Peter Elleray on the Bentley LMGTP

Many thanks to Peter Elleray
Images courtesy Alastair Macqueen
Does Peter Elleray need an introduction?  He's the Chief Designer for the 2003 Le Mans winning Bentley EXP LMGTP.  Mr. Elleray graciously sat down with us to mull over our questions.
Mulsanne's Corner:  What is your background?  How did you get involved in racing? Who was your first job with and what was your position?

Peter Elleray:  I graduated from Durham University in 1979 with a B.Sc. in Applied Mathematics – not Engineering - and spent 5 years working in aerospace, first as an aerodynamicist, and later as a designer on future projects, but it was really a means to an end – I’d wanted to design race cars since I was a schoolboy, as some of my teachers could no doubt confirm - but there wasn’t such an obvious route in those days, particularly if you had a Maths degree. 

My first professional involvement with racing was when I applied for a job with the Tyrrell team in 1982, as an aerodynamicist. I didn’t get the job (they ran out of money), but I ended up doing some analysis work for them on their ground effect tunnels, and they took me to some races and tests with them during the summer of 1982.

After that I got involved in Formula Ford racing in the north of England and during 1983 and 1984 helped a couple of amateur teams. In the summer of 1984 I got my first proper job in racing, with the Arrows F1 team. I was employed as a design and race engineer. In those days you got involved in everything, and at that point there were only 3 of us in the design office (!). In the winter of 1984 I ran one of the cars in testing and then in 1985 started to run the second car – which was Gerhard Berger’s  –at   the races.

MC: What was your specific role in the development of the Bentley GTP?

PE: I was given the job of Chief Designer after LM 1999. That meant laying out the concept design of each of the cars. So, working with the engine and X-Trac gearbox I began by drawing a complete car scheme. That’s the way it used to be done and its still the way I like to work.  I feel that if you haven’t generated the basic layout then you can’t really appreciate the conflicting requirements.  I work in 2D on that, although I no longer use a drawing board.  So my 2D CAD schemes (really electronic drawings rather than true CAD models) generate the basic layout of the chassis, suspension, side pods and so on, and are then handed over to the mechanical designers to engineer into the bits.  I’ve done a fair amount of suspension design work over the years so I was happy to delegate that and keep an eye on it.  After that I concentrated on the aero program.  We didn’t have an aerodynamicist after 2001 so I ran the program myself, which worked out quite well.  To a lesser extent I also led the composite design, specifying the majority of the body splits and basic lay-ups.  I was also fairly precise about the way I wanted the tub to be constructed and the lay-up of that; although we had quite a bit of analytical work done on it first by a consultant.

Once the cars were up and running my role switched to coordinating any updates or changes required, and to advising the race engineers if they were operating the cars in a way that didn’t make the best use of them – which rarely happened, we were very well served in that area.  I attended about 90% of the track running of the cars throughout the program, and it would be fair to say that when something broke or wasn’t working properly then I was expected to sort it out!

MC:  What is your design philosophy?  How do you start when given a clean sheet of paper?

PE:  I’ve been asked this a few times since June and never really been happy with my answers.

Although we have good analytical tools to help these days the Bentley project had nothing like the amount of analysis thrown at it that you get in modern F1, so to a large extent, and I think this is still the case in sports car racing generally, you still get an individuals own ideas in the makeup of the car.

But it’s very difficult to describe your own design philosophy.  I suppose I usually start with the aero side, and then try to fit a workable structure and suspension systems into that, making compromises as you go.  One very good piece of advice that I was given a number of years ago is to design the car you want, and then make it fit the regulations!  It’s surprising what you can get away with using this approach.  Of course you can only push it so far, and subconsciously you don’t draw something that you know you can never make-work.  I suppose it’s about pushing everything to the limit, which means knowing what the limits are really. That probably sums up the ’03 car quite well…

As regards a clean sheet of paper, I’ve been given one of those twice now (once for the ’00 prototype and then for the ’03 car). You still end up using your experience and being influenced by what has gone before.  If you look at both those cars in detail then there is virtually no lineage with their immediate predecessor (’99 R8C and ’02 Evo Bentley) but that’s inevitably where I started each time.  I suppose there is an evolution on the CAD screen as you go through the detail design which links it all together for the designers but which the rest of the world never sees, so they think “Why did they do that? ”. 

That would be truer for the ’00 car, which came together over a relatively long period and changed direction more than once, but the ’03 car was a little different.  I seem to remember having bits of it either on sketches or in my head for months or even years, and when I did finally sit down to do the concept scheme it came together very quickly.  If you look at the Wind Tunnel model for that one at its first test then it is almost identical to the car at Le Mans 15 months later…

MC:  Did the Bentley benefit from the work carried out on the Audi R8C?  How much of the design staff carried over from R8C to Bentley?

PE:  The prototype ’00 car, the one that never raced, did benefit from the R8C experience, but only in as much as the early car showed us what not to do!  Having said that most of the problems with that car were aero related, and although I steered clear of carrying virtually any R8C ideas onto the ’00 prototype, I did reintroduce a number of them, mainly in the suspension, on the ’03 car when nobody was looking.  Which rather tended to prove the old adage that you need to understand the root cause of a problem to deal with it satisfactorily! 

There was not a great deal of carry over of design staff between R8C and subsequent Bentley projects.  We were on the point of moving to 3D CAD for the ’00 car, and recruiting for that whereas quite a bit of the R8C was drawn to our schemes by contract staff in 2D.  The design team was fairly stable during the rest of the Bentley program.

MC:  Why did Bentley decide to go the LMGTP route?

PE:  Good question, ask Bentley!

I think that the visual presence of the GTP may have been as important as anything actually.  They were always very pleased to hear that we were considered to have the best looking car.  I think, if you look back through the history of prototype racing then the closed cars have always been charismatic, and, in as much as the racing was part of their marketing strategy then I’m sure this was a factor.

Once that decision was taken however, then there was an ongoing debate as to whether the GTP should be inherently faster than the open car.  I’ve looked at both in the tunnel, and schemed out open as well as closed concepts over the last few years, and I would say that actually the ACO did a very good job of equivalency on the whole.  A lot of what was written in the press over the years ignored the small print in the regulations, which was really the key to it.  I think we showed, over the period ’01-’03 that you can make the GTP work, but only if you can optimize those areas where the regulations give you a head start.

MC:  How did you tackle the wind tunnel testing of the Bentley from 2000-2003?  What scale model and in what facility?  How many hours were spent developing the ’01 package?  02?  03?  How much CFD analysis?

PE:  The tunnel program was based at Emmen in Switzerland for the duration of the project, which began in Sept. ’99.  We ran a 40% scale model at 50 m/s in their open jet tunnel.  The model was new for the Bentley program and ran 7 tests for the ’00 car, about 200 hours.  Then we had to redesign that car as a turbo car for ’01, and that took 5 more tests, maybe 150 hours.  By that stage the model had been completely rebuilt so that it shared no parts with its original incarnation, and the ’01 car shared very little with the ’00 prototype except the general shape.

We had some basic problems correlating track with tunnel at this stage in the development, and so we ran a test in April 2001 to try and sort that out prior to LM, and although we did sort it out the parts didn’t go on the cars till the autumn.  So we ran the 2001 race with an aero package that wasn’t right.

Then after LM 2001 I did a series of 3 tests that defined the ’02 aero kit, about 120 hours, with a further test in Feb. 2002 to map the car for the race engineers.  This time we got a much better correlation. In between that I took a hybrid model along to try out some ideas for the ’03 car on the original chassis, this was December 2001.

The ’03 program started in April ’02, using the hybrid model completely rebuilt with a close to definitive tub for the ’03 car.  There were 5 tests, again about 200 hours spent on the LM package, although as I’ve said earlier, it was probably 90% there after 70 or 80 hours.  Then we ran a test that was mainly the Sprint setup, another 40 hours, and finally two mapping tests, one for Sebring, and one for LM.

I got to like Emmen and the staff very much over that period. Once you got used to the way the facility worked then it was possible to achieve a lot with it.

We did make a little use of CFD, but never really pursued it.  We used it on the air box for the ’01 car, and did some overall car pressure and velocity plots for the ’03, which were actually very impressive, the streamlines equating well with what we were seeing in the tunnel.  I’d definitely like to follow this through next time.

MC:  Getting somewhat off track for a moment, you just mentioned, “we had to redesign that car as a turbo car for ‘01”, in describing the ’00 wind tunnel model.  What was the originally intended power plant? 

PE:  It was not quite that simple.  There were several possible options, but without being too specific they were all normally aspirated engines.  It is quite a bit more complicated to fit an inter-cooled, turbocharged engine to a car that has been conceived to run NA than to swap NA engines.  That was why so much changed.  The rest of the (non engine) changes were to either make it easier to manufacture, work on, or simple performance upgrades based on the test program.

MC:  Can you describe the un-raced 2000 car in more detail?

PE:  It’s the car in all the early (pre-release) spy shots.

001 is visually very similar to the ’01 car, but is actually almost completely different in detail. Really the same basic layout went through 3 phases.  001, (the prototype), the ’01 cars, and the ’02  (Evo) car(s).  There would be very little on a ’02 Evo car that would fit on 001.

The original car had a longer wheelbase, shorter rear overhang, slightly wider chassis, larger engine cover and was mechanically almost totally different.  I think we reused the front uprights and gearbox internals

MC:  In 2001 the EXP Speed 8 ran a single element rear wing at Le Mans though was seen testing with a double element wing.  In 2002 the Speed 8 appeared with a double element wing and sported louvers over the front and rear wheel arches in addition to overhang changes at the front (shorter) and rear (longer).  Specifically how did Bentley’s aero philosophy change from 2001 to 2002?

PE:  As I mentioned earlier we had some correlation problems in 2001.  Specifically we weren’t generating the front downforce on track that we were obtaining in the tunnel.  That was one of the reasons for running the single element wing in 2001, to keep the car in balance aerodynamically.  When we began to put some interim front diffuser parts on the car in the autumn of 2001 then we were able to go back to the 2-piece rear wing.  However, in that early hybrid test car we were developing too much drag for LM.  But at that stage I just wanted the drivers to experience the extra downforce and tell us if that was what they needed.  Actually it wasn’t so that was when the louvers appeared.

The overhang changes were to address the drag issue.  I developed a new front diffuser which was more aggressive, but also shorter.  That gave us the front downforce, but only with the louver package.  However they contributed little drag so that was OK.  The overhang lost at the front I put on the rear, and moved the wing back so that I could generate the rear downforce by leverage, and back the wing incidence (and drag) off.  It was also a shorter chord wing, which put the center of pressure even further back.  To generate the downforce it required quite a lot of camber, so it needed a flap and a slot gap.  It was not the same two-piece wing that we had run earlier. The rear louvers were a late addition (Feb. 2002) and gave a little more rear, although we couldn’t get them to work on the ’03 car when we tried! That was the definitive ’02 EVO kit.  There was a very low drag setup for the car as well, which used a different single piece rear wing and removed front louvers. We tried that at LM in 2002 and hit 337 kph down the first section of the Mulsanne (!) but couldn’t get around the corners.  That wing reappeared in ’03.

So, no change in philosophy, just working to find what was missing.  But it wasn’t the answer.  The car was too sensitive in heave and pitch and consequently had to be sprung too stiffly for the tyres, which got hammered as a result.

MC:  How pitch sensitive was the Bentley’s aero. platform and how did drivability improve ’01-’03?  How much variation in front downforce % was there for ’01-’03 cars across a ride height map?

PE:  This is one of those how ‘how many apples are there in a bunch of grapes’ questions…

The most extreme example of aerodynamic pitch sensitivity is perhaps porposing at high speed, but there are many other ways in which the performance is influenced by pitch sensitivity.  We all have our own method of quantifying that. 

Generally speaking, the more downforce that you make with the front diffuser, the more susceptible your aero package will be to pitch sensitivity and possibly porposing.  That is one of the key features of the 2004 regulations, its why the front diffuser has been raised by 50 mm...

Without going into specifics of front downforce coefficients or variation, one of reasons that the ’03 cars were much more drivable was because we learned how to keep that sensitivity under control.

MC:  Regarding the front diffuser, how powerful are the tuning strakes?  How pitch sensitive are they?   Over all this is a sensitive area to test in, how do you balance model sensitivity with the need to develop the diffuser?   And why is it that the correlation between front balance achieved in the tunnel and that achieved at the track is so divergent? 

PE:  It would be difficult to say how pitch sensitive strakes are.  They can be quite powerful in certain circumstances. All diffusers require some help from them, but I would question whether or not you should need to run more than a handful on each side.

I believe that the front diffuser scenario is related to surface finish, and hence boundary layer characteristics rather than a pure scale effect.  The pressure coefficients under the throat are the highest on the car, and super sensitive to the surface finish.   A mirror smooth diffuser will always give much higher peak suctions and hence more leverage about the front axle.  I've yet to see that replicated on a full size diffuser on track.  The difference seems to depend on how 'aggressive' the section is, and can be of the order of several percentage front balance loss on track.  I would hazard a guess that whatever their tunnels are telling them, that people are actually running their cars so that the trim (balance) in high speed corners, is roughly the same as their static weight distribution, i.e. low to mid 40's.  Being a boundary layer related phenomenon I would think that this is also tunnel dependent, just to muddy things a bit more.  You almost reach the stage where you have to correlate the tunnel to the first car you run.  If you have no prior warning of that then you can get quite confused.

It’s a lot easier to get a nice shiny carbon tunnel model to produce honking good front downforce than a real car, and my experience is that the difference is significant.  I've had much better correlations from track to tunnel, if I don't tunnel test with an idealised diffuser but my tunnel numbers look less impressive.  On track those cars have had high-speed aero performance that was at least as good as some others who claimed far better numbers.   That would include this years car, which didn't look to be giving anything away even in low drag trim, so...

MC:  Can the use of the diffuser strakes on the Bentley ‘03 be seen as tuning devices as say dive planes, being used to add or take off downforce as needed?

PE:  I’d rather not get into this, as it is still relevant under the 2004 regs.

MC:  The 2003 Bentley was a large step away from the more conventional ’01-’02 series. Would it be best described as evolutionary or revolutionary?  What ideas or parts carried over to 2003? What areas were most improved upon?

PE:  I don’t think you could describe the car as revolutionary, because everything on it had been done in some form before.  But I suppose it’s true to say that the overall package was quite different to what had gone before as a result of integrating some of those ideas on a coupe. 

There was very little mechanical carry over from ’02 to ’03, just the engine and the internal layout of the gearbox, but even there a lot of the components were different. We carried some of the torsion bar system over at the front – the rear was all new, being torsion bar for the first time – and the wheels and wheel bearings and hubs.  That was pretty much it.

As far as ideas being carried over, the layout of the cooling system, although not the ducting, was largely the same, and the front diffuser section was actually the same.  We carried on with the electrically assisted power steering from Kayaba, but all the bits were different.  The basic fuel system concept also carried over. 

In contrast to that all the key performance related parameters changed.  Weight distribution, aero concept (and aero map), suspension geometries, all these changed significantly.  And all at the same time, which is usually a recipe for disaster. Fortunately…

By and large this was driven by the knowledge that only a win was going to be acceptable.  If you know that your going to be in serious trouble if you finish 2nd then it probably encourages you to take a few chances…

Having said that, I was always pretty pleased with what was coming out of the wind tunnel, and as far as those areas that were most improved, then I would say that it probably starts with the aerodynamics.  The car had a good useable level of downforce over a wide range of ride heights and pitch angles, which meant that it could run softer than the 2002 car.  Whether that assisted Michelin in a way that we hadn’t assisted Dunlop I don’t know, but the race showed that we were able to look after the tyres and still be quick.  We had changed the suspension geometry over the winter as well, and I’m sure that played a part.

I think that much was definitely part of the design philosophy, but the other thing which was that we ended up with a very driveable, balanced car, although that came about during the test program rather than off the drawing board.  By the time we got to LM the car was very driveable with a very high level of grip.  That had really been a case of trying various combinations of roll and heave stiffness, 7 post rig testing and also playing with the weight distribution a bit.  I think the aero map being pretty stable over quite a wide range was also a function and then Michelin responded with the right tyres for the car and suddenly we were right there…

MC:  For 2003 a single element wing reappeared at Le Mans. What drove that modification?  Can you compare the aero package of the #7 and #8 cars at Le Mans ’03 and describe the reason behind the differences (driver preference/comfort inter-team competition, etc.?)

PE:  The 2003 car was on the limit as regards overhangs.  In other words it would have been impossible to move the rear wing back any more and remain legal, or for that matter make the front any shorter.  Also the top body shape was designed to reduce the turbulence of the flow onto the wing.  The net result was that we could knock 2 or maybe 3 degrees of wing off compared to the old car.  At those sorts of wing angles the single piece wing was actually more efficient than the two-piece.  At Sebring, where we ran a higher level of downforce we ran the 2 piece.

The aero packages on the two cars were just two of several options developed in the tunnel.  The basic aero concept did seem to be very versatile.  You could get a high downforce sprint setup, or a really low drag Le Mans, and keep in balance, and by playing around with louvers and turning vanes you could bias it between slow and fast corners differently, entry and exit etc.  I suppose it was the payoff for the investment in the tunnel program.  Both cars actually had a similar balance, but car 7 ran a higher level of downforce (and drag), map on map, although the mechanical setup was such that it ran in a different part of that aeromap and the end result was fairly similar.  Car 8 ran a setup that was close to the tunnel baseline, which was good, because it showed that the tunnel was giving good data.  The overall difference in lap time was not significant. Both setups had their merits.  It was actually quite rewarding to see the race engineers being able to dial in the cars to the different driving squad’s preferences in this way.  The old car had only really worked with one setup.

Car 7 ran the vertical wing endplate extensions.  The drivers felt that these stabilised the car in yaw, which would be logical, whereas on car 8 they didn’t like them.  I have to say that having poured over a lot of tunnel data with these off and on that I think the effect of these came down largely to yaw behaviour, certainly on the Bentley where the wing mounting plates were only just inboard of them.  In other words any advantage of a deeper endplate in making the wing more efficient is difficult to show when you look at the complete aeromap – at least with the profile we ran.

MC:  Using Relative figures (% improvement) how did the aero figures (lift and drag average over an aero map) evolve from the R8C to the 2003 Bentley steeping through the 2001 and 2002 Bentleys.  How did the Sprint variant compare?

PE:  I think we’ll draw a veil over the R8C, but in terms of the Bentleys, between 2001 and 2003 we found about 30% more downforce in high speed corners, and 75% more front downforce, which moved the split forwards 10%.  That would also be a good average across the aeromap, not forgetting that we might have been in a different part of it during 2001 than in 2003.  The efficiency went up by 25% over the same period, which meant that we ran slightly more top end drag in 2003, although this could have been trimmed out for the loss of a little downforce.  The ’02 car was as much as 70% of the way to the ’03 car in certain parts of the map, but not in others, and that was really its downfall…

The Sprint package used at Sebring added as much as 17% more downforce over a low drag LM package, and as little as 5% over a high downforce one.   The efficiency would have been about 10% lower, with a very similar balance.  The ’00 test car only ever ran in Sprint trim, and over that car the Sebring setup showed 75% more front with similar rear, a 13% balance shift!  The ’03 car also carried 5% less drag, and was nearly 30% more efficient.

So we have learned a lot about balance over three years!  All those numbers are derived from the same tunnel of course, so they should be reasonably indicative of the gains we actually made.

MC:  How much did the Bentley’s monocoque weigh?   Were any significant weight savings found from ’01-’02?  How much different was the ’03 car’s monocoque?   Did the tub’s torsional stiffness improve between the variants?   Additionally, how does the torsional stiffness of a closed LMGTP compare with an open LMP?

PE:  All of the Bentley monocoques have been in the 70 kg region, as delivered from the composite assembly shop at RTN.  This is the case for the earlier cars as well as the ’03 cars.

To a very large extent that weight has been driven by the FIA crash tests that the chassis has to pass before it can race.  The nose box test has never given too much of a problem, nor the squeeze tests, but we have found the roll hoop tests – which were drawn up for open cars - to be quite challenging for a closed one.  There is a serious bending moment applied to the front A pillars in this which, as a result, carry quite a bit of weight, all of it high up as well.  That’s an example of the detail of the regulations hurting closed cars.

The ’01 cars were produced from modified tools used to make the ’00 prototype (which had a larger cross section at the rear of the tub), but were fundamentally different inside.  The ’01 cars had composite rear roll hoops (the ’00 car had a steel hoop inside the tub), as well different front suspension damper channels and a different fuel cell bulkhead.  The lay-up was quite different too. However, they were conventional in as far as they used a top and bottom section joined by a tongue and grove joint at the waistline (just below the door aperture). 

The ’03 cars were fundamentally different. We had had a fairly large accident at Paul Ricard in 2002 with Eric Van De Poole and although he was unhurt I still felt that we could improve the way the chassis had dealt with the impact.  I decided to make the car more like an open car, with a monocoque coming right up in one piece to the driver’s shoulders and with a separate roll hoop (in carbon) on top of that.  The chassis was heavily waisted in plan view so this was actually closer to the width of two single roll hoops on an open car than a conventional twin hoop.  We then inserted our 360-degree a hoop at the dashboard and glued a small roof on top of the two hoops to join them together.  This meant that the hoops were homogeneous carbon extrusions, rather than being formed from angle sections inside the outer skin of the tub.  I figured that in this way they would be able to withstand repeated impacts better.  Fortunately we didn’t have to test this out, Johnny’s accident at Jerez being very similar to the nose impact test, only with significantly more energy …

The only other significant piece of structure in the ’03 chassis is the fuel cell bulkhead, as the suspension sits on top of the chassis, rather than inside as on the earlier cars.  To the best of my knowledge this is the first time anyone has built a closed GTP chassis in this way.

The earlier tubs were also designed around the pre-2001 footbox regulations, which were only 250 mm wide per side, whereas the new tubs had to be 330 mm wide per side.  By shuffling the driver around we managed to negate the extra width and in fact the ’03 tub is narrower than the original in the area of the A pillar!

The torsional stiffness of the chassis has also been very similar for all variants.  The closed coupe with a structural roof behaves as a shell type structure, and so it’s the cross sectional area of the complete chassis and skin thickness that determine the torsional stiffness and this has been pretty consistent from start to finish because of the crash regs.  Interestingly, the open car that we built when we were called Tom’s and working for Toyota had a very similar torsional stiffness figure despite the fact that an open car behaves as two beams in differential bending (and warping) about the cockpit opening. 

As regards the overall level of stiffness, it is an order of magnitude higher than the roll stiffness of the suspension, which is usually a good rule of thumb to work to…

MC:  For each year, starting with the ’00 prototype, what were the definitive variants/configurations that were developed (LM, “Sprint”, EVO, Sebring spec, etc.)?

2000:  No specific specification for the prototype, just lots of bits. Retrospectively we would have described it as being in Sprint trim throughout the test program.

2001:  Developed in the tunnel in LM trim and ran in LM trim up to the race. Ran the autumn/winter test program in a hybrid format, which was neither LM nor Sprint, but closer to the latter.

2002:  ‘EVO’ spec, which was LM only. 

2003:  First model developed with a specific Sprint trim, which of course we ran at Sebring. Sprint could be ramped up, at the cost of quite a bit of drag, with diveplanes, body gurneys and flap angle if required - it wasn't, interestingly.  After that race we ran it in LM trim for the rest of its life. There was a low drag specification but we never ran it fully trimmed out.

ęCopyright 2004, Michael J. Fuller