AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04

immediately acknowledged this second transmission. It s possble that the arcraft was n a known rado blnd spot when the first transmission was made. At 029:07 hrs the plot called the Lossemouth Radar Controller advising him that he was at FL95. The 4 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 Lossiemouth Controller confirmed that the aircraft was identified and informed the pilot that he, the controller, was providing a RAS. The pilot acknowledged the radar servce he was recevng and, at 029:34 hrs, he requested descent. By this time the aircraft was in the area where it was usual for the pilot to make such a request. However, the controller commented that, ntally, he nstructed the plot to “standby” because the arcraft had been handed over to him “a bit early”. At 1029:50 hrs he cleared the arcraft to descend to FL75 and nstructed the plot to report when level. The pilot acknowledged in a clear, unhurried voice. This was the last transmission heard from the pilot. The ATC controller observed G-TWIG’s descent rate on radar, whch appeared to be typcal for that flight. At 1032:59 hrs he advised the pilot that there was temporary loss of radar contact and, as a consequence, the ATC servce was reduced to a Flght Information Service (FIS). There was no reply from the pilot. Twenty seconds later the radar controller called the plot agan and mmedately another arcraft, a helicopter, transmitted on the frequency. Over the next mnute the Lossemouth Radar Controller and the helcopter’s crew conducted a dalogue durng which the periods of silence totalled 25 seconds. Followng that conversaton, the Radar Controller called G-TWIG eght tmes n the space of seven and a half minutes. On each occasion there was no reply from the arcraft and, durng that perod, there were no other transmissions on the frequency. From the ATC rado recordngs, the plot sounded lucd and calm from the tme he requested clearance to tax at Stornoway untl hs last transmsson at the top of descent. He did not transmit an emergency call and he gave no indication of any problems. Search and Rescue activity At 036 hrs Lossemouth ATC nformed the Scottsh Ar Traffic Control Centre (Military) Distress and Diversion (D&D) Cell at Prestwick of the situation. D&D attempted to contact the plot of G-TWIG on the aeronautcal emergency frequency, 121.5 MHz. There was no response. At 1046 hrs Lossiemouth also contacted the Aeronautcal Rescue Co-ordnaton Centre (ARCC) at Knloss and passed all the known detals of the arcraft’s disappearance. Further unsuccessful attempts were made to contact G-TWIG by rado from ground statons and another aircraft that was flying from Stornoway to Inverness some 25 minutes behind G-TWIG. Two Tornado aircraft were diverted from their training flights to search the vicinity of the last radar contact. While t was possble to make a vsual search of some of the valleys, the crews reported that cloud was coverng a plateau of high ground in the area. At 1107 hrs a Sea Kng Search and Rescue (SAR) helcopter was launched from RAF Lossiemouth. The coastguard helicopter based at Stornoway was also moblsed and the arborne search was augmented by mountan rescue teams from Dundonell and Kinloss. The arcraft wreckage was found by a mountan rescue team the following day at 1330 hrs. It was located at an elevaton of 2,480 ft amsl on Meall Feth na Slatach, a broad mountain ridge in a remote area of the Highlands, 30 nm to the north-west of Inverness. The severity of the mpact had scattered the arcraft over a wde area and into many pieces. When viewed from the air, even n good vsblty, the small sze and large spread of the fragments made the aircraft difficult to distinguish amongst the intermittent quartz type rocky outcrops. Four people who were fishing on Loch Vaich, 5 nm to the south-east of the crash ste, and a number of estate staff, who were workng n the area, all heard a loud bang or 5 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 explosion on the day of the accident at about 1030 hrs. The nose had come from the drecton of the crash ste but no-one had seen any sign of an aircraft. Later, some of them saw the two Tornado jet arcraft and an SAR helicopter which had been searching the area. Pilot information The pilot started his flying training in the USA in 1998 and qualified as an ‘airplane’ and instrument flying instructor on single and multi-engined light aeroplanes. In 2000 he returned to the UK to contnue hs tranng for a commercial pilot’s licence for aeroplanes. In March 200 he was ssued wth a UK Commercal Plot’s Lcence (Aeroplanes) and commenced employment as a co-pilot, flying the Dornier 228 on a short-term contract for an overseas operator, based in Aberdeen. That contract ended n July and he was offered employment wth another regional operator in Scotland. He declined the offer n the hope that he mght secure a poston on larger aircraft further south. The events of September 2001 and a subsequent downturn n the avaton market thwarted hs aspratons and he accepted a full-tme poston wth that same operator in June 2002. By all accounts he had much enjoyed the nearly two and a half years he had spent flying passengers and freight, predomnantly around Scotland and to the Northern and Western Isles. He had started on single-pilot duties on the company’s F406. Eleven months later he transferred to the company’s Jetstream 3 (J3) as a co-plot and n July 2003 he combned that duty wth hs prevous role on the F406. In October 2003 he was issued with hs JAR Arlne Transport Plot’s Lcence (Aeroplanes), vald untl 2008, and he completed command tranng on the J31. He flew the J31 exclusively until January 2004, while he accrued some experience as its commander. Then, once more, he combned hs dutes on the J3 wth single-pilot operations on the F406. He had commented that he would probably not experence such enjoyable flying again. In August 2004 he successfully underwent the selecton procedure for a short-haul jet operator who he was due to join in November. A week before the accdent the plot had swapped the ‘standby’ duty, for whch he was rostered on the date of the accdent, wth the F406 duty allocated to another pilot. It was understood by the other pilot that the request was made because t would then be the accdent plot’s last flight into Stornoway in the F406 before he left the company. However, his roster showed that he still had a J31 duty and three more F406 duties the following week. The last was on the Frday and would have nvolved the same routing as that on the date of the accident. Certanly, three of the ground staff n Stornoway were expecting the pilot to fly there on the following Friday’s F406 flight. There were a number of references n the plot’s tranng file to good performances and there was no record of him experiencing any difficulties during his conversion or recurrent training on either the F406 or the J31. He had revaldated hs F406 type ratng and hs Sngle Plot Aeroplane (SPA) instrument rating on 30 June 2004. His JAA Class One medical certificate, with no limitations, was valid until 5 November 2004. All his other annual and trennal checks were n date and, n all respects, he appeared to be medically fit and well. The plot had been on standby duty from 0800 hrs untl 600 hrs the day before the accdent but he was not required to fly. The following morning he reported at 055 hrs, gvng hm a 3 hours and 5 mnutes rest period prior to the accident duty and the benefit of no flight duty period since landing a J31 at 2015 on 6 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 20 October 2004. The pilot’s previous flight in an F406 had been on 18 October 2004. The plot was descrbed, by those who knew hm at work, as a steady, joval ndvdual, who was well-lked and respected. He was considered to be a conscientious, able avator and one who was partcularly known for adherng to standard operatng procedures and for being safety conscious. His family and his fiancée said that he was physically very fit and that he had a happy personal life. He had also carried out at least one other ‘exuberant’ departure in an F406 when flying single-pilot without a payload. Description of the aircraft and relevant systems The Rems Avaton F406 Caravan II s an un-pressursed utility aircraft. Its interior can be configured to carry passengers and/or freight, or surveillance equipment. The man entry door s on the left sde of the rear fuselage and is available in several configurations. The door on G-TWIG consisted of front and rear sections. The forward half was hnged at ts leadng edge and thus opened forwards. The rear section was split longitudinally n the mddle, the upper part openng upwards on a gas strut and the lower secton, contanng ntegral steps, opening downwards. This door also served as the normal means of entry and exit for the pilot(s). In addition, an escape hatch, ncorporatng the left sde cockpt wndow mmedately aft of the wndow, was provded for the plot, wth two addtonal escape hatches on the left and right sides of the cabin. Additional freight/luggage space was avalable n the nose and aft sectons of the engne nacelles, wth access to the latter beng va lockable doors on the upper surfaces. The nose baggage area was equpped wth two doors on the left sde and one on the right side. The landng gear s of conventonal, trcycle desgn, retracted and extended by hydraulc actuators powered by engine-driven pumps. The arcraft s powered by two PT6A-2 turboshaft engnes drvng McCauley three-bladed, varable ptch propellers. All PT6 engines consist of two independently rotatng sectons; the gas producer and the free power turbine. The former directs a high energy gas stream at the latter, whch drves the propeller through a reducton gearbox. Cockpit controls include a power lever and propeller rpm lever for each engine. The rpm lever s connected to a propeller control unt (PCU), whch incorporates a governor assembly. The latter controls engne ol pressure ported through a transfer tube to the nsde of the dome that forms part of the propeller hub. This results in forward movement of the dome, whch, because t s connected to the propeller blades via levers, causes the blade angles to reduce. However, dome movement s opposed by the combned force of an nternal sprng (the featherng sprng) and the effects of centrfugal counterweghts mounted on each of the blades. The propeller blade angle is thus set by the poston of the pston and wll vary accordng to the power and rpm selected by the pilot. A ‘beta system’ prevents the blade angles reducng below a pre-set value in flight, - the primary blade angle (PBA). The ‘beta range’ of propeller blade angles s the area of operaton below the PBA (4° n ths case) used on the ground for taxiing and reverse thrust. Control is by means of the power lever below the ‘dle’ detent and s connected to the beta valve, mounted on the front of the PCU, va a reverse thrust cam box assembly. It is the beta valve that regulates oil flow to the propeller dome in this mode of operation. In the air, when the blade angle reduces to the PBA, a flange on the dome contacts the ‘beta nuts’, whch are attached va rods to a brass slp rng on the propeller shaft. A carbon block, located in a groove in 7 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 the slp rng s connected, va a feedback arm, to the beta valve. Any additional forward movement of the dome causes the beta valve to reduce the ol pressure, thus preventing the blade angle reducing below the PBA. The governor wthn the PCU should prevent the propeller from overspeedng; however, each engne s also equpped wth an overspeed governor that prevents excessve rpm that could result from a falure wthn the PCU. The primary flying controls are manually operated and manly comprse cables, bellcranks, pulleys and quadrants. The elevator, aileron and rudder trim systems are all cable drven, wth screw-jack assembles attached to the trm tabs on each elevator, the left aleron and the rudder. They are operated via trim wheels on the cockpit pedestal. The arcraft’s elevator trm tab can be adjusted manually usng a trm wheel on the centre console or by the electrcal trim system. The electric trim system consists of an electrcally operated drve motor and clutch assembly, whch receves power through a two-way swtch (ptch up and ptch down) and an autoplot/electrc elevator trim disconnect switch. Both are located on the left arm of the pilot’s control wheel. Operation of the electric trim switch disconnects the autopilot. On G-TWIG (whch was equpped wth a Sperry 000A autoplot) operaton of the dsconnect swtch dsabled the electrc trim when the switch was depressed and released. The electrc trm then remaned dsabled untl the trm swtch was actuated once more. The flaps are selected electrically and operated hydraulcally by means of an actuator mounted on the rear spar of the wing centre section. The avionic fit on the F406 varies according to operator requirements. G-TWIG was equipped with an ARC (formerly Sperry) 1000A autopilot system. This was a relatvely unsophstcated devce, compared wth modern equvalents, but t could mantan a headng and alttude; addtonal features ncluded navgaton, approach and go-around modes. There was no ‘altitude acqure’ functon although clmbs and descents could be acheved by means of a thumbwheel on the control panel. This could be rotated so that the aircraft adopted the desired nose-up or nose-down attitude. An alternatve way of achevng the same result was to depress a ‘ptch sync’ swtch on the control yoke whch temporarily disconnected the autopilot. The aircraft was then manually placed n a new atttude whch was held by the autopilot on releasing the switch. The autopilot controlled the arcraft va servo motors operatng on the aileron and elevator circuits. It also trimmed the aircraft in pitch by means of the elevator trim actuator. Finally, a yaw damper was ncorporated nto the autoplot system, with an actuator operating on the rudder. The autopilot could be swtched off by means of a swtch on the control panel, a dsconnect swtch on the control yoke or by operaton of the electrc trm swtch, also on the control yoke. Accident site details The arcraft had crashed nto rough, undulatng terran at an elevation of around 2,500 ft. The ground was a mixture of peat bog and grassland, with rocky outcrops. The mpact area had grante beneath the surface, whch combned wth what was evdently a hgh mpact speed, had caused extreme fragmentation of the aircraft. A shallow crater had been formed, wth some wreckage scattered to the rear of t, but the majorty havng been thrown forwards over a dstance of approxmately 250 metres. The distribution of the wreckage suggested a steep mpact angle, estmated at around 70°, wth 8 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 the wreckage throw ndcatng an mpact track of approxmately 200°M, whch was at rght angles to the approxmately south-easterly course the arcraft had been following towards Inverness. Many wreckage items were lightly burned, indicating that a fireball had occurred at impact. This would have resulted from mstng fuel followng the dsntegraton of the wng tank structure, wth lkely gnton sources beng electrcal or hot engine exhaust gases. There was no evidence of a pre-impact fire. Wthn the broken rock of the mpact crater, t was possble to dscern the mpresson made by the wng leading edges. The remains of the wing-tip navigation lght bulb-holders were found at each extremty of the impression. This indicated that the wing was structurally ntact at the tme of the mpact although the degree of fragmentation of the wreckage meant that it was difficult to determne whether any panels from elsewhere on the aircraft had become detached prior to impact. The dstance between the two wng-tp mpact postons was 54 ft, compared with the wingspan of around 49.5 ft. Ths ndcated that the arcraft yaw axs was at an angle of approxmately 22°, left wng low, relatve to the ground at impact. The accdent ste was n a remote locaton and could only be accessed on foot or, weather permttng, by helicopter. Following the on-site examination, the Royal Ar Force Arcraft Recovery and Transportaton Flght gathered the wreckage together n groups of large bags, whch were formed nto under-slung loads for a seres of helicopter flights to a collection point close to a road. The wreckage was then taken to the AAIB’s faclty at Farnborough for a detailed examination. Detailed examination of the wreckage i) General The severely fragmented wreckage was sorted to extract identifiable system components such as airframe, power plant, flying controls, electrical equipment, and transparencies. Windscreen fragments were examined for evidence of bird remains but none was found. The remans of a number of cockpt nstruments and controls were also recovered and identified, although the degree of damage was such that ther examnaton contrbuted little to the investigation. The examination established that the flaps and landing gear were retracted and that all the extremtes of the arcraft were accounted for wth the excepton of the nose cone. However, since this was the first part of the arcraft to strke the ground, t s probable that t was damaged beyond recognition. Pieces of the forward fuselage structure mmedately aft of the nose and the weather radar antenna were identified. The main door had suffered severe damage. The only part that had survved reasonably ntact was the rear lower secton that ncluded the steps; ths showed evdence of longtudnal crushng, whch suggested that the door was n poston at mpact, and that t had been compressed between the tralng edge of the forward secton and the aft door aperture. This in turn suggested that the forward door section had been in position. Dstorton of the lockng mechansms of the nacelle baggage doors confirmed them as being secured at the time of the impact. Also, fragments of the forward nose baggage doors were identified by means of lettering painted on the external surfaces. The degree of fragmentation suggested that they were most probably closed at impact. The rearmost nose baggage compartment door on the left 9 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 side was not positively identified. Pieces of the pilot’s escape hatch and the over-wng cabn exts (all outward opening) were identified, although it was not possible to confirm that they were secured at impact. ii) Flying controls a) Primary flying control system The steep nature of the mpact had resulted n severe fore-aft compresson of both the horzontal stablsers and the elevators. It was noted that both elevator balance weights were present. The elevator controls at the rear of the arcraft conssted mostly of rods and bellcranks; there was no evdence of pre-mpact falures n any of them. The rudder surface had remained attached to the severely damaged fin and both ailerons were recovered. The fragmented nature of the wreckage meant that t was not possble to dfferentate between many of the peces of the flying control operating cables in terms of whether they orgnated from the aleron, elevator or rudder circuits. However, all the failures bore the characteristics of overload, with no evidence of pre-impact failure. b) Secondary flying controls Representative portions of the flap surfaces were recovered and identified, indicating that they were present on the aircraft at impact. The hydraulic actuator was found wth ts ram n the retracted poston, ndcatng that the flaps were retracted at impact. The aleron trm actuator was not recovered and identified, although it was established that its attachment to the aileron tab had failed in overload. Only a small pece of the aleron trm tab was found; however the elevator and rudder tabs were complete and had remained attached to their respective surfaces. The rudder trm actuator was found n ts approxmate mid-travel position. There were two elevator trm actuators on ths arcraft, operating tabs on both elevators. Both units were present n the wreckage and the lnkages to the tabs were intact. Each actuator comprised a ‘twin-pack’, whch conssted of two screw-jacks drven by sprocket assembles whch n turn were operated by chans that formed part of the elevator trim circuit. Operation of the ptch trm system (whether by means of the manual or electrc system, or by the autoplot), thus caused all the jack-screw assemblies to move in unison. A diagram of one actuator, together wth photographs, s shown at Figure 1. Rotation of the sprockets caused the sliders (whch were attached to rods that moved the tabs) to move back and forth: they extended for nose-down trim and retracted for nose-up trim. All the sliders were extended by a similar amount. Comparison with an ntact arcraft revealed that the slder postons equated to almost a fully nose-down trim condition. Durng the hgh-speed mpact, n whch the arframe must have dsntegrated extremely quckly, tenson n the trm operatng cable/chan system would have been lost due to foreshortening of the fuselage. However, as the tal secton broke up, there may have been scope for consderable snatch-loads to be appled to localsed lengths of cable close to the elevators. Whilst such loads may have moved the trm actuators, the smultaneous dstorton that was occurrng n the structure and tab lnkages would have ressted such movement leadng to overload failures in the cable. As a consequence, it is likely that little significant slider movement occurred during the impact. Therefore, the ‘as-found’ positions of the elevator trm actuators were most probably representative of the pre-impact settings. iii) Engines The engnes had broken up to the extent that the gas-producer sections were exposed. Most of the blades 20 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 Bearing Housing Slide assembly Sprocket Retaining plate Shaft Dual trim tab actuator Bearing Dagram of elevator trm jack Rght elevator trm jack Left elevator trm jack Figure 1 Detals of ptch trm jack 2 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 n the axal compressors had been torn off n a manner that indicated high rpm at impact. It was not possible to quantfy the power settng from the condton of the compressors. However, the degree of damage was the same n the compressor assembles of both engnes, indicating a symmetrical power condition. The remans of the engne casngs, whch had been severely compressed n the mpact, were cut open to expose the turbine sections. Once again, the symmetrcal nature of the damage was apparent, both on the gas producer and free power turbne dscs. Many of the engne components and accessores were examned n the presence of a representatve from the engine manufacturer. The filter elements in the fuel pumps were clear, the pump gears were ntact and the fuel control unt (FCU) drve couplngs were undamaged. The FCU’s themselves were severely damaged, although nternal components such as daphragms had remaned ntact, and the daphragm chamber n the unt from the rght engne was stll primed with fuel. Both cam-box assembles were recovered but t was not possble to determne whch assembly related to each engine. It was noted that on one unit, the beta arm together wth ts assocated roller, was n the reverse-ptch portion of the cam slot. Additionally, the locking wire was mssng from the pnch bolt, whch clamped the arm onto its splined shaft. The torque necessary to turn the pnch bolt, n a tghtenng drecton, was measured usng a torque wrench and was found to be around 5 to 8 lbf in. As a comparison, the locking wire was removed from the bolt on the other unt and the tghtenng torque was found to be around 40 lbf in. The Maintenance Manual Footnote  Translates power lever movement to the fuel control unt and the propeller control unt figure was 32 to 36 lbf in. Also the splines beneath the pnch bolt wth the mssng lockng wre were damaged to the extent that they had a worn appearance. It was not possble to determne whether ths was caused before or during ground impact. The ‘as-found’ torque value on the pinch bolt, at around half the specified figure, could not be descrbed as excessvely low, but t dd rase the possblty of a potental loss of synchronsaton, due to slppage of the lever on the shaft, between the power lever in the cockpit and the propeller pitch control. iv) Propellers and their control systems All sx propeller blade roots were found scattered around the accident site because the hubs had shattered on impact. All the blades were recovered wth the excepton of one outer secton, and all had suffered consderable leadng edge damage. The fracture face on the blade fragment, adjacent to the mssng secton, was ndcatve of an overload failure on impact. Although it was not possible to determne from whch propeller assembly some of the blades originated. The similarity of the damage to them all suggested a symmetrcal power condton, or at least a similar rpm, at impact. The propeller control units were identified but they were n such a severely damaged condton that they could not be tested. However, internal examination of the governors ndcated no evdence of pre-mpact mechancal falures and there were no flyweight contact marks on the nternal surfaces of the governor housngs that mght have indicated an overspeed condition. However, no significant pieces of the overspeed governors were found that could have confirmed this finding. In many accdents t s possble to determne a propeller ptch angle at mpact by establshng, wth the ad of wtness marks, the poston of the ptch change mechansm relative to an internal piston. Alternatively, a similar 22 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 process can be used to establsh the angular poston of each blade root relatve to the “spder” porton of the hub in which the blades are located. In this accident, the degree of fragmentaton was such that these methods were not available. However, portions of the feathering sprngs were recovered, together wth fragments of the steel tubes in which they had been located. It was found that areas of the nternal bores of the tubes showed evdence of ndentatons made by the ndvdual sprng coils during the impact. The average spacing between the col mprnts can vary accordng to the fore-aft poston of the dome, whch n turn s a functon of the propeller blade angle. The imprints were measured (see Figure 2), whch revealed that the spacngs were the same for both tubes, ndcatng that the left and rght propeller angles were very similar. Using the measured spacing of 8.33 mm, the propeller manufacturer was asked to determne the corresponding blade angle. Figure 2 Remans of featherng sprngs, showng col mprnts on tube bores 23 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 The manufacturer was also asked to calculate blade angles at the estmated mpact speed of 350 kt at both maximum engine power and flight idle engine power, at the temperature and altitude of the accident site. The assumed propeller speed was 1,650 rpm in all cases. The calculations yielded the following information: at flight idle power the blade angle should have been 48.7° and at maximum power the angle should have been 53.4°. The ‘as-found’ blade angle, for both propellers was 55.2°. It was stated that the blade angle would increase by approximately 2.7° for every 50 kt ncrease n arspeed, wth temperature and alttude changes resultng in comparatively smaller blade angle changes. The manufacturer addtonally stated that the propeller blade angle range went from 88.5° at the feathered position to -13.5° at full reverse, gvng a total angular range of 02°. An intact feathering spring has 25 coils and the amount of dome (and hence sprng) movement per degree of blade angle change was given as 0.7112 mm. Because there are 24 gaps between the 25 cols, ths corresponds to a change in the coil pitch of 0.0296 mm per degree, whch llustrates how the blade angle s hghly sensitive to changes in the coil spacing. Put another way, if the 8.33 mm measurement was subject to an error of ± 5% (ether through measurng error or movement at mpact), then the derved mpact blade angle would be subject to an error range of ± 4° or so. Thus, while it would be temptng to conclude from the apparent mpact propeller blade angle of 55.2° that the arcraft struck the ground wth the engnes at hgh power and at a speed n excess of 350 kt, the possble error range could also encompass a low power condton, albet at blade angles above the beta range. In addition, the scope for spring movement caused by the impact cannot be quantified except that t s lkely to be less for a steep, fast mpact compared to a shallow, slow impact. On the other hand, f movement dd occur, there would be no reason why t should be the same for both propeller hubs. The fact that the sprng col ptch was the same for both propellers gives some confidence to the deduction that they reasonably represented the pre-impact settings. The beta feedback lnkages were recovered from both engines, although the carbon blocks were missing. The blocks had each been mounted n a ‘horseshoe’ shaped bracket, whch n turn was attached to a pn that was located n a hole n the feedback arm and secured by means of a circlip. The twisted remains of the pin were stll attached to the end of the rght engne feedback arm. However, there was no sign of the pin from the left engne feedback arm and the locaton hole was noted to be in pristine condition. This absence of damage gave rse to the possblty of a pre-mpact dsconnect, due, perhaps, to the pn detachng from ts horseshoe bracket. According to both the engine manufacturer and the propeller manufacturer, n ths eventualty, a sprng n the beta valve housng would act to push the (now unrestraned) feedback arm forward, allowng the valve to port ol away from the propeller dome, thus feathering the propeller. From the analysis of the featherng sprng marks, descrbed earler, t s clear that this did not occur. Examnaton of an ntact engne revealed that even f the crclp somehow became removed from ts groove n the end of the pn, the provson of a gude pn mounted on the engne casng would prevent the feedback arm from lifting off the pin. Thus, in order for the feedback arm to become free, the pin itself would have to fail. This seemed unlkely, n vew of the fact that the jont would be subjected to low n-servce loads and also because of the consequence of the propeller being feathered. It was therefore concluded that the undamaged locatng hole n the left propeller beta feedback arm was the 24 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 result of a qurk of the mpact, n whch the pn was pushed cleanly out of the hole, due ether to removal of the circlip or failure of the pin itself. v) Autopilot The possblty of an autoplot malfuncton was consdered, whch, for example, mght have caused a sudden nose-down command that the plot was unable to oppose. The autoplot manufacturer’s orgnal Falure Mode Effect Analyss (FMEA) was obtaned durng the nvestgaton, and t contaned a number of potental falure condtons that would result n a sustaned control input in any of the axes. With regard to the pitch axs, many of these falures would cause the autoplot to dsengage when the ptch angle exceeded 2° up or down. However, in some failures the autopilot would not disengage, resulting in a ‘hardover’ condition. In these cases the FMEA stated that the system had been demonstrated to meet the Federal Avaton Administration (FAA) certification requirements in that the plot was able to overcome the servo motor force and hence retain control of the aircraft. The certification documentaton suppled by the manufacturer stated that, for the ptch, roll and yaw axes, the force levels had to be within 50 lbs, 30 lbs and 150 lbs respectively. Test flight measurements showed that the actual forces were 45 lbs, 25 lbs and 60 lbs. Although parts of the autoplot servos were recovered and identified, these yielded no useful information. The autoplot computer and other assocated electronc components had been destroyed n the mpact, and so could not be tested. However, the mode control panel was recovered in a relatively intact condition. Each of the push-button swtches contaned a capton segment, illuminated by light bulbs. These were examined under a mcroscope n an attempt to establsh f any of them were llumnated at mpact: all were found to have “cold” or unlit indications. Immediately before the accident, the arcraft had been followng a south-easterly course towards Inverness and t would have been standard practice to engage the autopilot in HDG (heading) mode. However, the aircraft was at an extreme attitude at impact and, even f the plot had not dsengaged the autoplot, t s probable that t would have dsengaged automatcally durng the descent as the ptch and roll angles exceeded the limits. vi) Miscellaneous items In addton to the lght bulbs from the autoplot mode control panel, the remans of the two adjacent warnng annunciator panels were recovered. Many of the warning segments were mssng but most of the mssng bulbs were found n the wreckage; however, t was not possble to establish which systems they belonged to. All the bulbs were examned under a mcroscope and all but two showed clear evidence of being OFF at impact. Some filament stretching was apparent on the remaining two bulbs. During a flight in a similar aircraft it was noted that n cruse condtons, no lghts were llumnated on the warnng panels apart from the ‘partcle separators’ caption. It was the normal practice of G-TWIG’s operators to leave the partcle separators, n the engne ntakes, n the ‘open’ poston so the lghts would have been illuminated. The engine air bleed valve regulators were found to be in the ‘open’ positions. The cockpt area had been extremely fragmented n the mpact and most of the swtches, controls and nstruments Footnote  When bulbs are illuminated, the heated filaments become extremely ductile and an impact can result in extensive filament stretching within the glass envelope. This feature can thus provide evidence that the bulb was lit at impact. 25 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 had been destroyed. For example, the face of one atttude ndcator was found, but there were no wtness marks that could have provided an impact indication. The brass rotors from two ar-drven gyros were found: one bore evdence of crcumferental scorng, ndcatng that t had been rotatng at mpact, when it would have come into contact with its casing. The other rotor had no crcumferental marks, although ths dd not necessarly suggest that t was statonary at impact. One gyro case was found; its internal surface had been heavily scored. It was not possible to identify whether these components orgnated from the atttude indicators or directional gyros. The drectonal ndcator from the captan’s sde was found in a relatively intact condition. The heading bug was postoned at 29°; the selected course towards Inverness. Calibration of the pitch trim system Because the ptch trm actuators were found n the full arcraft nose-down poston, t was decded to conduct an evaluation flight on a similar aircraft to assess the trm settngs for the same centre of gravty poston as the accident aircraft. Full nose-down pitch trim was appled wth the arcraft descendng through 8,000 ft at 205 KIAS. To prevent the aircraft’s nose dropping, a significant rearward force (about 30 to 45 lbf) had to be applied to the control yoke. This evaluation was somewhat subjectve but t demonstrated that control of the aircraft was manageable in this condition. Moreover, f the nose was allowed to drop, the arcraft could be recovered to a level atttude wth only one hand on the control yoke. The aircraft was then flown in several speed/attitude combnatons and, for each trmmed condton, the poston of the trm ndcator ponter was marked on an adjacent piece of adhesive tape. On the ground, the trim actuator extenson was measured for each of the marked postons and at the full nose-up and nose-down postons (although the aircraft was not flown at the full nose-up trim condition). The total linear travel of the actuator, which extended for nose-down trim, was 0.75 in from the nose-up to nose-down marks. With the aircraft in a cruse descent at 205 KIAS t was found that the actuator ram was 0.125 in away from the full nose-down position; n fact ths value was found to change lttle for the level flight condition. Also, during the evaluation flight, the rate of electrical trm operaton was notceably slower n comparson to typical manual operation of the trim wheel. Additional aircraft information The arcraft’s techncal log was recovered from the accident site. The pilot had calculated a takeoff weight of 6,787 lb. With the aircraft in the freight configuration, no cargo and only hmself on board, the centre of gravty would have been within the permitted range. It is estmated that at the tme the arcraft dsappeared from the radar screen, t had burned approxmately 200 lb of fuel and, consequently, weighed about 6,580 lb. At this weight, in a clean wing configuration and with the wings level, the aircraft’s stall speed would have been 83 KIAS. G-TWIG’s maximum take-off weight was 9,850 lb. At that weght and at sea level, the maxmum manoeuvrng speed is 162 KIAS. Abrupt control movements should not be made above that speed. The manufacturer’s Aeroplane Informaton Manual contans an emergency procedure for an Electric Elevator Trim Runaway. It states: 26 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 1. Control Wheel – OVERPOWER as required. 2. AP/TRIM Disconnect Switch – DISCONNECT immediately. 3. Manual Elevator Trim – AS REQUIRED. NOTE After the electric trim has been disconnected and the emergency is over, pull the electric trim (ELEV TRIM) circuit breaker. Do not attempt to use the electric elevator trim system until ground maintenance has been completed. There was also a note wthn Supplement A3 of G-TWIG’s Plot’s Manual whch stated that n the event of any Kng 275/325 autoplot malfuncton, the battery master switch may be turned off. No such note was included in the secton dealng wth emergency procedures for the Kng autoplots or n the Flght Manual Supplement for the Sperry 1000 A autopilot fitted to G-TWIG. Whle experence has shown that t s possble to control the arcraft at the maxmum operatng speed wth full nose down elevator trim, a definitive figure for the force required at the control column was not forthcoming. Cabn heatng s provded by dvertng hot compressor bleed ar from the engnes and mxng t wth cabn ar to obtain the desired temperature. This mixed air is also routed to the wndsheld defrostng and defoggng outlets. The flight load limitations for the aircraft at maximum gross weight with the flaps retracted are minus 1.44g to + 3.6g. With the flaps at the takeoff position, these limits are reduced to 0g and +2.0g. An exercse conducted n 2000 at the Internatonal Test Plots School, based at Woodford n the UK, examned the lateral and drectonal stablty and control characteristics of the F406. The report did not reveal any adverse handlng qualtes and the lowest score given by the pilot using the Cooper-Harper Handling Qualtes Ratng Scale, on a declnng scale from one to ten, was three. This equates to an aircraft characteristic for whch mnmal plot compensaton s demanded to acheve the desred performance n a selected task or required operation. This score was given by the testing plot when assessng the arcraft’s behavour whle mantanng 30º angle of bank turns to the rght and, secondly, when rollng out of rudder-free aleron-only turns. This reflected comments by other pilots, who have flown the F406, that the aircraft type, which had been in production for 19 years, did not possess any vices. It had been mentoned that the arcraft type s more responsve n ptch than t s n roll but ths was an observaton, not a criticism of the aircraft. Aircraft handling procedures For takeoff and clmb the propeller speeds are set to 1,900 rpm, the maximum. For the climb and cruise flight phases, the propeller speeds were normally reduced to 1600 rpm. The normal climb speed for the F406 is 140 kt. In the cruse, the Operatons Manual nstructs crews not to exceed the maxmum cruse torque shown n the Aeroplane Flight Manual. For the conditions estimated at FL95 on the accident flight, maximum cruise torque at a propeller speed of ,600 rpm should have gven an aircraft speed of 205.5 KIAS, equivalent to 234 kt true airspeed (KTAS). This compares with the aircraft’s normal cruse speed of between 200 and 205 KIAS and somewhat less than the arcraft’s maxmum operatng speed of 229 KIAS. During this phase of flight it was customary for the plot to engage the alttude and headng hold modes of the autopilot. 27 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 The operator’s Operatons Manual nstructs plots that ‘before visible moisture is encountered with an OAT between +4ºC and -30ºC’ they are to ‘ensure that all aircraft anti-icing systems are ON and operating.’ These ant-cng systems nclude ptot heat, stall vane heat, the engne ntake nertal separators, the propeller de-cng systems and the electrical windshield anti-ice systems. The Operatons Manual also provdes the followng gudance on the operaton of the arcraft de-cng system in flight: ‘Position de-icer switch to AUTO when ice has accumulated to a thickness of approximately half an inch on the leading edges. No adverse aerodynamic effect will be produced by the operation of the de-ice boots other than a slight increase in prestall buffet and speed ….. NOTE: Since wing and horizontal stabilizer de-icer boots alone do not provide adequate protection for the entire aircraft, known icing conditions should be avoided when possible. If icing is encountered, close attention should be given to the pitot static system, propellers, induction systems and other components subject to icing. The de-ice system will operate satisfactorily on either or both engines. During single-engine operation, suction to the gyros will drop momentarily during the boot inflation cycle.’ The arcraft Informaton Manual states that an ‘accumulation of a ½ inch of ice may cause a cruise speed reduction of up to 30 knots as well as a significant buffet and stall speed increase.’ Before commencng descent, t s lkely that the plot would have obtaned the latest meteorologcal nformaton for Inverness from the arport’s Automatc Terminal Information Service (ATIS). To initiate descent, the normal practce s for the plot to lower the nose of the arcraft by rotatng the ptch command wheel on the autoplot control panel, whch also disengages the altitude hold mode of the autopilot. Power is also reduced. Using this method, the pitch atttude change s proportonal to the amount of rotation of the pitch command wheel. If the aircraft’s ptch atttude had exceeded approxmately 20° up or down, a dsconnect functon should have automatcally disconnected the autopilot. The ptch command wheel sgnals operate through the autoplot servo actuator, whch drves the ptch control circuit. This is separate from the elevator trim control. An alternatve method of changng the ptch atttude s to depress the ptch synchronzaton button, located on the rght arm of the plot’s control wheel, and manually select a new ptch atttude, before releasng the button and allowing the autopilot to maintain that attitude. The pilot can also fly the aircraft manually by disengaging the autopilot. On ths company’s operatons t was typcal for the aircraft to descend at 220 KIAS. The Operations Manual advsed crews that: ‘crew and passenger comfort is aided by the avoidance of steep descents and rates of descent above 800 fpm should be avoided.’ The Informaton Manual explans that, f a baggage door s left unlatched, t may open as the nose of the arcraft is raised during takeoff. However, the door will not ht a propeller nor wll there be any unusual handlng characteristics. In such a situation the airspeed should be kept below 120 KIAS. 28 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 The operator’s plots receved recurrent tranng n techniques for recovery from unusual positions. Meteorological information Durng the nvestgaton a meteorologcal aftercast was obtaned for the area around the accdent ste on the morning of the crash. At 1000 hrs the synoptic stuaton showed an area of low pressure centred between the Shetland Islands and Norway, whch fed a light, unstable, north-westerly airflow over the route from Stornoway to Inverness. The weather was mainly cloudy with occasional showers. Surface visibility was 10 to 20 km reducing to 4,000 m in showers. A band of more persstent ran lay to the south of the route, algned west to east from Skye to Aberdeen. The cloud conssted of few/scattered stratus at ,200 to ,500 ft amsl, scattered/broken cumulus or strato-cumulus at 2,500 to 3,000 ft amsl and broken strato-cumulus with a base at 5,000 ft amsl. These layers may have ncreased n amount and extent over the mountains. Photographs taken by some holidaymakers on the day of the accdent, 5 nm to the south-east of the accdent ste, appear to show a cloudbase at about 2,500 ft amsl when compared wth the elevaton of the mountains in the pictures. These conditions were reflected in the meteorological observatons taken at Stornoway and Inverness arports around the time of the accident. Of the two, Inverness had the worse weather. It s possble that there was some dynamc turbulence over the tops of the mountans, as a result of the wnds and the extent of the hgh ground, and t s hghly lkely that there was some convective turbulence in the cloud. The freezng level was at about 5,000 ft amsl and arframe cng was consdered to be lkely n cloud above that level. The wind velocity at 5,000 ft amsl and at 10,000 feet amsl was 320º/20 kt. At 5,000 ft the air temperature was -0.3°C. and at 10,000 ft it was -9.4°C. The air pressure at mean sea level was 990 mb. The pilot of another aircraft, flying from Stornoway to Inverness about 25 mnutes astern of G-TWIG at FL75, stated that he had experenced smooth condtons and no icing during his flight. When he was established in the cruise at FL75, he recalled that he had been flying between layers of cloud. He estimated that there was a fairly dense layer of cloud between 500 ft and ,000 ft below hm and about 6 octas of cloud approxmately ,500 ft above him. He did not encounter any precipitation until he was overhead Inverness. Medical and pathological information The post mortem report concluded that there were no pathological findings to help determine the cause of the accdent and that the plot ded as a result of the multple injuries sustained in the accident. It was impossible to say whether the plot was conscous or unconscous n the period preceding the accident. There was no evidence of any underlyng dsease and toxcology analyss showed no abnormal indications. Recorded data The arcraft dd not carry any mandatory recordng devices and there was no requirement to do so. A GPS unt was found n the wreckage but t was of a type that does not record track information. The sources of event data avalable were recorded radar tracks from Stornoway and Tree radar heads, a report from a controller who was vewng the unrecorded radar returns from the Knloss and Lossemouth radar heads, and radio communication recordings. 29 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 Post-accdent poston data was taken from a GPS unt carred to the ste durng the nvestgaton to pnpont the impact location. In-flight GPS and radar recordings were taken from another aircraft flown in the area at a later date to evaluate the radar performance lmtatons in the area. Radar system characteristics In order to understand the analyss of the radar data used n ths nvestgaton a few of the basc system characteristics and limitations are given below. There are two types of radar system currently used for cvl aviation in the UK, primary and secondary radar. Radar heads have one or both of the prmary and secondary systems and both use rotating antennas. Primary radar sends out pulses and detects when one bounces back from an aircraft. Primary radar tracks provide slant range and bearing from the radar head only. Secondary radar sends pulses to a transcever on board the arcraft whch then responds wth an arcraft dentty code and additionally, if selected, the aircraft’s pressure altitude. Thus secondary radar tracks provde arcraft dentty and alttude as well as slant range and bearng; however, the aircraft equipment must be operational. Another lmtaton of secondary radar arcraft equpment s that on arcraft of ths sze, there s only one transponder antenna. This is installed on the bottom of the aircraft, provdng reasonable coverage durng manoeuvrng, but at more extreme atttudes t can cause loss of secondary radar sgnal dependng on the orentaton of the arcraft to the radar head. Other relevant radar characteristics are the lne of sght of the radar head to the arcraft and the resolution and accuracy of the radar track position. Radar needs drect lne of sght to an arcraft n order to detect it. High ground between the aircraft and the radar head nterrupts the passage of radar pulses and creates a radar shadow. This effect is exacerbated with dstance between the arcraft and radar head because of the curvature of the earth. Each radar poston does not represent a pont n the arspace but a volume of arspace whch for convenence may be visualised as a box with dimensions defined by the resoluton and accuracy of the range, bearng and altitude systems. The range and altitude sides reman farly constant wth regards to resoluton and the effects of errors. However, although the angular bearng resoluton s constant, the horzontal dstance (wdth) ths represents ncreases wth dstance from the radar head. In ths case, the resoluton of the recorded radar data was limited to 1/16 nm in range and 0.088° in bearing. These ncrements are qute large compared to the dstance travelled in the 8 seconds between each radar sweep. Thus the dstance travelled between each radar sweep s not a single value but a band of possible values. This resolution tolerance also affects speed and heading calculations. So, gven ths resoluton tolerance, determnng arcraft manoeuvres between ndvdual returns cannot be done in detail. Trending flight parameters over many sweeps during steady flight can be done with more accuracy because the band of possble values becomes smaller compared to the distance travelled. Radar altitude resoluton s always lmted to the 00 ft ntervals of the arcraft’s transponder resoluton whch provdes smlar limitations as per range and bearing. A further relevant lmtaton of secondary radar s that t rejects, and therefore does not track, secondary radar returns reportng an alttude change of ,000 ft or more since the last sweep. 30 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 Radar data derived flightpath The recorded radar tracks from Stornoway and Tree are gven n Fgure 3 together wth the type of radar, the locaton of the radar head, the advsory route beng flown and the accident site. The Tree radar tracks, whlst provdng both prmary and secondary radar returns, were fragmented due to shadowng by terran half way between the radar head and the flight path. Another problem with the Tiree data was that the forward moton of the arcraft was algned wth the bearng resoluton of the radar whch, at these distances, is very poor compared to the range resolution. However, this did make the Tiree source good for assessing the aircraft’s across-track motion. Figure 3 Geographical locations of the accident site, radar tracks, advisory route flown and relevant radar heads 3 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 The Stornoway radar provded contnuous secondary radar data whch covered all the Tree data tracks and more. The aircraft flew away from the Stornoway radar head and so ts forward moton was algned wth the ‘tighter’ range resolution of the radar. Therefore the Stornoway data was used for the general flight overview and speed calculations. Figure 4 shows these in detail. Figure 4b The Stornoway secondary radar track derved parameters Figure 4a The Stornoway secondary radar track wth reported alttude 32 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 The Tiree data correlated with the Stornoway data. The reported altitude was also verified by comparing the ntermttency of the Tree data wth the lne of sght lmts of the Tree radar head, gven the terran between the aircraft and the radar head. The track initiated at 1017 hrs at FL38. The aircraft clmbed to FL95 wth an average clmb rate of 1700 ft/min. During cruise the aircraft maintained a ground speed of 240 kt equatng to a true arspeed of 220 kt and an indicated airspeed of 192 kt. The aircraft tracked slghtly to the left of the centrelne of advsory route W6D. The aircraft was cleared to descend to FL75. The descent was initiated and averaged 750 ft/mn untl FL88 (approxmately 8,200 ft amsl) at which point the descent rate started to fluctuate, approxmately 50 seconds before the arcraft track was lost. Due to the coarse nature of the altitude data, it was difficult to determine the flight path between individual radar returns. However, the average descent rate between the last two recorded ponts was between ,500 ft/mn and 3,000 ft/min. The last radar point was at 1031 hrs wth the arcraft at FL78 whch was approxmately 7,200 ft amsl. Fgures 5 and 6 overlay both the Stornoway and Tree data to provide a more detailed profile of the aircraft’s flight path during the last portion of the flight. Figure 5 Overview of the final radar track points from Tiree and Stornoway against the impact site, impact orientation and local terrain. 33 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 Pertnent ponts to note from the radar tracks are: 1. Despite being vertically separated by nearly 5,000 ft, the arcraft mpact was wthn a few hundred metres of the final radar return. 2. The aircraft turned left relative to its previous flight path in the last few radar sweeps. 3. Reaching the impact point required a significant change n headng after the relatve moton of the last radar points. 4. None of the radar heads recorded, or were observed to dsplay, the arcraft after t descended through FL78 despte havng line of sight capabilities significantly below this level. 5. The Tiree secondary radar did not detect the arcraft at FL78 despte Stornoway secondary radar and Tiree primary radar detecting it. Also, the observer of the Knloss secondary radar dd not recall seeng any returns below FL81. Additional information No one saw the mpact and there were no mpact signatures recorded on seismographs. The pilot was 76 nches tall (6 ft 4ns) but hs seated heght was not determined. The maximum distance between the pilot’s seat cushon and a strnger supportng the cabn roof was 38 inches. The seated height of person of similar stature to the accdent plot was measured at 36 nches from the seat cushion (depressed) to the crown of his head). Figure 6 View of the final radar track points from Tiree and Stornoway, as vewed from a pont to the South of the accdent ste 34 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 Analysis Overview G-TWIG and ts plot both seemed to be operatng well until the fifth sector of the day when, shortly after starting descent for Inverness at about 85 KIAS, the arcraft’s rate of descent increased and it started to turn left. The aircraft struck the ground near the final radar return but almost 5,000 ft below t and on a headng at rght angles to its intended track. The available evidence indicated that the arcraft struck the ground n a steep spral dve to the left. The extreme fragmentation of the wreckage ndcated a hgh mpact speed, probably n the order of 350 kt. There were no radio messages from the pilot during the spiral dive. Radar data analysis The tme of ground mpact could not be establshed so analyss of the radar returns was the only method wth which to estimate the likely flight path and deduce whether the aircraft flew directly from the last radar return to the point of impact or whether it flew a more circuitous route. Loss of radar returns Gven the lne of sght the radar heads had n the area of the accdent, the radar tracks stop at a greater heght than expected. In order to explain the sudden cessation of radar returns, the last few recorded ponts of prmary and secondary radar are analysed separately. Primary radar The only source of recorded prmary radar was the from the Tiree radar head. This indicated that Tiree detected a primary return from the aircraft one sweep after the final secondary return at FL8 whch, gven the Stornoway secondary radar track, s lkely to have occurred at the time the aircraft was at approximately FL78. Tiree radar can ‘see’ down to at least 5,500 ft amsl at the accdent location. The lack of further primary radar returns ndcated that ether the arcraft atttude at the tme of the next sweep was such that it presented insufficient area to create a return, whch s unlkely, or that the arcraft had descended below the Tree lne of sght lmt n the 7.87 second interval between the sweeps. To descend from FL78 to 5,500 ft amsl in 7.87 seconds required a 1.2g downward acceleration (a person seated in the aircraft would experience -0.2g tending to lift them off their seat). This fact implies that the aircraft was providing a significant downward thrust. Secondary radar The first anomaly associated with the secondary radar data s that Stornoway was the only radar head to detect the aircraft at FL78. The explanations considered were as follows: 1. Random track drop. Radar occasionally drops aircraft tracks randomly. However, it s unlkely that two radars would randomly drop the track of the same aircraft. It is feasble that ths s a product of nterrogatng the arcraft at the exact same tme but ths s also unlikely. 2. Antenna obscured. The secondary radar loses track of the arcraft f t s at an extreme atttude wth the radar lookng at a transponder blnd spot above the arcraft or, when lookng drectly along the antenna axis from underneath the aircraft. Given that Knloss and Tree were lookng at the arcraft from postons approxmately 20º apart, t s unlkely that an extreme atttude could present the upper blnd spot to both radars at the same time. If one of the radars was lookng drectly along the antenna axs 35 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 from underneath, t s unlkely that the other radar would smultaneously be lookng at the transponder blind spot on top of the aircraft. 3. Transponder inoperative. Because their recorder clocks were not synchronsed, the relatve tmngs of the three radars sweepng the aircraft were unknown. It is possible that the transponder became noperatve just after the Stornoway detecton at FL78 and just prior to the Kinloss and Tiree radar sweeps. The noperatve state s unlkely to have been drectly lnked to the prmary causal factors of the accdent because the loss of arcraft trackng occurred after the arcraft departed from ts expected heading and altitude rate. However, the noperatve state could have been lnked to a cascade of falures or to acton as a result of dealng wth other factors, possbly leadng to the interruption of electric power. The second anomaly s the lack of secondary radar returns below FL78. Explanations considered are as follows: 4. Transponder inoperative (as above). 5. The aircraft’s descent rate was so high that it dd not pass the reasonableness check of the altitude rate by the radar head. (If the reported alttude of an arcraft changes by ,000 ft or more between consecutve sweeps the return s rejected and not transmtted to the control centre.) To meet this condition after the FL78 detecton would requre an average vertcal acceleraton to the mpact pont of approximately 0.7g or more (ie a person in the aircraft would experience +0.3g instead of the normal 1g). Whilst this does not require an acceleraton force greater than gravty, t does not preclude it. However, it does require that normal wng lft forces are drastcally reduced or no longer acting significantly upwards. Gven the physcal evdence of speed, ths would imply a significantly nose-down or nverted atttude, or an arframe dsrupton such that the wings no longer imparted lift. Potential explanations for the accident The evdence from the accdent ste ndcated that the arcraft had struck the ground n a steep, left wng low atttude, on a track some 90° to the rght of the track towards Inverness, at a speed well n excess of the maximum permitted. The most logical explanation for ts dsappearance from radar was a very hgh rate of descent. In attemptng to evaluate what mght have happened to nduce ths hgh-speed dve, three categores of causal factors were consdered: an arcraft defect, an environmental factor and a piloting factor. Aircraft defects There was no evidence of an in-flight fire or explosion. The possibility of an in-flight structural failure was elmnated by the fact that all the extremtes of the arcraft were accounted for and the wng was structurally intact at impact. However, it was not possible to be so certan about the forward baggage doors although, as a causal factor, the possblty of a door becomng detached, penetratng the wndscreen and ncapactatng the commander, seemed remote. The airspeeds probably acheved pror to mpact would have been well n excess of the maxmum permtted and the assocated control forces would also have been abnormally high. However, n the event that the commander was able to make a significant control input, it is probable that the aircraft would have suffered an in-flight structural failure. 36 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 The fragmented nature of the wreckage meant that t was difficult to establish with confidence the operating state of some of the aircraft systems. For example two gyroscope rotors were recovered; one bore evdence of circumferential s coring whilst the other did not. Thus the evdence that one of them was rotatng at the tme of the mpact, when t came nto volent contact wth ts casng, was countered by the absence of such evdence on the other. Whilst this was most probably an oddity of the mpact, t put n mnd at least the possblty of a falure of the pneumatc supply to one or all of the relevant instruments. If such an event occurred, in addton to presentng msleadng nformaton to the commander, t s lkely that the autoplot would make erroneous control inputs to the aircraft. For example, f the atttude ndcator drfted to the extent that t gave a false nose-up ndcaton, the autoplot would apply a nose down correcton, whch could result n an excessve rate of descent. If the aircraft was flying in IMC, then the commander mght not mmedately recognse that something was wrong. However, such a scenario would lkely result n a relatvely gradual departure from the intended flight path; the available evidence suggests a more dramatic event. Smlarly, t was not possble to establsh, wth certanty, that electrcal power was avalable on the arcraft, although the fact that the transponder was operatng during the early part of the descent suggests that it was. In any case, falure of the electrcal system would not logically be followed by a sudden loss of control. Investgaton of the propeller hub components led to the concluson that both propellers struck the ground at smlar blade ptch angles and, as a consequence, wth essentially symmetrical engine power applied. The nature of the evdence was such that the derved blade angles (approxmately 55° n both cases) were subject to potentially large errors. Whilst this reduces confidence in the arspeed calculatons, t at least suggests the engnes were developing a significant amount of power, rather than flight idle power. If the propeller blade angles were at 55°, the impact speed may have been close to 400 kt. Investgaton of the ptch trm system revealed that the elevator trm actuators were near ther fully nose- down postons whereas the approprate settng for the weight and balance conditions was 0.125 in from the fully nose-down position. There are only three possible reasons for the as-found postons of the actuators: the commander trmmed to ths poston; a fault n the electrc trm system caused an uncommanded trm nput; or there was a fault in the autopilot. There appears to be no logcal reason why the commander would trm to such a nose-down settng at the normal arspeed used in a descent. However, the as-found trim setting may have been appropriate to some higher airspeed. It was not possble to dscount an electrc trm system malfunction although flight tests indicated that the control forces could have been overcome wth lttle difficulty. Similarly, the most serious potential fault in the autoplot, a spurous nose-down nput followed by falure to dsengage automatcally when the ptch angle exceeded 2° nose-down, could not be discounted. If that had happened, the commander would have had to overcome the force of the servo motor n addton to the aerodynamic force. Whilst this force may have been significant, possibly in excess of 40 lbf, the commander would have had the opton of swtchng off the autoplot and manually re-trimming the aircraft. Switching off the autoplot va the electrcal master swtch mght explan why the arcraft’s secondary radar return was lost but t does not explan why only one more prmary return was received. Moreover, had the commander been combatng a run-away trm system, t seems lkely that he would also have reduced engne power and rolled the aircraft’s wings level to recover from a dive. 37 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 Environmental factors The arcraft was probably n cng condtons although it may not have been accreting ice. In those conditions the arcraft’s ant-cng systems should have been operatng and, f there was an ce buld up of between ¼ and ½ an nch on the leadng edges of the wngs, the commander should have been able to operate the de-icing boots without any adverse effect. He should also have been aware of the attendant warnngs n the Operations Manual. The reduction in aircraft speed that could accompany an ice build up may be reflected in the radar data f the commander had selected maxmum cruise power on the engines. There was no indication of any significant turbulence and the commander of another arcraft whch was followng the same route at FL75, some 25 mnutes astern of G-TWIG, reported experiencing smooth conditions. Moreover, there were no thunderstorms n the area whch mght have produced a lightning strike. Therefore, severe atmospheric conditions seem an unlikely explanation. Collson wth an object, perhaps one penetratng the wndscreen leadng to plot ncapactaton, was consdered but there was no evdence of any other ‘foregn’ objects, including birds, within the wreckage. AAIB experience ndcates that collson wth any szeable object leaves identifiable traces within the aircraft so this also seems an unlikely explanation. Piloting factors The commander was due to leave the company n just over a week’s time to join a larger short haul jet operator. In dong so, he would have been leavng behnd two and a half years of enjoyable flying on turboprop aircraft, operating passenger and freight flights on a regional network. At his request, he had changed the standby duty, for whch he was rostered on the date of the accdent, with the F406 five-sector duty that had been allocated to another pilot. In view of his comments that he might not enjoy such flying in the future, it is understandable that the commander mght have wshed to make the most of any remaining opportunities. The commander’s private lfe was happy and company staff at Stornoway descrbed him as being in his normal, jovial mood. They also remarked on his conscientious approach to his duties. There was no evdence n hs tranng records of any difficulties during his conversion or recurrent training and, by all accounts, he was fit and able, with an exciting future ahead of him. Equally, the aircraft type was not known to dsplay any characterstcs whch could place particular demands on a pilot. G-TWIG’s take off from Stornoway was unusual but the commander had flown a smlar manoeuvre at least once before wth no adverse effect on the aircraft. Also, it would not have been the first time that a pilot had performed an eye catching departure in an empty, light aircraft. Consequently, there was no reason why the commander mght have taken hs own lfe, ether delberately or nadvertently through some form of unauthorised manoeuvre. The clmb and subsequent cruse at FL95 seem to have been unremarkable and all the commander’s rado calls were lucid and calm. He did not transmit an emergency call and he gave no indication of any problems. He mssed one rado call towards the end of the cruse phase but ths may have been when the arcraft was n a known rado blnd spot or when he was lstenng to the Inverness ATIS frequency. His acknowledgement of the ATC clearance for the arcraft to descend from FL95 to FL75, his final radio call, was delivered in a clear, unhurried voice. The arcraft had returned from Stornoway ,000 ft above the level it had cruised at on the outbound leg. On both sectors the commander would have had the cabn 38 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 heating on. However, there was no evidence from the post mortem that the commander had been ncapactated by fumes. If the elevator trm had malfunctoned n the early stages of the descent, t would have been possble for the commander to overcome the nose-down trm forces; moreover, he could have stopped an electrc trm runaway by isolating electrical power to the trim motor. It is not known how manageable the control forces would have been at speeds above the maxmum permtted but the commander could have used the elevator trm wheel to assist with recovery from a high speed dive. If the arcraft’s atttude been dsturbed by an encounter wth localsed turbulence or vertcal wndshear, the pilot had sufficient skill and experience to recognise an ‘unusual poston’ and take the approprate recovery action. That would probably have been to throttle back both engnes, roll the wngs level and ease the arcraft out of its dive. However, both engines were still developing significant power at impact, the wings were not level and the dive angle was about 70°. These parameters were inconsistent with an attempted recovery. One plausble causal factor for ths accdent could be that the commander was affected by a sudden mental or physcal ncapactaton that manfested tself n involuntary movements. For instance, if the aircraft had entered a localsed vertcal ar current leadng to a negatve g excurson, even f hs seat harness was securely fastened, t s possble that ths unusually tall plot could have struck hs head on a hard strnger supportng the cabin roof about two inches above his head. He was almost certanly wearng a communcatons headset whch mght have gven some cushonng to the crown of hs head but a hard mpact on an unprotected regon of his skull could have been temporarily debilitating. A severe encounter could have rendered hm unconscous and f he started to regan conscousness, any nvoluntary arm and leg movements might have been sufficient to ‘upset’ the aircraft. Amongst other control inputs, nvoluntary movements mght explan why the electrc elevator trim operated to near its full nose-down extent. The commander was not heard to make any emergency radio call, although the frequency was briefly blocked after the arcraft had dsappeared from the radar screen, and there were no sgns that he was attemptng to recover from the steep, spiral dive. Conclusion Durng a gentle descent from FL95 to FL75 n nstrument meteorologcal condtons G-TWIG rapdly entered a dramatc and sustaned manoeuvre from what initially appeared to be controlled flight at normal descent speed. Despite a determined and thorough investigation, because there was insufficient evidence from which to draw a firm conclusion, the cause or causal factors for ths rapd devaton from controlled flight could not be identified. Safety Recommendations Internatonally agreed standards dd not requre G-TWIG to carry either a flight data recorder or a cockpit voice recorder but the nvestgaton of ths accdent would have been greatly enhanced if audio and basic flight parameter recordings had been available. For accdents where there has been extensve dsrupton of the arcraft, t may not be possble to determne the causal factors from wreckage analyss and wtness evidence alone. Yet with aircraft of G-TWIG’s weight category undertakng commercal ar transport, nstallng a traditional flight data recorder, with its array of remote sensors, would be mpractcal and economcally unacceptable. An alternative and potentially more 39 AAIB Bulletin: 7/2006 G-TWIG EW/C2004/10/04 practcal soluton would be to record the actvty of the pilot(s), flight controls, flight instruments and instrument panel selectors using imagery techniques. The addton of audo recordng to the mage recordng system would enhance the avalablty of evdence for accident and incident investigation. However, before approprate recordng equpment can be developed, a minimum performance specification must be developed. To that end, n the report on the accdent to G-BGED (AAIB Bulletn /2005) the AAIB made the followng recommendaton: Safety Recommendation 2005-062 It is recommended that the European Aviation Safety Agency [EASA] develop standards for appropriate recording equipment that can be practically implemented on small aircraft.’ Also, two safety recommendatons, 2004-084 and 2004-085, were made as a result of the nvestgaton nto the accdent to helcopter G-CSPJ (AAIB Bulletn /2005), and these are reproduced below: ‘Safety Recommendation 2004-084 The Department for Transport should urge the International Civil Aviation Organisation (ICAO) to promote the safety benefits of fitting, as a minimum, cockpit voice recording equipment to all aircraft operating with a Certificate of Airworthiness in the Commercial Air Transport category, regardless of weight or age.’ ‘Safety Recommendation 2004-085 The Department for Transport should urge the International Civil Aviation Organisation (ICAO) to promote research into the design and development of inexpensive, lightweight, airborne flight data and voice recording equipment.’ In a letter to the AAIB, dated 4 October 2004, the Department for Transport gave ts full support to these recommendations. Wth EASA assumng responsblty for matters of arworthness wthn the European Communty, the followng two recommendatons were made n the G-BXLI report (AAIB Bulletn /2006): ‘Safety Recommendation 2005-100 The European Aviation Safety Agency should promote research into the design and development of inexpensive, lightweight, airborne flight data and voice recording equipment.’ ‘Safety Recommendation 2005-101 The European Aviation Safety Agency should promote the safety benefits of fitting, as a minimum, cockpit voice recording equipment to all aircraft operated for the purpose of commercial air transport, regardless of weight or age.’ Recommendatons 2005-00 and 2005-0 are appropriate to this accident. As yet, no response to these recommendations has been received from the EASA.