not mention this discrepancy in my report, which was published in 1944 (ref. 8.5). The Navy was also interested in towed bodies for carrying magnetometers to detect submarines. As a result of their interest, I made tests of a large towed body with a movie camera in its nose to study its motion. Through the course of my career, I always kept an interest in towed bodies and kept a bibliography of reports published on this subject.
Despite my analysis, towed bodies continued to break loose and get lost, and I finally realized that the problem was instability of waves transmitted down the cable rather than instability of the body itself. Studying this type of instability got into the realm of partial differential equations, a subject with which I never had much experience. The problem of transmission of waves down a string or cable, however, is a classical problem in acoustics that was treated in the book, Theory of Sound by Lord Rayleigh (ref. 8.6). I studied Lord Rayleigh's analysis that showed how to divide the problem into ordinary differential equations dealing with the time and position dependence of elements of the cable. My analysis showed that oscillations originating from unsteady flow in the region of attachment of the cable to the airplane would be amplified exponentially as they traveled down the cable if the airspeed was greater than the speed of wave propagation. The speed of wave propagation, in turn, varied as the square root of the tension in the cable. With a weighted body on the end of the cable, the tension did not increase very much with airspeed, and as a result, beyond some value of airspeed, the oscillations became unstable. I wrote a second report, which was published in 1949, treating the oscillations of a towed cable (ref. 8.7). After WW II, when much of the German wartime research was made available, I discovered that an almost identical report had been written in Germany (ref. 8.8). These reports revealed the true cause of the loss of the trailing bombs. As a result, another method of static pressure measurement, called the trailing cone method, was developed and has since been widely used. The drag of a cone on the end of the cable increases as the square of the airspeed, and therefore keeps the cable stable as the airspeed is increased. The static holes used to measure the air pressure are placed in the tube attached to the towed cable, sufficiently far ahead of the cone to avoid interference from this source.
I never received much recognition for my knowledge of towed bodies because so few people ever work in this area. Some discussions were held with companies concerned with in-flight refueling because instability of the refueling hose was a matter of concern. I was also selected as an anonymous reviewer of an AIAA paper dealing with the stability of towed cables.