Features - Technical

NOVEMBER 7, 1997

1997 Formula 1 Technical Review

BY PETER WRIGHT

The 1997 Formula 1 season will be remembered for the controversy over the final race to determine the Drivers' World Championship. However, this will be to devalue the year as not only being one of the most competitive and exciting for some time, but also one full of technical interest, although much of it was well hidden.


The 1997 Formula 1 season will be remembered for the controversy over the final race to determine the Drivers' World Championship. However, this will be to devalue the year as not only being one of the most competitive and exciting for some time, but also one full of technical interest, although much of it was well hidden. Next year will see radical changes, as the narrow cars and grooved tyres are introduced to control speeds, and a broad sweep of restrictions on the application of control systems for drive-by-wire throttles, differentials and brake balance systems, brings these systems back to what was intended.

This year has seen a huge leap in lap speeds, as the effects of a tyre war on tread compounds has taken effect. In terms of technical significance measured by the stop watch, nothing else comes close to the tyres. 100 bhp (the claimed difference between Tyrrell's Ford V-8 and the top engines) is worth 2-3 seconds per lap, while the 1997 tyres are around twice that time quicker than last year's tyres. It is a pity we cannot find out the details of construction and compound behind these advances, but judging from the number of blistered Goodyear tyres, they have sometimes had to go right to the edge to stay ahead of Bridgestone.

Like tyres, we are not able to find out much about Formula 1 engines - they too are the products of major motor industry companies who are skilled in preventing intellectual property from escaping their organisations. The latest 3-litre engines are small, light and high revving and, as can be seen by the embarrassing Ford, Illmor and Mugen-Honda blow-ups, they too are right at the edge. Weighing round 100 kilograms and revving to 18,000 RPM to produce over 750 bhp, the structural, thermal and lubrication margins are small. More and more they rely on sophisticated control systems to ensure that the temperatures, pressures and flow rates remain within tight tolerances. Everything that can be precisely controlled by a computer - ignition, fuel temperature and flow, air flow and intake lengths, valve timing etc. - has been investigated in the search for power and driveability.

In contrast, aerodynamics, the third of the big three performance parameters, are exposed to the eyes of all, but understood by very few. The front wing is still the dominant component in the aerodynamic package, as evidenced by the effect of new front wings on the performance of the McLaren and Ferrari. When Adrian Newey fitted a "Williams-style" front wing to the McLarens, and the Ferraris gained a "Benetton-style" flexible front wing towards the end of the year, much of their sudden improvement in competitiveness was attributed to these changes. Apart from these developments it has not been a vintage year for aerodynamics, the main excitement being generated by some of Tyrrell's attempts to gain more down force to make up for a lack of power.

With the performance of the top teams so equal, the quest to optimise every aspect of the cars has been intense. Any possibility of controlling the characteristics of a sub-system has been exploited to the full and the extent to which such control systems are within the Technical Regulations has been subject to greater pressure than ever before, with the FIA's software inspectors having a busy year checking the legality of the systems. The three main systems to have been developed by the leading teams are:

Brake balance:

I have described in a previous article the why's and wherefore's of fore and aft brake balance systems. However, since that time it has transpired that both Williams and McLaren have been experimenting with lateral balance control systems. McLaren have been using a car fitted with an additional pedal to the left of the brake pedal (in the position where the clutch pedal would normally be fitted). It has been suggested that this pedal applies braking to the rear wheels only, and that it may be possible for the driver to select just one rear wheel, either left or right or both together. The driver can use it as a means of steering the car by applying the brake to one or other rear wheel as he enters a corner, either with power applied or not, helping to reduce understeer. Of course, when this system is used, the differential must be "open" so that it does not fight the brake - see later. The other potential use is to control rear wheel torque with minimum lag, which would normally occur when using the throttle to do so, thereby providing the driver with a precise method of controlling wheel spin without infringing the traction control ban - very neat!

It is not known how Williams' system works, but it is believed to have a similar effect, but using just one brake pedal, the mode being selected by the driver.

The main problem with both balance systems is what to use to control balance. In the latter systems, only the driver can determine which rear brake is the appropriate one to apply, as steering and lateral acceleration feedback are banned. He controls the switch. With fore and aft balance, it would be nice to use either speed or longitudinal acceleration to control the percentage braking front and rear, but again it is not permitted. It is likely that brake pedal load or line pressure are used as this is the best estimate for the maximum possible brake torque, which is itself determined by speed. FIGURE: 1 shows a possible relationship.

Drive-by-wire:

Around the middle of this year there occurred a quiet revolution that passed with little comment until the last few races. And yet it embodied one of the most significant technical developments to take place since the great control system ban, at the end of 1993.

One technology that survived that ban was drive-by-wire, but the control laws and permitted parameters used for feedback were severely limited in order to ensure that it was not used for traction control. As the procedures for checking software matured and confidence was gained in the feasibility of policing these regulations, the FIA began to consider easing the restrictions to permit developments that were in line with road car development, but did not take away driver involvement in car control.

One of the prime purposes of road car drive-by-wire systems is to refine the process by which the driver demands torque from the engine, and the engine, along with its EMS, delivers it in the most efficient and environmentally clean manner. Mechanically-linked throttles can only control the air flow to the engine at any given moment, while the EMS strives to provide optimum fuelling. Even so, the driver needs skill to modulate the throttle pedal to control the engine output to his needs. Drive-by-wire turns the throttle pedal into a true torque demand - a fixed throttle pedal gives a constant torque output, irrespective of engine RPM, temperature etc, within the overall capabilities of the engine. To do so, the control system must include engine RPM as one of the feedbacks.

Until the middle of this year engine RPM was one of the banned parameters in Formula 1 drive-by-wire algorithms, because of the fear that it could be used as a surrogate for rear wheel speed in a traction control algorithm. Then it was changed so that it could be used as a feedback, provided it was only used to adjust throttle opening according to an engine map, and in no way could it be interpreted that it was being used to modulate the engine in response to wheel speed or wheel acceleration.

What advantages does this concession give? The answer is twofold: there are significant driveability advantages; and there is potential for tuning the engine for higher peak power.

First it is necessary to understand the driveability problem. A racing engine is tuned, using intake and exhaust lengths and valve timing, to give maximum torque over a working RPM range. With the current Formula 1 engines/transmissions, which have a maximum RPM of around 18,000 and 6 or 7 gears, this range is around 3,000 RPM. The torque curve approximates to FIGURE 2. Below the working RPM range, the engine is "out of tune" and torque is significantly lower, transitioning into tune over some quite small, RPM range. Logically, the driver would never use a gear that resulted in an RPM below the lower limit of the working range, in order to maximise acceleration at all times, modulated only by the throttle pedal to control wheelspin and handling. However the range of throttle pedal movement available to control torque from zero to maximum (actually less than zero, as closed throttle gives negative torque due to engine drag) is so small that the sensitivity, or gain of the throttle as a control input is too great for most drivers to be able to control handling. To overcome this problem they select a lower gear than optimum for a corner, and force the engine down into the "out of tune" region. Because the torque is so much lower, the throttle pedal gain is also much reduced. Now the driver can control the car. Once full throttle can be applied coming out of the corner, the engine accelerates so quickly in the lower gears, that it is soon up into the optimum RPM range and little time is lost at less than maximum acceleration.

However, if while on part-throttle, trying to control rear wheel grip, the RPM rises into the steep transition region, torque can increase so quickly that it catches the driver out and spins him, particularly on a wet track. This is the region engine tuners try to shape in such a way that the transition is as smooth and gradual as possible. Non-linear, drive-by-wire or mechanical linkages that reduce throttle pedal gain for the initial movement of the pedal, help the driveability. A typical characteristic is shown in FIGURE:3.

Feed back RPM into the drive-by-wire control algorithm and suddenly the problem goes away. Now the throttles can be progressively closed by the computer as RPM rises to give a constant torque, irrespective of RPM, for any given pedal position. If the driver puts the pedal in position "A", in FIGURE:3, wanting around 1/3rd full torque, the drive-by-wire throttle system will control the engine throttles with a characteristic that looks something like an inverse torque curve - FIGURE:4. The engine can be run in the transition RPM range, and precise throttle control used to give any torque up to the maximum. No longer will there be a sudden torque increase as the engine starts to come into tune, and by the time full torque is demanded, the engine can be in the working RPM range and ready to deliver it.

Not only is the driver happy, but the engine tuner can discard all the compromises he has had to make to smooth and fill the transition region, at the expense of tuning the engine for maximum torque. He can now fully optimise the working range - the width of which is only governed by the number of gears available.

Transmissions - differentials:

The section of the transmission that has received the most attention this year is the differential. In order to understand why so much effort has gone into putting it's operation under the control of the chassis computer, it is first necessary to understand what the differential can do to the car's dynamics.

It was not long after man decided to propel a vehicle by driving two, laterally spaced wheels mounted on a common axle, that he discovered the need for a mechanism to allow the wheels to rotate at different speeds, as they followed different paths around a corner, and yet both transmit torque to the ground. The differential was invented in 1827 by Pecquer, for use on steam powered traction engines.

The open differential delivers equal torque to each wheel, irrespective of each wheel's RPM. In a corner, when load is transferred from the inner to the outer wheel, in order to maintain equal torque the slip ratio of the inner tyre must increase to compensate for decreasing vertical load. Eventually the slip ratio exceeds that for peak longitudinal forces, and the wheel spins. At that point the differential ensures that the torque transmitted to the outer wheel equals that of the inner, i.e. low, and any increase in acceleration becomes limited.

If a resistive torque is applied to the inner, limiting wheel, the differential must transmit more torque to that wheel and hence will also transmit more to the outer wheel. The inner tyre is not able to transmit more thrust, but the outer one can and so the sum of the two will be greater. The resistive torque can either be applied via the brakes (as in "fiddle-brakes" on trials cars and the latest types of traction control on road cars that use the ABS system to differentially apply the brakes) or as a friction torque, generated between the two output shafts of the differential. All differentials have some degree of internal friction and are therefore not truly " open", but it is the intentional application of friction that is the basis of all limited-slip differentials.

The characteristics of a limited-slip differential depend upon how the friction is generated and how it changes with such factors as differential wheel speed, total torque, differential torque, temperature etc. A wide variety of differentials have been developed over the years, each with a different set of characteristics that are determined by the mechanical arrangement. As such, they tend to have fixed characteristics, be sensitive to manufacturing tolerances and wear, and to sometimes have undesirable side effects.

I would refer the reader, once again, to "Race Car Vehicle Dynamics" by Milliken and Milliken for descriptions and analyses of the small number of differential types that have stood the test of time and are still in popular use for racing cars and high performance road cars. It is necessary to understand these devices and the differences and limitations of their characteristics, prior to considering what electronic control can do for them. This is particularly important in Formula 1, where the limitation is imposed that the differential cannot be active, but can only emulate a mechanical device. No such restriction exists for World Rally Cars, which are free to use complex, active-feedback differentials in the front, rear and middle of their 4WD systems.

Current mechanical limited-slip differentials do a pretty good job of controlling single wheelspin on a race car, even in the wet. So why go to all the complication and potential reliability of an electro-hydraulically controlled device? The answer lies in the differential's ability to steer the car by controlling differential torque across the axle. The difference in thrust at the wheel, side to side, generates a moment about the Centre of Gravity that can be large enough to significantly affect the stability and control of the car. Among the first to recognise and exploit this were the cinder track and midget racers in the USA. Because of the limited grip, and the fact that the corners all go one way, the cars are fitted with locked differentials or "spools" (actually, no differential at all) and different diameter rear tyres (stagger). This generates a turning moment into the corner and overcomes the inherent power understeer of a short wheelbase, front-engined car. The technique has been further refined by oval track cars and was even used in Formula 1 in the 1970's, when cross-ply tyres that varied in static and dynamic radius were used, albeit with limited-slip differentials.

Depending upon the torque distribution either under power, or on the overrun during braking, a limited-slip differential will either resist yawing or introduce a turning moment in addition to any steering input. Under hard braking when the car is unstable, a yawing resistance is a stabilising moment and hence desirable. Salisbury differentials can achieve this with suitable ramp angles and pre-load, but the effect is quite small. Strong lock-up under power will have the same affect, but in this case it can be undesirable as it adds to power-on understeer. What is desired is a system that provides just the right torque distribution, and hence yawing moment, to meet the requirements of the driver and the response of the car at any given moment.

In an unregulated world a fully active system would be used, sensing and feeding-back all driver inputs (throttle, brakes and steering), car response (speed, lateral and longitudinal acceleration and yaw acceleration) and rear wheel response (slip ratio, vertical load and torque) to determine the optimum torque distribution for the rear wheels. The differential mechanism alone, however complex the internal friction control is, would not be sufficient to ensure that the torque could be varied from a 0/100% to 100/0% split across the axle. The rear brakes, differentially applied, would also be needed. Of course, in Formula 1 none of this is permitted.

What is allowed is an electronically-controlled differential whose characteristics emulate a mechanical system. Making it electro-hydraulic allows the desired characteristics to be very closely controlled and easily set-up, and permits the engineer or driver to select from a library of characteristics. In practice a Salisbury type limited-slip differential is fitted, which hydraulically loads the clutch pack to modify the internal friction between the two output shafts. (Note: my article on gearboxes had a cross-sectional drawing that showed an hydraulically controlled differential.) The hydraulic pressure is modulated by a Moog valve, controlled by the chassis computer. Exactly what the control algorithms are only the individual teams know, but the options are limited due to the need to copy a mechanical analogue. Some software engineers may have vivid imaginations when it comes to describing how a mechanical differential might work, but in reality there are really only two basic types: those that apportion torque according to the total torque, and those that do so according to differential wheel speed. To emulate these functions in software therefore, requires a means of determining input torque and differential wheel speed.

Differential wheel speed is straight forward, providing it can be shown to the FIA's software inspectorate that it is differential wheel speed that is being used, not mean wheel speed - that would look too like an element of a traction control system.

Torque measurement is much harder. A reliable, race-worthy system for torque measurement on a racing car has yet to be developed, though systems are used in testing on both the gearbox input-shaft and the half-shafts. That leaves the necessity to calculate engine output torque from all the engine parameters that influence it, and derive the torque from a map of engine performance that has been calibrated on the dynamometer, and then multiply it by the gear-ratio to determine final drive input torque.

The differential clutch pack load is calculated according to a function of input torque, differential wheel speed or both (mechanically possible, but practically a bit of a contraption). The functions must be fixed around the circuit, variable only by the driver selecting a different set of characteristics. As such they will be less than ideal at all times, but should be more optimised than a purely mechanical set-up.

Electronically-controlled differentials in Formula 1 are not something that is going to dramatically reduce lap times or produce a discernible difference in the way the cars are driven. More stability under braking and an increased ability to turn in under power, are the potential benefits.

The rest of the transmission hardware in Formula 1 has, at least temporarily, reached a design plateau. Such is the level of development that there are one or two layouts that provide the chassis designer with some degree of freedom, without making a significant difference to the functioning and performance of the transmission. That is not to say that the continuous process of refinement has ceased, indeed numerous small gains in weight, efficiency, reliability and gear-change times that are won in the laboratory, eventually come together into a measurable gain on the stopwatch. At the same time, the transmission offers control and software engineers their greatest opportunity to show their skills, with the operation of clutch, gear-change and differential all firmly in their hands.

With Williams, Ferrari and Prost retaining transverse layouts, while all the other contenders have returned to a longitudinal arrangement, there is no consensus on which is the optimum for the current cars. Prost still uses the ex-Benetton gearbox, so it maybe expediency that dictates their choice. Ferrari's CFRP/Titanium/CFRP sandwich favours the transverse layout as this concentrates the hot, highly stressed parts of the casing, that cannot yet be in CFRP, into the smallest volume. It is questionable whether the stiffness benefits of carbon fibre and titanium compared to magnesium, are realised in a design that necessitates extra casing joints. Ferrari have overcome their initial problems with cracking around the suspension pickups and the unit appears very reliable now. How the specific stiffness of the whole (stiffness per unit weight) compares with Williams' box is an unanswered question. Certainly no other team has chosen to follow in Ferrari's footsteps down this exotic and costly path.

Just why Williams have retained their transverse layout is their secret. Maybe it is because there is no real advantage in a longitudinal's weight or efficiency. Maybe they know something about weight distribution that the others do not know, and do not wish to take advantage of pushing the engine and gear masses forward. Ratio changing on a longitudinal requires the box to be split from the engine and inevitably takes longer than on a transverse layout. It could be simply that, with the gearbox still the Achilles' heel of the car (e.g. Villeneuve at Imola), they would rather stay with the devil they know.

Gearbox R&D efforts are focused on reliability, design for lower weight and transmission efficiency. All three come from very careful detail design, specification and quality control. Material and heat treatment specifications have to be more closely controlled to exploit further material properties, reducing weight without compromising durability. Bearing, shaft and gear tolerances, allied with precise control of lubrication, can yield efficiency gains and lower running temperatures.

Nearly all the teams use Xtrac for the manufacture of key gear and shaft components. The exception is Ferrari, who source all their components in Italy. Jordan splits their suppliers between Xtrac and Hewland. Williams work closely with their technical partner Komatsu, for the manufacture of final drive gears, tapping into the specialist steel manufacturing skills for which Japan is renowned.

The choice of fitting 6 or 7 forward ratios is provided for in some of the gearboxes, notably Benetton and Jordan. The decision on how many to actually fit for a given event, is mainly a function of matching engine characteristics to the circuit and the driver's preferred style. With gear-change times down to around 2 hundredths of a second, any development to reduce them yields a diminishing return in exchange for potential greater risks of a miss-shift.

The ability of the carbon clutch to continue to shrink in diameter and weight, whilst transmitting ever greater power, still amazes me. The current 4.0 inch diameter AP Formula 1 unit must be getting near the limit, and it is unlikely that the engine designers will be able to lower the crankshaft centreline much more. The tiny clutch would not look out of place on a 250cc motorcycle engine. Clutch duty cycles have been made slightly easier with the advent of computer controlled gear changing, where the clutch is not operated at all during up-changes, and is controlled precisely by the computer on down-changes. It is the start that takes clutches to their limits of temperature, and they are still expected to accommodate the bulk of the slipping needed to maintain engine RPM when leaving the start line. Tyres lose grip if slipped too much while the carbon plates do not, unless their surface temperature becomes too high. Ultimately it is the thermal capacity of the clutch that will determine where the size limit lies.

New for this year has been the ability to fit an Anti-stall system, taking over control of the clutch to prevent the engine stalling if the driver spins the car.

Suspension:

Tyres-suspension-aerodynamics-chassis: one integrated system, made up of highly inter-dependent parts, that determines the performance of any racing car except when maximum power or maximum braking is applied. If one part is changed, the effect ripples around the others until a new order is arrived at, where the whole is back in optimised harmony. This year a major player - the tyres - have undergone substantial changes, while the other parts struggled to keep up.

The suspension determines the attitude at which the tyre is presented to the track and the vertical load on it, both quasi-steady-state and oscillating due to track undulations. It also feeds back to the chassis the loads that the tyres generate in response. In turn the chassis responds, taking up an attitude that in turn determines the aerodynamic loads on it and alters the attitude and loads on the tyres. Both messages and responses, from chassis to tyres, go by way of the suspension and are shaped by it.

The tyre war between Goodyear and Bridgestone has changed both the construction of the tyres and the forces that they are able to generate, through the development of tread compounds. Tyre characteristics and their interaction with the dynamics and aerodynamics of a car are not a precise enough science, even in Formula 1 to look at the characteristics of a new tyre and dial straight in the appropriate changes to the suspension etc. that are needed to re-optimise the whole system. Instead the engineers must dig deep into their boxes of tools and go round and round the loop of changes until they find out just what the new tyre wants, and tune out the negative side effects of those primary changes. If they are lucky, the changes will be within the adjustment range of the suspension geometry and spring and damper settings. If not it may require a whole new geometry (new pickup points in the monocoque or on the gearbox, or new uprights) or, worse still, a different weight distribution or aerodynamics.

In a previous article I described how suspension engineers were equipping themselves with the option to run a third spring and/or damper on either axle, in order to separate out roll characteristics from those controlling vertical motion. Behind this trend there is a more fundamental suspension/tyre issue that really came to a head when Jacques Villeneuve joined Williams last year. Since running active suspension, Williams had been pursuing a policy of developing suspension and aerodynamics to allow the suspension to be as soft as possible for mechanical grip. Active suspension had allowed the compromise between suspension stiffness and controlling the ride height to be resolved in the favour of both features. Since it's ban in 1993, they have worked hard to ensure that the aerodynamic demands did not limit how soft the suspension was set up. In early 1994, when Senna first drove for them, the FW16 initially possessed less than ideal aerodynamics, preventing Senna from dominating in the way that people had expected. The problem was identified and resolved and, guided by the smooth driving style of Damon Hill, they have set the standards for both high and low speed grip since that time.

For Villeneuve ultimate grip is less important than the ability to place the car exactly where he wants it, when he wants to. For this he needs a highly reactive car, and that means one that has stiff suspension. It is a question of achieving a balance between the greatest potential performance and realising as much of that potential as possible. Conventional theory indicates that ultimate grip comes from varying the load on the tyre as little as possible i.e. a softly sprung and optimally damped suspension. If the oscillating load has an amplitude equal to the steady vertical load on the tyre, the load will be zero, once per cycle. At that point the tyre cannot grip and will slide. The analysis becomes very complex at this stage, but in essence the conditions at the contact patch are "stick-slip", rather than steady "sticking". Oscillating loads of the same magnitude as the steady load are common on a stiffly sprung car at low speeds, and this condition usually appears to reduce the grip.

However, there are characteristics of the tyre that under certain conditions prefer the "stick-slip" situation. For a start, it heats up the tyre faster, and when optimum tyre temperature is not attainable, may raise the temperature and so achieve higher grip. For Qualifying, when overheating the tyres is not a terminal problem, many drivers stiffen the suspension compared to race settings. Forcing a race compound to behave like a qualifying one by treating it harshly to achieve optimum temperature, and a more responsive car, are both desirable for a single quick lap. For the race however, drivers tend to adjust to a smoother driving style, and a set-up that looks after the tyres on a car that is heavy with fuel, is of paramount importance.

There are two aspects of the tyre war that emphasise these differences. Firstly, the compounds being developed are more marginal, and so need careful treatment during the race. Secondly, with two types of rubber from different manufacturers, scattered in small lumps all over the track, there is a real problem with tyres picking up this incompatible debris and ceasing to perform as advertised. A stiff car, which is more aggressive towards it's tyres, seems to have less problem in scrubbing off any pickup. A typical case of real life interfering with theory, and one that keeps the race engineers busy during the free practice periods, changing springs and damper settings to suit track, tyres and driver.

Most of the top teams develop their own dampers, working with damper manufacturers to make the more specialised cylinder and valve components. Because there is still considerable mystique surrounding damper technology, details of what goes on inside them is well guarded. Apart from the introduction of the third damper, there are two main areas of damper R&D. The first is to develop a development tool or procedure, to assist in the quick setting up of dampers and their associated springs, for a given circuit. The rate at which circuits are being altered and re-surfaced these days means that databases, gleaned in the preceding years are not much help. After the 50% wind tunnel, the next "must have" is a 4-post, electro-hydraulic, road simulation rig. With the car placed on it, and aerodynamic loads imposed via two further servo-actuators, it is possible to excite the suspension with an actual track profile, at any simulated speed. The ability to measure any parameter and study particular conditions in the laboratory is a valuable tool in understanding suspension dynamics and validating models. In particular, it can be used to study the second issue:

Without active control, it is not possible to change the damping to suit varying conditions. With rising rate suspensions, the stiffness increases as the suspension is compressed by the aerodynamic downforce and ideally the damping would increase too. For a long time, a way of achieving this has been sought and a clever mechanism is bound to appear soon, if it has not already done so.

Two regulations new this year, have had an influence on suspension linkage design. In 1996, Tyrrell decided to test the FIA's interpretation of "aerodynamic effect" in regard to components that moved relative to the entirely sprung part of the car. They made the entire top front wishbone into a streamlined shape, forcing a new regulation to define suspension linkages as having no greater aspect ratio (defined in this case, as chord to thickness ratio of a transverse section) than 3.5:1. There is a sufficient aerodynamic advantage in streamlining the linkages, especially the top one, that there became an even greater reason to manufacture them from individually shaped CFRP mouldings, rather than standard section, streamline steel tubing.

There are a number of constructors however who prefer to retain steel for most if not all suspension members, adding a CFRP sleeve to streamline them. This may have something to do with the second regulation, which states that the forward, bottom wishbone strut must be the strongest in tension, and be able to articulate, without necking out the inner joint, at least 60¡ or until the wheel hits something solid on the car (i.e. the side impact structure for the front wheel). This regulation is a reinforcement of the need to try and prevent the front wheel striking the driver's head, in the event of a front/three-quarters impact into a barrier. CFRP linkages tend to shatter in an impact, and are unlikely to retain the wheel. Steel with sufficient ductility stands a chance, and provided that the front lower link is intact, will stop the wheel swinging up into the cockpit. It has not been possible to define a test that will ensure success under all possible types of impact, and so a sensible requirement has been stipulated. Only time will show whether it is the right one. Once again one is reminded that safety is a statistical science, based on experiment, and lest designers should forget this, there have been a higher than usual number of suspension failures this year - notably Mika Hakkinen's rear suspension failure at Spa, and Katayama's Minardi at Monza. It is likely that the change in loads brought about by the higher grip tyres have caught out the designers, illustrating once again just how close to the edge is Formula 1 design. However, it was in pre-season testing that a most interesting phenomenon occurred, catching out two of the most experienced teams of engineers.

Winter testing on the new 1997 cars, fitted with rear impact absorbing structures for the first time, threw up a vibration problem. While all the teams who experienced the problem managed to solve it, it serves to illustrate a technical phenomenon that is well understood in the aerospace industry, but not so well understood in motor racing.

Rear wings have, until this year, been mounted on twin plates bolted to the gearbox. The plates support an airfoil shaped cross beam, to the ends of which are bolted the rear wing endplates. The rear wing assembly is attached between the endplates. The new rear impact absorbing structure occupies the space used previously by the twin mounting plates, it did not take designers long to decide to mount the cross-beam direct to the conical impact structure.

The impact structure, looking like a reversed, similarly functioning nose cone, is very stiff and is attached rigidly to the rear of the gearbox case. The cross beam is bonded to it.

The teams that have experienced problems - , and it is probably no coincidence that two of them have the Renault engine - have not been specific about the details, but it would appear that they suffered cracking around the attachments of the cross beam to the wing endplates. Fortunately the cracks were detected in time and no catastrophic failures of the rear wing mountings occurred.

Whether it was a new excitation frequency or increased amplitude in the new Renault engine, or whether it was the new rear wing mounting structure mode shape that induced the problem is not clear. I suspect it was the latter as the new engine ran in last years Williams and Benettons during winter testing without problems.

1997 was a year of refinement to chassis and engines and a relearning of what it takes to participate in a tyre war. Plenty of hard-earned experience gained in the R & D laboratories and on the track - much of it to be filed away at the end of the year as it will be of no use in 1998.

Major regulation changes (not only reducing the overall width, but also stipulating new, increased minimum monocoque dimensions) and a substantial tightening up of what is permitted in those crucial control systems, means that the cars for next year are very different machines. Most teams have built mule vehicles and run them during the year. In doing so they have gained vital design input knowledge for the changes to dynamics and aerodynamics, brought about by the narrowing of the car by 200 mm overall, and the effect of grooving the tyres. Initial tests brought strong reactions from some drivers - probably more to do with their team's ability to predict the set up of such radically different cars, than a representative view of how the new cars will perform. As the cars have been developed and the tyre companies have learnt just how hard the compounds need to be to stand up to the grooving effect, the comments have become more muted. Some drivers even enjoy driving the cars.

The FIA is also considering bringing in a means of checking that bodywork and wings do not deflect unduly and beneficially under aerodynamic loads. CART has had such a regulation for years, but Formula 1 has avoided it. Ferrari may regret the day that they introduced their flexing front wing!

One thing is clear: the changes look as if they will achieve the FIA's stated intent - to bring lap times back to 1996 levels. How long this will be retained, in the face of the relentless technical development of Formula 1, remains to be seen.