WORLD WAR II FIGHTER AERODYNAMICS BY DAVID LEDNICER EAA 135815 Previously, we have explored the aerodynamics of modern homebuilt aircraft. Here, we will instead look at a different class of aircraft - World War IT Sighters. As time progresses, many of the valuable lessons learned in the original design of vintage aircraft are being lost. It is the purpose of th study to use modern aerody- namic analysis tools to recover some of this lost knowledge. reat strides in air- craft design were made in the era of 1935-1945, and this iy most evident in the design of fighter aircraft of this period. For this reason, an evalu- ation of three prominent fighter aircraft of this era, the North American P-51 Mustang, the Su- permarine Spitfire and the Focke Wulf Fw 190 is presented here. As so much misinformation has appeared on these aircrafi, refer ences will be cited to support the data discussed here, Wing Geometry Ina sense, these three aircraft types represent three stages with a single generation of fighter de- velopment. This ean be most easily seen in the wing airfoils used on the aircraft. The Spitfire, designed in the mid 1930s, used the NACA 2200 series of airfoils, which was new at the time, The wing root air- foil is aNACA 2213, transitioning toa NACA 2209.4 at the tip rib. The Fw 190, which was designed at the end of the 1930s, used the NACA 23000 series of airfoils. The wing root airfoil isa NACA 23015.3 and the tip airfoil a NACA 23009. The P-51’s wing, designed in the early 1940s, uses an early laminar flow airfoil which is a NACA/NAA hybrid called the 45- 100. The wing root airfoil (of the basic trapezoidal wing, excluding the inboard leading edge ext sion) is 16% thick, while the airfoil at the tip rib is 11.4% thick. With the inboard leading edge exten- sion, the wing root airfoil on the P-51B is 15.2% thick and on the P-51D 13.8% thick, The later model P-51H used a NACA 66,2 (1.8)15.5 a=.6 at the wing root and a NACA 66,2-(1.8)12 a=.6 at the tip and has no inboard leading edge extension. It is interesting to note that ap- proximately 2 degrees of washout was used on all three aircraft. However, the distribution of twist SPORT AVIATION 85Incidence (degrees) Figure 1~ Wing twist distr et «|| Utlons for the P 00 oz os os ef 08 @7 Ge ob 10 | S1B,P-51D, ‘Semispan Fraction Spire and Fw varied for each aircraft. The Spitfire wing has a constant incidence (2 de- grees) to the dihedral break, where the twist starts. This aireraft actually has 2.25 degrees of washout, distrib- uted linearly (Fig. 1). The Fw 190 wing is unusual in that 2 degrees of washout exists between the root and a point at 81.5% semispan, Outboard of this location there is no more washout, the incidence holding fixed at zero degrees. The basic trapezoidal wing of the P-51B and P-51D has 2 degrees of washout, with the tip rib at -.85 degrees of incidence. However, addition of the drooped inboard lead- ing edge extension modifies the appearance of the twist distribution. Lift distributions for the three aireraft show the results of these twist distrib- utions (Fig. 2). These lift distributions were calculated, using VSAERO, with the aircraft trimmed at 360 kts and 15,000 feet of altitude to repre- sentative Gross Weights and CG locations. The Spitfire wing is famous for having an elliptic planform. Indeed, the chord distribution is elliptical. An examination of the resulting eircula- tion distribution for a trimmed condition mentioned above, shows that the loading distribution is not el- liptical, though it is probably the most ‘optimum of the three aireraft from the induced drag standpoint. The reason for deviation from elliptical is the 2 degrees of washout that have been added to the elliptical planform, which shifts the loading inboard, The elliptical wing planform appears to have been chosen primarily to pro- vide greater wing depth in the inboard portion of the wing, while keeping the airfoil thickness-to-chord ratios 86 JANUARY 1998 low. This depth was necessary to house the outward retracting landing gear and wing gun ammunition boxes. P-51 Mustang Analysis ‘The original North American Avia tion drawing set for the Mustang are available from the National Air and Space Museum. A friend of mine liv- ing in England, Arthur Bentley, had obtained the set and was kind enough to sort through it for the drawings that were of relevance to my endeavor. It 'was found that models of the P-51B/C and P-51D/K were relatively easy to prepare, as the North American AV tion drawings contained surface coordinates, in a familiar Fuselage Station/Buttline/Waterline system. However, the NASM drawing set did not appear to contain the wing defini- tion. After quite a bit of searching, I ‘was put in touch with the Ed Horkey, who had been the Chief Aerodynami- cist on the P-SI at North American, Ed was kind enough to supply the wing definition drawings upper surface extends fairly far back on the wingis chord. This indicates that the wing should be capable of supporting a fairly large amount of laminar flow. The P-51 Mustang is renowned for being one the first air- craft to make use of airfoils designed to be capable of having extensive runs of laminar flow. Both the Spitfire and Fw 190 use airfoils that do not support substantial amounts of laminar flow. ‘A two-dimensional cut through the ‘wing pressure and skin friction distri~ butions calculated by VSAERO on the Mustang (Fig. 5) show that, at a representative cruise condition, the ‘wing was capable of sustaining long laminar boundary layer runs, with transition occurring at roughly 47% of chord. However, this calculation is for an ideal case, for a wing without fas- teners, gaps, misalignments or surface waviness. During World War II, a Mustang was flight tested by NACA with a wake rake behind the wing at roughly 66.7% semispan (Ref. 1). The results of this test show that, in ser- ‘vice the airoraft was unlikely to have a substantial laminar flow on the wing and transition occurred in the first 15% of the chord, Testing in an as- manufactured condition showed slightly lower drag and further, when the wing was refined fo remove wav ness and surface imperfections, a drag level was measured indicative of 2 substantial region of laminar flow Wartime windtunnel tests of the Mus- tang’s wing airfoil in Germany gave similar results (Ref. 2). Early models of the P-51 experi- enced boundary layer separation in the radiator inlet duct. Pilots reported a rumbling noise emanating from the for both the P-S1B and P-51D, The pressure dis- tributions calculated = by VSAERO on the P-51B and P-51D are shown i and 4. Part noteworthy is the re- ‘gion of strong suction on the P-SID bubble ‘canopy. This region is not present on the less bulged P-51B canopy. On both ait- craft the suction ae region on the wing Figure 2 - Calculated wing loading comparison with the craft trimmed at 360 kt and 15,000 feet altitude to repre ‘sentative gross weights and CG locations. Semispan Fractionductwork behind and beneath the cockpit on early model Mustangs. To investigate this phenomena, a complete Mus- tang fuselage was installed in a wind tunnel at the newly opened NACA Ames Re search Center. It was found that the rumble was the result of the separated flow in the cooling inlet duet striking the radiator (Ref. 3), Changes, both in duet shape and the addition of a deep boundary layer splitter on the inlet eliminated the rumble and improved the aireraft’s cool- ing. The results of these quite common on automo- changes can be seen in the biles and is related to the VSAERO boundary layer calculation, Were present on the drawings, but slope of the windsereen, The Spit. which shows that boundary layer on preparation of the fuselage proved to fire’s windscreen is at a 35-degree the upper surface of the cooling sys- be difficult as a global coordinate sys- angle to the forward deck, while the tem does not separate until far back in © M Was not used. For instance, Fw 190°s is at a 22-degree angle and ihe duct (Fig, 6). The boundary layer bulkheads could only be located by ae- the P-S1’s is ata 31-degree angle. Ev- on the lower surface of the duct, start- cumulating ing fresh behind the oil cooler makes it distances from a to within inches of the water radiator known reference, and intercooler before separating. The it. @ system more losses in this system are much lower kin to that used in than that of the Spitfire. This efficient — the design of ships. cooling system arrangement is credited _ The surface pres with much of the Mustang’s superior Sure distribution the base of the windsereen The computation indicates that the boundary layer sep: rates approximately 6 inches in front of the windscreen, due to the increasing pre sure in this region (Fig. 8). The boundary layer traces that stop at separation have been restarted on the wind- shield at the point where the static pressure is the same as that at separation. Such a sep- aration is not present on either of the other two air- craft reviewed _ here. However, this is a featu Figure 3 - Pressure distribution calculated on the P-51B Mustang, performance over the Spitfire calculated for the The Mustang has long had a repu- Spitfire IX is shown tation for being longitudinally i" Fig. 7. Unlike the unstable at aft CG locations resulting Mustang, the chord from the addition of a long-range Wise extent of suc~ fuel tank added behind the pilotis tion on the win, seat, Results of a wind tunnel test of Upper surface cal 4 P-51B (Ref. 18) place the aircraftis be Seen to be rela- power-off stick-fixed Neutral Point _ tively small, limit- at 39.11% MAC, which agrees quite img the amount of well with the VSAERO results, laminar flow the which places this point at 38.97% Wing can suppor. It MAC. P-5IBs could be flown at CGs is interesting that the as far aft as 31.55% MAC (Ref, 4), gfeatest suction on Stick-fixed to stick-free effects and the entire aircraft ap- | power effects account for roughly Pearson the bulged | 50 |—>__- ee Figure 4 - Pressure distribution calculated on the P-51D Mustang. Wi5% MAC difference, canopy. Other strong suctions appear at Supermarine Spitfire the corners of the Analysis windshield, which was made up of Arthur Bentley also was able to sup- panels of flat armor ply me with the original Supermarine glass and had sharp drawings for the Spitfire, The Spitfire corners. drawing set contained definition for One of the first various models, ranging from the Spit- things to come to fire I to the Seafire 47. It was decided light inthe VSAERO to build the panel model to represent a analysis of the Spit- Spitfire IX, which could be fully de- fire is a region of Figure 5 - Calculated Mustang wing airfoll pressure distribu- fined from the drawings. Coordinates separated flow at Hon and boundary layer transition locations In cruise for ideal SPORT AVIATION 87idently, the Spitfire’s the boundary layer separates shortly windscreen is too steep. _ after entering the duct, resulting in a An experimental wind- large drag penalty (Fig. 9). Experi- screen, rounded and of mentally. it was determined that the shallower slope, was fitted Spitfire cooling system drag, ex- toa Spitfire IX in 1943 pro- pressed as the ratio of equivalent g-drag power to total eng: 12 mph, ata Mach number power, was considerably higher than 79 (Ref. 5). A similar that of other aircraft tested by the windscreen introduced on RAE. This was attributed to “the pres the Seafire XVII, is cred- ence of a boundary layer ahead of the ited with a speed gain of 7 duct tends to precipitate separation mph, at 400 mph (Ref. 6). and makes the ducting problem more duced a speed increase of colin ine Separation Supermarine is often re- difficult” (Ref. 8). Similar problems ! —— | garded as being one of the are present on the early model Messer- Figure 6 Calculated boundary layer separation inthe {itst companies to make schmitt Bf 109, up through the F Mustang cooling system use of the breakthroughs model. A complete redesign of the made by Meredith at RAE cooling system, during development Farnborough in of the Bf 109F, resulted in the use of @ the design of boundary layer bypass duct, which duets for cooling significantly improve systems (Ref recovery at the radiator face (Ref. 9), Infact, the Spit- ‘The Spitfire has long had a reputa- fire’s radiator ducts tion of being longitudinally neutrally ere designed us- stable. Results of wartime flight tests these guide- ofa Spitfire VA by NACA (Ref, 10) confirm that the aircraft was indeed the VSAERO cal- longitudinally neutrally stable at a culation indicates typical CG location. The NACA re the boundary layer port mentions that no chan; onthe lower sur- elevator position was necessary to face of the wing maintain longitudinal trim when © changing airspeed, implying that the cooling system CG was positioned at the location of inlet. Running into the stick-fixed longitudinal Neutral the severe adverse Point. The CG location in this test (increasing) pres- was at 31.3% MAC. VSAERO anal sure gradient ahead sis of the Spitfire places the power-off ofthe radiator, — stick-fixed Neutral Point at 36.66% the pressure Figure 7 - Pressure distribution calculated on the Spitfire IX. 88 JANUARY 1999. Standard esumates of power ‘cts show that the Neutral Point will shift forward 4-5% due to these effects, which accounts for the differ- ence between the VSAERO and flight test results, The NACA testing also found there was a stable gradient of stick force with increasing airspeed This means that the Spitfire was stick- free longitudinally stable. Bobweights in the elevator control circuit helped turn the stick-fixed neutrally stable airplane into an airplane with a small degree of stick-free stability. As the pilot mostly is aware of stick-free sta~ bility and low margins of stability are associated with high maneuverability, this was a satisfactory situation. Focke Wulf Fw 190 Analysis Arthur Bentley was once again the source of my geometrical informa. tion. In this everal years ago he had prepared a set of Fw 190 draw- ings for a modeling magaz working from the original Focke Wulf drawings. Initially, | first modeled a radial engined Fw 190 A-8, but I later modified this model to represent an inline engined Fw 190 D-9, in this case using actual Focke Wulf draw- ings. Despite sparse fuselage cross section information, this model was constructed with relative ease. The pressure distribution calcu- lated on the Fw 190 A-8 and Fw 190 D-9 are shown in Fig. 10 and 11 Here, like on the Spitfire, the chord- wise extent of suction on the wing is limited by the choice of airfoils and ‘not much laminar flow is supported Also, as on the Spitfire, the bulged canopy of the Fw 190 D-9 has a re- gion of strong suction, not present on the Fw 190 A-8 At the time that the Fw 190 first appeared in combat, in 1941, it was superior to the contemporary fighters on nearly every count. When the RAF captured the first flyable Fw 190 in 1942, a thorough evaluation revealed the Achilles Heal to be a harsh stalling characteristic, which limited its maneu- ver margins. Captain Eric Brown states (Ref. 11) The stalling speed of the Fw 1904- 4 in clean configuration was 127 mph (204 km/h) and the stall came sud deny and virtually without warning, the port wing drop: ping so violently that the aircraft al- most inverted itself. In fact, if the Ger- man fighter was pulled into a g stall ina tight turn, it would flick out into the opposite bank and an incipient spin was the in= evitable outcome if the pilot did not have his wits about him. The stall in landing configura- sion quite different, there be- ing intense pre-stall buffeting before the starboard wing dropped compara- tively gently at 102 mph (164 km/h). The results of an USAAF evalua- tion of the Fw 190 (Ref, 12 and 13) report the aircraft to have a gentle stall, However, these reports admit that the Fw 190 stalled abruptly when maneuvering, The reason for this re- ported difference in non-maneuvering stall behavior is unknown, A compar- ison of the local wing lift coefficients, calculated by VSAERO, at stall with the estimated stalling lift coefficients of the airfoils two-dimensionally (Fig. 12) shows that approximately the inner 40% of the wing reaches Cmax at the same aireraft angle of at- tack. A wartime Focke Wulf report (Ref. 14) indicates that at higher load- ing conditions (i.e., when pulling more gs) elastic deformation of the Fw 190 outer wing shifts the load dis- tribution outboard. This would cause more of the wing to reach its stalling lift coefficient simultane- ously. Combined with the sharp stalling features of the NACA 230XX airfoils, this would produce the harsh stall found in by Capt. Brown. A gen- te stall would be evidenced by a more gradual progression of the 2D stall spanwise. Initial VSAERO calculations were made on a model of the Fw 190 A-8 This version of the aircraft was pow- eted by a BMW 801D radial Naturally, the question arose as to how the aerodynamics of this aircraft differed from the later, Junkers Jumo 213A powered Fw 190 D-9. The Jumo Figure 8 - Calculated Spitfire windshield boundary lay aration. Separation is calculated to take the windshield where the streamli where the separated flow is estimated to reattach higher up the windshield is shown by where the streamline traces, 1ce at the bass traces end. The location engine, an inline, is much longer than the BMW engine, giving the D-9 a clongated nose, which was counter balanced with a 500mm plug added to the aft fuselage. The VSAERO model was modified to represent a D-9 by ‘making these changes and by adding the bulged canopy found on Fw 190 D-9s, It was found from the VSAERO results that the fuselage stretch de- signed by the Focke Wulf engineers resulted in a slight increase in stick fixed stability, with the Neutral Point moving from 35.8% MAC on the A-8 to 40.4% MAC on the D-9, It should be noted these results do not contain propeller effects, which were not modeled. Flight testing of an early model Fw 190 indicated that the air- craft was “just statically stable; stick fixed and free, engine off; and stati- cally unstable to a slight degri engine on” (Ref, 11). During the con- tinued development of the Fw 190 series, the aircraft's CG moved rear- ward as fuel tanks and other equipment was added to the aft fuse- lage (Ref. 15). This Neutral Point shift during development of the Fw 190D model would have been quite valuable in maintaining the continued growth of the design. Drag Comparison There are many conflicting claims as to the equivalent flat plate drag area (f) of these fighter aircraft, Based upon my research, what I believe are SPORT AVIATION 8°a Mustang reengined. by Rolls-Royce with a Merlin 65, The P-51B, with an improved cooling system configuration is even faster than the Spitfire IX The difference in perfor between the Mustang and the Spitfire is attributed to several fac- These include the superior configuration of the Mustang’s cooling sy’ tem and the Spitfire’s tors. Figure 9 - Fw 190 caloulated lift coefficient distribution at 1g stall the most accurate values are shown in Table | The wetted areas of the aireraft are calculated by VSAERO, and exclude the ducts for cooling systems. Notable is that the Mustang has the largest wetted area of this group of aircraft, but has the lowest drag. Evidence of this is that with the same version of the Rolls-Royce Merlin and propeller installed, the Mustan X was measured to be 23 mph faster than the Spitfire IX (Ref. 16). The Mustang X was an Allison powered TABLE 1 Figure 10-- Pressure distribution calculated on the Fw 190 A-B. relatively high level of ex- rescence drag, generated by open wheel wells, a nonretractable tail wheel and other design details (Ref. 17-19). One _ popular piece of aerody namie folklore is ti CDswet value achieved with the Mustang. Vari ous sources quote this value as rang. ing from .0038 to 0043. A review of available wind tun nel and flight test drag data for the Mustang demon strates the need for having all details of the aircraft present if the drag is to be accurately mea- sured. Subscale wind tunnel tests of the P-51A and P-S1B resulted in val- ues of Cr,wer at a representative cruise lift coefficient, in the range of 0046-.0047 (Ref. 20-22). However these tests usually were of models lacking exhaust stacks, surface discon- tinuities, ete. Measurements made in full-scale wind tunnel tests of the P- 51B (Ref. 23) and flight tests of the P-SIA (Ref. 24) and P-S1B (Ref, 21) resulted in a value of Crgyeq Of ap: proximately 0053 Conclusion Important design features of three prominent World War I fighter aircraft have been examined by the use of a ‘moder Computational Fluid Dynamics method. It is hoped that the results pre sented here will help demonstrate some of the valuable important era in fighter aire This information, while hist has relevance in today’s world of @ craft design. Important lessons to be learned are: + Airfoil choice and surface quality are important in achieving the advantages of laminar flow + Cooling system duct design for liquid cooled engines must be con. ducted carefully to avoid losses. + Attention to aerodynamic detail such as windshield slope, can overcome the disadvantage of excess wetted area. + An abrupt stall can be avoided if at- tention is paid to airfoil selection and wing twist + As seen with all three of these air- planes, longitudinal stability and control problems are common, but ean be avoided by the resourceful design essons learned from an Author's Note This article is dedicated to Edward Horkey and Jeffery Ethell, who both contributed information vital to this work. Ed died as a result of injuries sus tained in traffic accident in July 1996 Jeff was killed in the erash of a Lock heed P-38 Lightning in May 1997. Far too young to have participated in World War II, I have long been fasci- nated about finding out how the f aircraft of this war were designed. The deeper I have gotten into this pursuit, the more information I have uncovered that has proven to be valuable in my daily work as an aerodynamicist. I have be come convinced that a study of dynamics! is an important part of an aerodynamicistis ongoing edu iistorical Aircraft Spitfire IX P-51B Mustang P-S1D Mustang Ew 190 4-8 Fw 190 D-9 0 JANUARY 1999 Wetted Area 831.2 f2 874.0 f2 882.2 ft2 738.0 f12 761.6 f12 nee Ref, 0065 16 0053 2 0083 27 0071 26 0063 26cation. To this end, one of my goals has become to try and disseminate the knowledge I have unearthed, this article being an effort towards this end. For ing further information in this regard, I recommend taking a look at my incomplete Guide to Airfoil Usaget at: //amber.aae.uiue.edu/~m ig/ads/aireraft.heml. As mentioned in previous articles, | am an aeronautical engineer, specializ ing in applied computational fluid dynamics. Based in Redmond, Wash- ington, | work for Analytical Methods, Ine. My aerodynamic (and hydrody- namic) consulting projects at AMI have included submarines, surface vessels, automobiles, trains, helicopters, aircraft and space launch vehicles. I can be reached at: [email protected] or: Analytical Methods, Inc., 2133 152nd Ave NE, Redmond, WA 98052 References 1) Zaloveik, J.A., “A Profile-Drag tion in Flight of an Experi- mental Fighter-Type Airplane - The North American XP-51,” NACA re- port, November 1942. 2) Bussmann, K., “Messungen am Laminarprofil P-51 Mustang,” Aero- dynamisches Institut der Technischen Hochschule Braunschweig, Bericht 43/4, January 1943, 3) Matthews, H.F., “Elimination of Rumble From the Cooling Ducts of a Single-Engine Pursuit Airplane,” NACA WR A-70, August 1943. 4) Morgan, Eric B. and Shacklady Edwatd, Spitfire The History, Key Pub- lishing, Stamford Lines, England, 1987, 5) Anon., “Aerodynamic Dimensional Data on P-SIB-I-NA, P-51B-5-NA and P.51C-1-NT Ainplanes.” North American Aviation Report No, NA-5822, August 6, 1943, 6) Smith, J., “The Development of the Spitfire and Seafire,” Journal of the Royal Aeronautical Society, Vol 51, April 1947. 7) Meredith, F.W., “Note on the Cooling of Aircraft Engines With Spe- cial Reference to Ethylene Glycol Radiators Enclosed in Ducts,” ARC R&M 1683, August 19. 8) Hartshorn, A.S. and Nicholson, M.A., “The Aerodynamics of the Cooling Aircraft Reciprocating En- gines,” ARC R&M 2498, May 1947. 9) Morgan, M.B. and Smelt, R Aerodynamic Features of German Aircraft,” Journal of the Royal Aero- Invest nautical Society, August 1944 10) Phillips, W.H. and Vensel, J. Measurements of the Flying Qualities of'a Supermarine Spitfire VA. Airplane,” NASA WR L-334, September 1942 11) Brown, Capt. Brie, “Viewed From the Cockpit; Tank’s Second Iron,” Air International, Vol. 10 No. 2. February 1976. 12) Foster, John Jr. and Ricker Chester $., “Design Analysis No. 9 The Focke- Wulf 190,” Aviation, Oc- tober 1944, 13) Van Wart, F.D., “Handbook For Fw 190 Airplane,” USAAF T-2 Technical Report F-TR-1102 ND. March 1946. 14) Gross, P.. “Die Entwicklung der Tragwerkkonstruktion Fw. 190, Bericht 176 der Lillenthal-Gesellschaft 2 Teil, January 1944, 15) Bentley, Arthur L., “Focke Wulf Fighter,” Scale Models, July 1978 16) Birch, David, Rolls-Royce and the Mustang, Rolls- Royce Heritage Trust Historical Series No. 9, Derby, England, 1987. 17) Private Commu nication, J. Leland Atwood, October 1994 18) Anon, “Estima tion of the Increase in Performance Obtain- able By Fitting a Continuously Variable Radiator Flap,” Royce Experiment: Department Report, August 10, 1942, 19) Private Commu- nication, Ed Horkey November 1994, Rolls- 22) Staff of the RAE High-Speed Wind Tunnel, “High Speed Wind. Tunnel Tests of Models of Four Single Engined Fighters (Spitfire, Spiteful. Attacker and Mustang), Parts 1-5,” Aeronautical Research Council R&M No. 2535, edited by W.A. Mair, 1951 ) Lange, R.H., “A Summary of Dr Results From Recent Langley Full-S Tunnel Tests of Army and Navy Airplanes. NACA ACR L5A30 (WR L-108), 1945, 24) Staffs of the High-Speed Tunnel and High-Speed Flight Sections, “Research on High-Speed Aerodynamics at the Royal Aieraft Establishment fom 1942 to 19. Aeronautical Research R&M No. 2222, edited by W.A. Mair, September 1946 25) Anon., “Performance Calculations for Model P-S1D-5-NA Airplane (NAA Model NA-109),” North American Aviation Report No, NA-8449, December 1, 1944 26) Anon., “Widerstandaten. von Flugzeugen” (Drag Data for Aircraft), Focke Wulf data sheet ° le Figure 11 Pressure distribution calculated on the Fw 190 D-9, 20) Anon., "Wind [~ 20 Tunnel Data For XP. S1B Airplane (NAA Model NA-101),” | 78 North American Aviae | _ <4 tion Report No. NA- $548, October 9, 1943, 21) Nissen, J.M. Gadeberg, B.L. and Hamilton, W.T “Correlation of the Drag Characteristics ofa Typical Pursuit | 92 Airplane Obtained Semispan Fraction From High-Speed ° ° Wind-Tunnel and Flight Tests,” NACA Report 916, 1948, cooling system. Figure 12 - Boundary layer separation calculated in the Spitfire SPORT AVIATION 9)