Department for Arm-Waving Aerodynamics

Something Controversial from the WFP DAWA (Department for Arm-Waving Aerodynamics) - de Havilland props gave the altitude problem, not engines.

The Whirlwind was famously criticised for being ‘useless at height’. The disappointed RAF presumed that the Peregrines just weren’t performing as they should, once they had run out of supercharger at 15,000 feet and boost begun to drop.


But something didn’t look right to DAWA – the Peregrine had a ‘full throttle height’ of 15,000ft, and that is indeed where the supercharger started to lose its ability to maintain boost – but only (on the tested prototype – see later) at the same rate as a Merlin III. Meanwhile, graphs from Westland in AIR 16-326 showed that the aircraft speed actually continued to rise above the 15,000 feet at which boost, and horsepower, began to drop then began falling 1,800 feet higher. Looking around, DAWA found the case of the Spitfire I. The manufacturer’s full throttle Height for the Merlin III was 16,250 feet at original rating, but the Spit I reached its maximum speed at around 19,000 feet (2,750 feet higher).


This made DAWA realise that altitude performance wasn’t exclusively linked to engine performance – though conventional wisdom at the time had it as such. In fact it was also affected by the aircraft’s propeller. Clear evidence of this comes from the table of Spitfire II airscrew performance at – these are the results of trials by Vickers, the RAF and Rolls Royce:

Fig.1 – Comparative heights of maximum True Air Speed attained, Spitfire II aircraft, 3,000rpm, Rated Boost, Merlin XII, various propellers.


These figures clearly demonstrate that the same aircraft, with the same engine at the same RPM, with the same official ‘full throttle height’ of 17,550 feet had a ‘maximum speed altitude’ (my phrase) that varied by 3,000 feet, between 16,000 and 19,000 feet (the latter using the De Havilland 55409B, the standard for the Spitfire I). The trials showed this height to vary with blade thickness.

One thing stands out about the standard Peregrine ‘power egg’ – the DH propeller blades are remarkably thick for a high- performance fighter. According to the de Havilland specs still available at Farnborough (and used for our project’s CAD model, below left), thickness to chord ratio at 0.7 radius is around 0.096. For comparison, the Spitfire I had very similar


profile de Havilland blades (above right, from ARC RM.2357), of type 55409 as opposed to the Whirlwind’s type 54409 – but with a crucial difference – a t/c ratio of 0.076 at 0.7 radius

Fig.2 – WFP blade design curves, DH 54409

Fig.3 – Comparison of Thickness Distributions, various propellers, ARC Memo. Ref.7

Without going into the maths here (let’s just keep it arm-waving) what this means, based upon NACA report 463 (excerpt graph below) into the specific aerofoil profile of the 54409 (RAF6) and several others on the subject of propeller compressibility, is that at the max 3,000 rpm, at 16,800 feet and 355 mph, the Whirlwind’s propeller (who’s speed is a

function of both rotational and forward speed) has hit critical Mach. Not Mach 1, but critical Mach for the prop’s aerofoil – one that was designed in the 1920’s.

Fig.4 – Aerofoil drag rise (V/Vc = Mach No.). Ref.1

In other words, a lot of the thick propeller blade has already hit compressibility issues at 355 mph at the indicated altitude, and they all come on quite rapidly as the aircraft transitions 350-360mph at 16,000 to 17,000 feet. Of course, Mach drops with altitude – so the limiting critical speed does too. The speed above this critical height will always be restricted by this very ‘draggy’ effect..

All this led me to realise that something interesting happens due to the same un-planned-for compressibility drag in ‘climbing condition’. Here the minimum drag coefficient at De Havilland prop tip goes from 0.2 to 0.4 at the tip speeds encountered in this condition (Mach 0.71 to Mach 0.78), as the aircraft (hypothetically) climbs from 10,000 to 30,000 feet. To counter this, and a lesser effect inboard, there is a corresponding dropping of the angle of attack of the constant speed prop blades (by about 1.75 degrees) to maintain the revs at the climbing speed of 2,850rpm – causing the aircraft to loose forward speed for any given angle of attack (by as much as 10%), and thus reducing the climb.


This dropping of the angle as the blade hits compressibility and searches for a lower drag regime has another, less obvious effect – which explains the boost problems reported. The intakes for the carburettor – and then the supercharger – lie directly behind the blades, inboard of each engine – and boost is a function of intake pressure as much as anything else. Lower propeller efficiency when the prop is in front of the intake means lower pressure going into the supercharger, which then becomes become less effective in adding to manifold pressure.


That the propeller affects air volumes entering leading edge ducts is clearly shown by 1945 NACA tests on the Grumman Tigercat’s oil cooler intakes (below). The Tigercat tests showed that where threre was insufficent thrust from the propeller in front, oil tempertaures rose.

Fig.5 – Tigercat at the NACA research station,   Ames. Ref.8

Any such problem would be compounded in the climb by the peculiar effect of aircraft pitch – called the ‘P-factor’. Any additional angle of attack resulting from the aircraft being put into a climbing attitude has the effect of accelerating the ‘down-going’ blade relative to the surrounding air stream (increasing the effects of compressibility, including the compensating automatic reduction of blade pitch), and crucially decelerating the ‘up-coming’ blade. Further pitch change of the faster downgoing blade (and thus the whole unit) now isn’t just a mechanical effect of the constant speed unit – it’s now, at this raised Mach – also a natural tendency of the blade

aerofoil to pitch ‘downwards’ above critical speeds, a phenomenon known as ‘Mach tuck’ when applied to aircraft wings in transonic flight.

Fig.6 – blade slipstream affecting flow into inlet behind. Ref.8

Fig.7 – Oil inlet temperatures in the climb, Port and Stbd.

Unfortunately there is no data on the boost pressure variations with height of a production Whirlwind, but the oil temperature plot

of a production Whirlwind ( right) is indicative – the  serious relative  rises above 18,000ft were on the right hand side, and the intake for the carburettor shared the same inlet as the oil cooler. It was known that all production Whirlwinds’ starboard engines overheated like this, while the tested prototype’s didn’t.

At high altitudes, where propeller effectiveness has been drastically reduced and the blade pitch is ‘against the stops’ at 28 degrees to deal with the increased drag from the down-going blades at high Mach, DAWA calculates (on the back of a beer mat) that there was precious little thrust available overall from the inefficient blades and in fact a small negative thrust – effectively a drag – develops in front of the right-hand intake, at likely angles of attack and known climbing speeds above 18,000ft.Boost drop on the non-critical engine wouldn’t give too much yaw (again with reference to P-factor theories) and a veteran asked said that as a fighter pilot he was too busy looking outside to notice such things as long as both engines were working. Still, DAWA suspects that boost pressure deficiencies would show up first on the right hand side, just like the excess temperature

Left: Fig. 8 – Positioning of the intakes immediately behind the blades.


Right: Fig.9 – Situation in the climbing condition above 204 mph, 2,850rpm, 18,000ft, 6 degree angle of attack. Overall there is still some propulsion, of course, but in front of the right-hand intake (both cooling and carburettor) there is a problem..

Fig.10 – De Havilland 4/4 ‘bracket’ hub, standard on the Westland Whirlwind with DP54409 blades

There may even be a tertiary factor in reducing the ability to climb. As our thick down-going blade passes through Mach

0.8, the ‘zero lift’ angle rises dramatically. This means that the blade, already at negative pitch, stops producing lift at all when this angle is met – and starts getting even draggier. The blade will not coarsen to correct – the constant speed mechanism wasn’t an ‘intelligent’ system, it just reacted to the rpm-slowing effect of drag by making the pitch finer. This is now working the ‘wrong’ way, once this point has been passed.


Rough calculation (remember, this is arm-waving, not a paper submitted to the RAE) indicates that this happens to our nominally 6-degree angle of attack, climbing Whirlwind at around 27,000 feet. It won’t go any higher, in that condition.


In a letter from Lord Dowding, CinC Fighter Command, to the Ministry of Aircraft Production written in October 1940 when the future of the Whirlwind was in the balance, Dowding quoted a report from the ‘Squadron Commander’, presumably the O/C 263 Squadron. It deserves quoting at length here, as the ultimately decisive problems with boost and airscrew revolutions may both be explained by the above effects, without there being anything wrong with the engines:

  1. I have now received a report from the Squadron Commander concerning the performance and attributes of the Whirlwind, and I have been particularly anxious to get a report on its performance at height because the rate of climb curve showed an ominous flattening out at about 25,000


  1. The report contains the following paragraphs:-


“There is no tendency for the aircraft to spin in steep turns, but at heights over 26,000 feet the performance falls off rapidly and it is difficult to maintain height…. It must be emphasised, nevertheless, that the performance of the Whirlwind above 20,000 feet falls off rapidly, and it is considered that above 25,000 feet its fighting qualities are very poor. The maximum height so far attained is 27,000 feet but on every occasion that a height test has been carried out there has been a minor defect, either in airscrew revolutions or in lack of boost pressure..”


  1. The limiting factor in the recent fighting against ME109’s in the South East of England is the performance, manoeuvrability, and climb, at high altitudes, and a difference of 2,000 feet in Service ceiling is a very important advantage.


  1. It therefore seems to me quite wrong to introduce at the present time a fighter whose effective ceiling is 25,000 feet.


So, why wasn’t any deficiency noted by Martlesham Heath, who gave the aircraft a ceiling of 31,000 feet and didn’t refer to rpm problems and boost issues – or temperature issues either? Squadron pilots were asking the same question – saying that the altitude performance just wasn’t what it should have been, or even what it was when tested previously.

This is because the prototype example of the aircraft carefully and rigorously performance tested and used to create official figures had different prop blades from all other Whirlwinds!


Quite bizarrely the prototype sent to Martlesham Heath had a one-off Rotol design, not the props hung on all production Whirlwinds – and nobody seemed to think that this mattered. Indeed Martlesham themselves, seemingly in response to a lost letter, re-affirmed emphatically that L6845 was representative of production aircraft to the point where it might be considered the first production Whirlwind.

Left: Fig.12 – Production standard de Havilland 54409 blades on prototype L6844, which was never performance tested


Right: Fig.13 – One-off Rotol blades of unknown type, carried on prototype L6845, tested and the source of RAF trial figures

Perhaps missing the effects of compressibility on propellers, still a new field at the time, was understandable – the dates of the references at the foot of this article (especially NACA Technical Report 639) show that the crucial investigative work was being done around 1938, exactly the same time as

the already-chosen props were being hung on the prototypes. Nevertheless, testing with the ‘wrong’ propeller seems downright perverse.

By comparison, the Spitfire I hit the same sort of problems (minus intake pressure issues), only higher up. The DH blades were similar, yet thinner, and thus hit compressibility drag issues at a higher altitudes. Certainly the blade was thinned even further by DH, to get an even higher critical altitude and speed out of the Spitfire I. By the time later marks and the Mustang et al were doing 420mph-plus at altitude with two-stage superchargers and blade tip speeds of over Mach 1, blades were of very thin section and NACA-developed profiles that negated compressibility drag and lift-loss completely.

Fig.11 – Pages from NACA Technical Report 639. The work of David Biermann and Edwin P. Hartman – the authors of NACA report Refs 2,3,4 &5 – in the NACA 20ft wind tunnel in the late 30s and early 40’s is still referenced by propeller theorists worldwide. The slow dissemination of the lessons learned around compressibility and blade thickness came too late for the Whirlwind. Beirmann and Hartman used the Navy 5868-9 blade design (on the left of the five illustrated above) as a standard test blade. Apart from being fractionally thinner than the 54409, this blade clearly formed the basis of the DH design (via Hamilton Standard, of course) – something of a lucky co-incidence, as only fragmentary de Havilland design data survives (at the Library of the Farnborough Air Sciences Trust), and the American data fills in the gaps.

In hitting these limits, the Whirlwind – and the Spitfire – ran into the issue of being faster than almost everything else. The difference was, whereas Rolls-Royce, the RAF and De Havilland all got together to work out how to make the Spitfire even better, no-one seemed that interested in helping out Westland with their baby. If only someone had tried nice thin blades on the Whirlwind (or even re-fitted the Rotols), rather than blaming the engines.



After writing this, the DAWA actually stopped waving arms about long enough to read a relevant document – Thanks to Brian Marsh’s hard work at the NA, we now have access to AVIA 6/13712 – the unpublished Part III of the Farnborough wind tunnel tests carried out on the first prototype in 1940. This includes an actual comparative flight-trial record of maximum speed of the prototype with the DH prop. It was 350mph.


The report noted that the wind tunnel drag tests on the same actual airframe and some thrust calculations to replicate the effect of propellers indicated that its top speed should have been 355mph (like the Rotol-equipped L6845), and the only way they could account for it was a difference between calculated and actual airscrew propulsive efficiency. The report said it would need to drop from the theoretical 77% to 73% to fit the real world top speed of 350mph.


It also gave some specifics factored into the calculation – including a blade thickness of 0.09 t/c. As we know, it was actually 0.096 – perhaps Farnborough used data from the original American Bu Aer 5868 pattern blade rather than the thicker British version, the DH DP54409 actually used on the aircraft.


Now, DAWA used some actual maths to work out the difference in drag between a 0.09 (wrongly used in the equation) and 0.096 (real world) blade at the appropriate relative Machs for this speed and height (tip speed 0.86 Mach), and using strip theory and lots of NACA blade performance charts in quite a crude way came up with a difference in drag co-efficient of 24%.


Plugging the new figure back into the 75 year old sums, the propulsive efficiency of the aircraft dropped from 77% to… (drumroll, please) 73.4%!


This shows that blade thickness differences of around 2mm were critical, in several ways.

Fig.14 – L6844 undergoing tests, 1940. Ref.6. It has been proposed to display P7056 like this, in the same tunnel in Farnborough. But that’s enough controversy for one newsletter.



  1. NACA TR 463, The NACA High-Speed Wind Tunnel and Tests of Six Propeller Sections – Stack 1934
  2. NACA TR 639, The Effect of Compressibility on Eight Full-Scale Propellers Operating in the Take-Off and Climbing Range – Biermann and Hartmann, 1938
  3. NACA TR 641, The Negative Thrust and Torque of Several Full-Size propellers, and their Application to Various Flight Problems- Biermann and Hartman, 1938
  4. NACA TR 642, Tests of Full-Scale Propellers in the Presence of a Radial and a Liquid-Cooled Engine Nacelle, including Tests of Two Spinners- Biermann and Hartmann, 1938
  5. NACA TR 650, The Aerodynamic Characteristics of Six Full-Scale Propellers having Different Airfoil Sections – Biermann and Hartmann, 1939
  6. RAE BA Departmental Note 38 (unpublished), Tests on Whirlwind L6844 Part III, Estimation of Top Speed and Drag from 24ft Tunnel and Performance Tests – D.W. Bottle and T.V. Somerville, 1940
  7. ARC R&M 2357, 24-ft. Tunnel Tests on a Rotol Wooden Spitfire Propeller B. Haines 1942
  8. NACA A5C10, Investigation of Slipstream Effects on a Wing-inlet Oil-cooler Ducting System of a Twin-engine Airplane in the Ames 40- by 80-foot Wind Tunnel – D.R. Chapman, 1944

More Aerodynamics - with pictures!

Thanks to Matt and Gunnar

The reason for the acorns

Tests on the prototypes showed severe tail vibration at speeds over 380mph. This was famously fixed with the ‘acorns’, but once a solution had been found through trial and error there was little further thought given to how it worked.

Aerodynamic theory has moved on since then, as has the ability to model aerodynamic behaviour. so here is the reason behind the problem and the fix:

The tail presents two symmetrical aerofoils at 90 degrees to each other. Where they are both at their thickest the airflow is accelerated by up to 40%, and the pressure drops accordingly. At flying speeds there is subsequent flow separation, rapid flow deceleration, and extreme turbulence. This is the classic ‘interference drag’.

Tests on the prototypes showed severe tail vibration at speeds over 380mph. This was famously fixed with the ‘acorns’, but once a solution had been found through trial and error there was little further thought given to how it worked.

Aerodynamic theory has moved on since then, as has the ability to model aerodynamic behaviour. so here is the reason behind the problem and the fix:

The tail presents two symmetrical aerofoils at 90 degrees to each other. Where they are both at their thickest the airflow is accelerated by up to 40%, and the pressure drops accordingly. At flying speeds there is subsequent flow separation, rapid flow deceleration, and extreme turbulence. This is the classic ‘interference drag’.

You can see this problem zone as the blue blob on the (unmodified) tail in the pressure plot on the left (blue=low pressure)

You can see this problem zone as the blue blob on the (unmodified) tail in the pressure plot on the left (blue=low pressure)

Copyright WFP 2011/2015

The graphic below shows a velocity plot ‘slice’ close to the fin/horizontal tail junction, at 400mph. Red=fast flow and blue= slow. Air passing close to the junction is accelerated up to 480 mph in this low-pressure zone, then immediately slowed to under 15 mph (the dark blue), in the space of a few inches. As Gunnar observed, it’s not surprising pilots reported rudder vibration at speed.

The plots below are Gunnar’s model at 400mph. Adding the acorns reduces the severe pressure drop (the blue area) and the local velocity increase, indicated by the reduction in the orange area (this is the portion of local air accelerated to over 525mph)

Copyright WFP 2011/2015