Flight Test: de Havilland Dragonfly



De Havilland’s 1930s equivalent of today’s bizjet is not just a leading contender in the most beautiful aeroplane stakes, but a pleasure to fly



Walking across the wet grass to the Dragonfly, gleaming in the wan light of an overcast autumn morning, I’m struck by just how incredibly beautiful an aircraft it is?and also how appositely it is named. While we’d been waiting for the rain to stop, my host Henry Labouchere had claimed that “it’s a Renoir, a Matisse of an aeroplane,” and with the sun just starting to illuminate it, I fully understand what he means. An Art Deco masterpiece with wings, it’s absolutely gorgeous. Designed in the mid-thirties, the prototype made its maiden flight on 12 August, 1935. At £2,650 (a small fortune in today’s money) it wasn’t cheap, being essentially the small bizjet of its time and the private aircraft of such giants of British society as Sir Philip Sassoon and Lord Beaverbrook. Contemporary publications described it as ‘a limousine airliner’ with accommodation for four passengers and a pilot. It could also be used for training as there was provision for dual control, and although aimed at the luxury end of the market, most were used eventually for either commercial or military purposes. Production totalled 66, and ended in 1938.

As we begin the preflight I realise that there isn’t a better person to introduce me to the Dragonfly than Henry. Not only is he one of the world’s most experienced exponents of DH biplanes (both as a pilot and engineer), but he also restored and cares for Delta Uniform, and literally knows it inside out.

He explains that although the DH.90 does bear more than a passing resemblance to the better-known DH.89, it is most certainly not a scaled-down Dragon Rapide as both the design and construction methods are very different. Firstly, where the Rapide used a spruce and plywood box fuselage, the Dragonfly’s is a pre-formed plywood monocoque shell strengthened with spruce stringers and constructed using high-strength synthetic bonding resins. This technique had been pioneered on the famous DH.88 Comet racers, and would also be used to great effect on the immortal Mosquito. Other, more obvious differences are that the leading edges of the fabric covered wings are swept back slightly, with a higher aspect ratio and asymmetric spans (the upper being considerably wider), while the engine nacelles are not as deep as the Rapide’s because the main fuel tanks were relocated to the lower centre-section. Moving the tanks away from the engines had the added bonus of making the type less susceptible to catching fire after an accident. Access to the cabin was also improved by deleting the nacelle/wing root bracing struts and inner bay rigging, and strengthening the deep centre-section with very heavy-duty spars. This also allowed cantilever-type undercarriage units to be built into the nacelles, while the undercarriage legs are covered by suitably stylish trousers. The wing spars are boxed up with spruce booms and plywood webs, making the Dragonfly both strong and light. Access to the engines is good, as the nacelles hinge open on both sides to reveal a brace of de Havilland Gipsy Major 10 Mk1-IIIs. These air-cooled, inverted in-line fours produce 145hp each at 2,450rpm and turn wooden two-blade fixed-pitch Hoffman props (recently refinished by Hercules). As they don’t feather, losing an engine could certainly be construed as inconvenient, although as we’re pretty light Henry says the Dragonfly will maintain height ? and might even climb. However, at max weight on a hot day “the best you can hope for is an engine-assisted descent, preferably into a satisfactory field”. On the plus side, operating ‘off piste’ wouldn’t be too onerous because, as the aircraft was conceived when even large international airports were unpaved, the main wheels and tailwheel are quite large and the undercarriage track fairly wide.




Brakes taken from a Pucara

As delivered from the factory in 1936, the brakes were cable-actuated drums but Henry has fitted hydraulic disc brakes from an Argentinean Pucara COIN aircraft. Ailerons are on the top wing only, while the two-piece split flaps are located between the engine nacelles. They don’t look as if they’d be especially effective at generating additional lift, and Henry confirms that they are primarily drag producers.

Most Read

Walking back towards the tail, I notice the 601 ‘City of London’ squadron badge on the fin and also that vertical surfaces bear a strong resemblance to the Mosquito’s. Henry explains that both aircraft were designed by Eric Bishop and that “the Dragonfly is the Mosquito’s second cousin”, while the Dragonfly’s owner Sir Torquil Norman was once a member of 601. (Later I ask Sir Torquil “Why a Dragonfly?” He replies “I remember as a child sitting in a little biplane pedal car made by my father on the apron at Heston, and admiring a Dragonfly. I wanted one ever since!”) The fin looks slightly on the small side, although the mass-balanced rudder is quite large. The curvaceous tailplane carries a broad-chord elevator and I note the absence of trim tabs with interest. The landing light is right in the nose; it and the rest of the electrical system are powered by a turbine-driven generator set into the port upper wing’s leading edge.

With the walk-round complete we climb up into the cockpit via a large door on the port side. It’s a big step up, so there’s a sort of metal stirrup next to it, which is blown up into trail by the air loads when in flight. An interesting feature is the large vertical bar just inside the door. It’s a rather curious item, and Henry said that John Cunningham told him that it is a post-design fix for a minor torsional problem. While probably the least elegant aspect of the whole aeroplane it clearly fixed the problem and is also “jolly handy” when climbing in. The interior is surprisingly spacious, and features a bench at the back, a single seat in the middle and two pilot’s chairs. It is also well-lit as there are no fewer than nine separate transparencies for the cockpit, and a further two windows on either side of the cabin. The baggage bay is behind the bench seat and is accessed via a huge door on the starboard side just aft of the wings.



Having settled onto the left seat with the help of some strategically placed cushions (the seat is “somewhat adjustable”, but cushions are more efficacious) I acquaint myself with the layout of the instruments and controls. In typically British fashion, one is faced with myriad black dials set into a black panel. It is dominated by a large centrally-mounted turn and slip, with the artificial horizon directly above it. The ASI is to the left of the horizon and the directional gyro to the right, while the altimeter and VSI are respectively to the left and right of the T & S. It’s not quite the classic RAF ‘basic six’ (introduced in 1937) but is very close.

The three fuel gauges and an ammeter are on the left of the panel, while the oil pressure and rpm gauges are on the right. Incidentally, if you think that the tachometers look more like voltmeters graduated in rpm this is because ? strictly speaking ? they are. There are also (in common with many other Gipsy-powered aircraft) no oil temperature gauges. A P11 compass is found directly underneath the panel, while the park brake (which looks as if it’s been borrowed from an old Austin) is on the far left.

An interesting instrument anomaly is that more than fifty percent of the ASI is redundant, as the scale goes up to 250mph with 155 at six o’clock but Vne is 144. As for controls, the pilot has a spectacle type yoke mounted on a crossbeam carried by a large column that rises from the floor, while the throttles and mixtures are mounted on the left cockpit sidewall with the tailwheel lock selector underneath. This is a Henry-modification (he’s a firm believer in tailwheel locks) and as the type has a predilection to ground-loop, it seems prudent. An interesting aspect of operating old aeroplanes is that sometimes the controls move rather a long way ? and the Dragonfly is no exception. The control yoke ? and also the throttles ? travel over quite a range. Big, silver button-topped levers on either side of the control column operate the elevator trim and flaps: there are three flap positions ? Up, Take Off and Land. Henry emphasises that although he rarely uses any flap for take off, full flap is always used for landing, and that there is a significant change in pitch trim when the flaps are lowered or raised. Pitch trim is provided by a spring-bias system, but trimming in yaw is almost never done, as it’s very difficult to reach the rudder trimmer (a small handle by the co-pilot’s left ankle). This is also a spring-bias arrangement. Henry says it feels (if you can reach it, I couldn’t!) “like winding up a clockwork train”. The bespoke rudder pedals ? designed by Henry when he installed the hydraulic disc brakes ? are quite narrow. Pressing on their top extremity applies the brakes.




Complex fuel system

Between the seats is a surprisingly complicated looking fuel selector panel, which has two levers (each with four positions!) for the two main 136-litre tanks plus an on/off selector for the 114-litre auxiliary tank at the back of the cabin.

It is so complex that comprehensive instructions regarding its operation are printed on a large placard on the cockpit wall. With all three tanks full the range is an impressive 885 miles, but the rear tank can only be utilised if the load in the cabin is substantially reduced. On wing tanks alone the still-air range is 625 miles.

Both engines start with a discreet cough and a faint puff of smoke as the propeller blades fan into a blur. The panel comes to life as the oil pressures rise, the artificial horizon wobbles on its bearings and the aircraft begins the miraculous transformation from a disparate collection of wood, fabric, rubber and metal into a living, breathing flying machine, 76-years young. Fuel and oil are coursing through its veins, a crackle in my headset indicates that the intercom is working and I can feel the propwash drumming insistently against the elevator through the yoke. It’s eager to fly, so I touch the throttles forward and set off towards the downwind side of the aerodrome. The field of view over the nose is good and the undercarriage feels quite well damped as we taxi slowly across the wet grass with the engines chuffing contently to themselves. When unlocked, the tailwheel simply castors, so steering is by deft, delicate touches of differential brake and throttle.




The magic number is 75

At the run-up point we run through the pre-takeoff checks, test the mags and discuss the various speeds and rpms. Having flown a few old aeroplanes, I’m aware that many of them have a ‘magic number’ (essentially, the same speed is used for several different phases of flight) and the Dragonfly is no exception ? 75mph (65kt) is the speed for Vy, Vyse and Vref. As I swing carefully into wind and reach down to lock the tailwheel I’m slightly uneasy about flying something quite so extraordinary. G-AEDU was built in 1937, and has had a fascinating career ? indeed its history could probably fill a small book. Sir Torquil and Henry have flown it all over Europe, and in 1995 even as far as Oshkosh (see Pilot June 1996) where it was the star of the show. It is also the only airworthy DH.90 in the Northern Hemisphere, and although Henry is in the other seat, he has neither brakes nor throttles. As if sensing my trepidation he reminds me that “it’s not difficult, just different” ? and I know he’s right. I roll forward a couple of metres to ensure that the tailwheel has locked into trail, advance the throttles smoothly and the muted muttering of the motors swells to a roar as the Dragonfly begins to gather speed. Ambient conditions are ISA +2°C with around 10kt of wind on the nose, and we’re about 255kg below MAUW with C of G well forward. Acceleration is quite good with just a slight tendency to swing.

I ease the yoke forward a long way to pick up the tailwheel as the airspeed comes alive, then back to neutral. Catch the incipient swing with a bootful of rudder, the wings take the weight and suddenly we’re flying, skimming low across the grass as we wait for the airspeed to reach 75 before easing up into the climb. Climbing away at 65kt (the ASI is in MPH, but I prefer knots) the VSI is showing well over 750fpm, so at 500ft I reduce power, then draw the port throttle slightly further back (the port engine revs about 50rpm faster than the starboard) to ‘synch the props’ while simultaneously lowering the nose and letting the airspeed build.



Thus far the handling had been somewhat underwhelming, with the ailerons in particular feeling rather spongy. However as the speed increases a subtle but significant transformation in its character occurs and it becomes very nice indeed. I mention this to Henry who agrees, saying that “it doesn’t like to go slow”. While sitting stationary and lifeless on the ground the Dragonfly had been merely a machine, albeit an aesthetically attractive one. But now it’s alive and where it should be; up in the sky. As the speed builds I’m already starting to sense some of its personality and character. It’s a real gentleman’s aerial carriage and the absolute epitome of aeronautical elegance. It positively oozes class. With a couple of thousand feet beneath our elegantly-tapered wings, I swing the Dragonfly into a great sweeping circle above the Cotswolds. My initial impression of a fair field of view is reinforced in flight, while the controls feel nicely balanced, with little ‘stiction’ in the control circuits and low breakout forces. It’s also more nimble than I’d expected, with the handling (especially in roll) much tauter. As with so many aeroplanes of this era, it likes plenty of rudder, but it definitely doesn’t feel like a 76-year old aircraft, which can have distinctly dissonant handling. The controls have plenty of authority, and even quite steep turns are handled with aplomb, although it does need a big handful of power and plenty of back pressure to stop from sinking once the bank angle goes much past thirty degrees. An assessment of its stick-free stability is quite tricky, as the decidedly forward C of G (we have no passengers and the aft-mounted auxiliary tank only contains about twenty litres) means that the pitch trim lever is on the backstop, while the air is slightly bumpy.

With 2,050rpm set I think that the ninety knot cruise is pretty fair, but Henry gruffly announces that he can make it do another eight knots at the same rpm ? and he does. Suitably chastised, I vow to pay more attention to T & S indicator. It’s also interesting to note how smooth the Dragonfly feels, although of course wooden airframes do absorb vibration better than metal.

The next item on the flight test card is single-engine performance, so I close the port throttle gently, bank slightly into the ‘good’ engine and concentrate on maintaining exactly 65kt. As predicted, the Dragonfly responds with a shallow climb, although we really are quite light. One engine at MAUW on a hot day must be a real ‘seat-chomping’ experience.




Slipperier than it looks

Exploring slow flight is interesting. Now, you’d think that with all those wires, struts, trousers and wheels out there in the breeze there’d be plenty of drag, but this aircraft doesn’t fly quite as you’d think: it’s slipperier than it looks, and slowing down takes a while. With flaps up, the power at idle and a steady deceleration of one knot per second there is very little buffet until the wings just can’t hold the weight at around 47kt. Dropping the flaps reinforces the points Henry emphasised on the ground; there’s a big change in pitch trim and the airspeed collapses. It rather reminds me of the Miles Magister. As expected, the flaps are primarily drag generators not lift producers, and the stall speed only reduces by about three knots. The Vfe is 74kt, but in deference to its age Henry recommends 65.

The cameraship now appears and I retract the flaps, add power and edge gingerly into formation. I’d been a bit concerned about this but ? as it has proved in many other aspects ? the Dragonfly is a much better formation aircraft than you’d imagine. If anything, it’s really quite good, as long as the speed is maintained. For example, when we’re on the outside of a turn it’s fine, but when we’re on the inside I can feel the handling deteriorating as the speed reduces. This aircraft really talks to you, and although I have no idea of our actual speed (when in formation I never take my eyes off the cameraship ? in fact I don’t even blink much) I can sense something’s not quite right and, just as Henry says “you’re getting slow”, I add power. It’s quite remarkable, but within a relatively narrow speed envelope the handling degrades from being quite taut to distinctly slack. I ask the cameraship pilot for another ten knots whenever we’re on the inside of a turn.




A real aristocrat of the air...

Cruising back towards Rendcomb at 3,000ft with the power set to 2,050rpm gives a TAS of 112kt for a fuel flow of about 55lit/hr. I’m really starting to feel more comfortable with the Dragonfly, and a glance at the T & S confirms the slip needle is centred. All through the day Henry has been hugely enthusiastic about the aircraft and I can see why. What a machine! It’s a real aristocrat of the air ? an ‘airistocraft’ if you will ? and if it could talk its accent would be distinctly ‘Received Pronunciation’!

With Rendcomb looming large in the windscreen it is time to start thinking about the landing. Vintage machines are at their most idiosyncratic in the circuit, and the Dragonfly is no exception: With power and trim set for 65kt, when you lower the flaps (Henry recommends putting them all the way down in one go) it produces a significant change in pitch trim and a marked reduction in airspeed. Your control inputs are not entirely unlike being on a cross-trainer, because firstly your right hand pulls the flap lever back and your left pushes the yoke forward, then your left hand pulls the yoke back as your right pushes the trim forward. Then you swap hands on the yoke, and return your left to the throttles. As Henry reiterated several times; it’s not difficult, just different. Once it’s back in trim and all the levers are where they should be it’s simply a matter of using power and pitch as appropriate.

Speed stability is good, and I take care to nail the speed to 65. Lose an engine below Vyse and the bottom will fall out of your world. Start bleeding speed back to sixty on very short final and then chop the throttles over the hedge while drawing the yoke back… and back… and back. The controls certainly do have a long throw but (due mostly to Henry’s coaching) we run out of speed, height and elevator at about the same time and the Dragonfly subsides gracefully onto the ground.

I ask Henry if I can fly another circuit without any ‘prompting from the wings’ and he smilingly agrees. This time I’m ready for the swing as the tail comes up and the Dragonfly runs straight and true. Unfortunately I still make a bit of a pig’s ear when flapping and trimming (it’s definitely different) but everything is as it should be as we swing onto base. We’re correctly configured, the speed and angle are both right and ? crucially ? Henry says nothing. A final turn and the Dragonfly starts sliding down the imaginary slope that leads to the waiting aerodrome. I pinch some power off, then squeeze it back on again almost instantly. Henry says nothing. Over the hedge! Throttles and yoke back together. Wait… wait… hold it… a bit of a bump, and we’re down. The landing is slightly firmer than my first effort (I imagine that the Dragonfly probably feels it was a rather vulgar arrival, and it’s used to making an elegant entrance) but it’s acceptable. I suspect that on flat ground and nil wind you can comfortably shave at least 5kt off ‘over the hedge’, but as the field has an upslope we need a little extra energy in the flare.

The engines stop as softly as a butler’s apology, and I can almost picture Noel Coward lounging insouciantly on the bench seat behind. When asked if he’d enjoyed the flight he’d reply “well, aeronautically it was a great success. Socially it left much to be desired!”


...and very much a “dear friend”

In conclusion I fully understand Torquil and Henry’s tremendous enthusiasm for this machine (Torquil describes it as “a dear friend”). It looks great, goes well and does an awful lot on only 290hp. And when you consider that it was designed 78 years ago…

As we slide the hangar doors shut I take one last lingering look at the de Havilland DH.90 Dragonfly. The ‘limousine airliner’. The most beautiful biplane in the world.


Enjoyed this article? Why not buy the latest issue of Pilot, or become a subscriber?