This article appeared in Racecar Engineering, The International Journal of Motorsport Technology (RCE V19 N12).

As a computer scientist, I find it hard to comprehend that you cannot always quantify engineering. Computers represent, in many respects, a form of perfect engineering, free from the burdens of materials, tolerances and the infinite physics of an environment. In theory, a given operation is entirely predictable whether it is invoked once, twice or a thousand times.

The same cannot be said for most forms of motorsport engineering. Despite absolutely every possible effort being taken to minimise variance, there is no guarantee that two engines, two chassis or two sets of tyres, for example, will ever perform equally – despite being manufactured to an identical specification. And so, I was wondering, if you cannot quantify the manufacturing itself, is it possible to at least quantify the performance gain between different engines?

Formula Ford and the legend of Patch

Ever since the hey day of Formula Ford, there have been stories of engines that have dominated the formula. And the legend that is Patch demonstrates this well.

Patch was a Kent engine with remarkable pedigree. It is most well-known for powering four different drivers to successive Formula Ford Festival wins - Roberto Moreno (1980), Tommy Byrne (1981), Julian Bailey (1982) and Andrew Gilbert-Scott (1983). Additionally in 1981, it also believed to have powered a young 21-year old Ayrton Senna to the RAC and Townsend-Thoreson championships, in the same works-run Van Diemen that Tommy Byrne piloted to victory in the Festival.

However, the origins of Patch are rooted earlier with a young South African driver called Trevor Van Rooyen.  In 1977, Van Rooyen piloted a semi-works run Royale RP24 with a self-built (but Minister badged) engine. That year, Van Rooyen won the RAC championship and in the process won a staggering 33 races. However, the following year (1978), Van Rooyen’s engine was destroyed in a test session and the South African returned to his native land. For the record, Van Rooyen went on to enjoy a distinguished career which included winning the 1985 South African Formula Two championship.

Leading Engine builder Graham Fuller (Minister International) subsequently repaired the broken block, welding a ‘patch’ of aluminium where the block was damaged. The engine then returned to racing during the 1980s and to claim its place in history.

Whilst many would argue that much of the latter successes of ‘Patch’ was due to powering a works Van Diemen chassis, the dominance of which would invariably attract the fastest drivers anyway, the legend of ‘Patch’ continues to be known as the greatest advantage a Formula Ford driver could possibly have.

Formula Ford engine power and torque curves

The following is the dynamometer chart from my Formula Ford (1600 Kent) engine when it is was rebuilt in January 2009.

The vital statistics are that at the time of the rebuilt, this engine produced 105.8 BHP at 5800 RPM and 148.8 NM/torque at 4400 RPM.

The problem is that with only data from a single engine, who is to say whether or not this is a strong engine? What’s more, any comparison would only be valid if it were also from the same dynamometer. However, after a year of racing in Formula Ford I would suggest that immediately after being rebuilt, this engine was average. It was not poor but it was also not great and as the season progressed and the engine ran hotter, it became an obvious disadvantage.

But my question is, what did this disadvantage really do to my laptime?

rFactor for real world simulation

Image Space Incorporated have been producing world class simulations for over a decade and the latest incarnation of which is rFactor. Consumer distribution is almost exclusively online and the success of which is demonstrated by the large, dedicated and vibrant community of contributors it has amassed. GMotor2, the 3D and physics engine behind rFactor is utilised and licensed in many other popular sim racing titles including GTR2, ARCA Sim Racing and RACE – The Official WTCC Game. What’s more, Formula One teams including Williams F1, Red Bull and Ferrari use a (albeit customised) version of rFactor (and gMotor2) for their in-house simulators.

I could have approached this problem using a headless simulation such as Bosch’s LapSim. However, as a driver, the benefits of using an interactive simulation are too attractive. None the less, in the future, I might still do this and it would be interesting to further validate the results.

A more accurate Formula Ford for rFactor

In an attempt to answer this question, I have modelled the physics of a Formula Ford 1600 that is as close to my Swift SC94 Formula Ford as possible. In the process, I have accurately modelled the suspension geometry, imported the power and torque data from my engine and re-profiled the tyre slip curve to one that resembles the Avon ACB10.

To demonstrate its potential, the following is a comparison of two similar laps (within one tenth of a second) of the Silverstone National circuit. The blue trace is the actual speed (mph) sampled from my Swift SC94 on 27 June 2009 whilst the red line is the same speed sampled from a simulated lap of the same circuit.

It would be unrealistic to assume that the two traces would perfectly overlap. This particular lap of data was sampled during race conditions and at the time I was dicing closely with two other competitors. There is also a degree of precision lost in comparing data sampled from two different loggers (one of which being virtual) as there is a small variation in actual sample frequencies. Finally and inevitably, there are small inaccuracies in both the Formula Ford physics and indeed the ISI model of the Silverstone circuit (Formula One teams will have access to much more accurate surface data, for example).

However, for the purposes of answering my question, it is sufficient and accurate somewhere in the region of about 2-5% at terminal speed (comparing speed and RPM).

The following video is two sample laps of Silverstone National circuit using the simulator (one of which produced the simulated data shown above).

One final note before I move on. Whilst I would love to contribute this work to the rFactor community and make it available for download, I can only take credit for remodelling the physics of the car. The artwork and 3D model itself originates from the impressive netkar PRO and doing so would be a violation of their copyright and intellectual property. The original conversion of the car encountered similar problems and for those reasons I will continue to respect the rights of these parties but thank them for their hard work.

Building three more powerful engines – artificially, that is

I have remodelled the engine power and torque curves three times – with two, five and ten more horse power. This is obviously theoretical; a real engine producing more peak power would do so with a entirely different power curve. If anybody has dynamometer data from their own Formula Ford engine please drop me an e-mail.

Running the experiment

I have run four separate 20-minute simulations of Silverstone. In order to minimise driver inconsistency I will be using the worlds best test driver – the computer. This enables me to run four near identical sessions. Internally the rFactor AI uses predefined way-points on the circuit to drive each lap.

A summary of each session is shown in the table below.

The following chart shows the individual lap times for each session run.

Finally, the following two charts show the speed and RPM traces respectively, for the fastest lap in each session.

Making sense of it all

I don’t think there is any particular surprises in the results. The fastest overall lap time was set using the most powerful engine. Ten extra horse power yielded a lap time of 62.626 seconds or 1.105 seconds faster than the best lap time set with the standard engine. Similarly, the average lap time was slightly faster still at 1.113 seconds.

An extra five horse power produced a fastest lap time of 63.298 seconds, 0.433 seconds faster than the standard. On average, it was 0.463 seconds faster.

With just two more horse power, the best lap time was 0.183 seconds faster than standard and on average 0.202 seconds faster.

Broadly speaking in this experiment, one horse power equates to roughly a 0.1 second per lap performance. Obviously, this will vary from circuit to circuit.

Back to reality

During the National Formula Ford race weekend in June I qualified in 7th position with a best time of 63.986. This was 1.216 seconds off the pole position time of 62.770, set by Rory Butcher. For the record, I did not test before this weekend and this qualifying session was the first time I had driven a Formula Ford 1600 around the Silverstone National circuit. I was also driving with a left-rear slow puncture caused by a broken tyre valve core.

If we assume that there was a couple of tenths to be found through familiarisation with the circuit, that brings us close to the fastest simulated time of 63.731. Likewise, Butcher’s pole position time of 62.770 is equally close to the best simulated lap time of 62.626, set with ten extra horse power. However, I would suggest this is probably more coincidence than it is scientific fact. An interesting coincidence, none the less.

Concluding with the caveats

Ultimately, there are many caveats to the conclusions drawn from this experiment. I will reiterate that the remodelled power curves are artificial and in many ways, unrealistic.

There is no allowance for engine wear – at the Silverstone race weekend my engine had run for around 400 miles and probably no longer produced 105.8 BHP at 5800 RPM. However, this works both ways and it could actually mean a horse power differential closer to ten, after all.

The same gear ratios were used for each run (2.40, 1.74, 1.43 and 1.17). These are also the same ratios used during the race weekend. With ten extra horse power, I would expect to run longer gear ratios and suggest this would see a further reduction in lap time. The flat line at the end of the straights on the RPM trace confirms that fourth gear is indeed too short.

Ideally, this experiment would need to be repeated a high number of times before the data is collated and analysed. Unfortunately, one reality of using a real-time simulator such as rFactor is that each 20-minute simulation takes 20-minutes to complete. A headless simulator, such as LapSim, would be able to repeat simulations at a much faster rate and as such produce more reliable results.

As far as I am concerned, this experiment has demonstrated that despite its caveats, a small horse power advantage offers a quantifiable and not insignificant reduction in lap time on the Silverstone National circuit. But even more, it has demonstrated what a brilliant and accurate simulation rFactor really is.

Related posts:

1. Formula Ford (Duratec): Snetterton testing
2. Formula Vee: Pole Position, Fastest Lap and 2nd at Silverstone
3. Articles

This article appeared in Racecar Engineering, The International Journal of Motorsport Technology (RCE VN18 N9).

At first glance, it is difficult to imagine improving a Porsche 996 GT3 Cup Car. If perfection was ever attainable, wouldn’t it already be close? However, Quaife’s latest “Q-Tek” (QBE61G) gearbox promises just that and offers an affordable alternative to the more traditional Hollinger or Sadev gearbox.

Like its Porsche Motorsport counter-part, the QBE61G is a six speed sequential transaxle gearbox featuring a larger than standard 85mm shaft capable of applications in excess of 500bhp and 450lb/ft torque. As standard it is supplied with a plate-style limited slip differential although Quaife’s newer automatic torque biasing differential is an optional upgrade. It features a wholly-mechanical sequential shift and internal lubrication via an integrated, internal oil pump.

Mike Quaife believes they have identified a requirement for this type of gearbox. He explains: “We could see a market amongst club and national level Porsche competitors for a more economical gearbox. It is designed within the dimensions of the original Porsche 996 meaning its integration is a straight-forward conversion.”

Indeed, they have achieved that. A race ready QBE61G, including bell housing, gearbox mounts, flywheel, triple plate Superclutch and ancillaries costs 10,950GBP (+VAT). Adding a limited slip differential brings it to a total package price of 11,900GBP (+VAT). Compare this to the equivalent Porsche Motorsport Sequential box for a not insignificant saving of 8,761GBP.

And it’s not a compromise, either: “The QBE61G achieves a 20 millisecond shift – that’s twice as fast as the standard Porsche gearbox.” explains Mike. “Infact, it’s so fast that in testing the automatic ignition disengage was disabled to make the car drivable on down shifts.”

He continues: “In testing, its reliability has been impressive. We’ve completed almost 30-hours of testing, in the UK and Spain without a single failure.”

This quality of build is underlined by its potential for further applications: Quaife offer a helical set of ratios for use on the road (and a helical gear limited slip differential) and the QBE61G could even find its introduction into GT3 endurance racing such as the LMES or FIA GT.

“Overall, we wanted to provide value for money with the QBE61G,” concludes Mike. “When you also consider its performance and reliability too, I think we even surprised ourselves.”

A Drivers Perspective

The first drive in any new car is a daunting experience. You drop the clutch, exit the pitlane and unleash to a world of uncertainty. Racing drivers perform in the subconscious – they are perfect choreographers of laps practised, performed and executed hundreds of times before. Every steering, brake and throttle input, every direction change and gear shift is rehearsed to the point that in itself, it is an almost instantaneous and automatic reaction to the changing behavior of the car or the circuit. Broadly speaking, it can be taught with seat time. However, until this point, you are very much driving with a conscious mindset.

A Porsche 996 GT3 Cup Car is about as hostile of an environment as any sports car debut could be. Reputations are built for a reason and a 420bhp 3.4-litre flat-6 mounted beyond the rear axle is along way from the nimble, mid-engined single-seaters that I am used to.

Every shift must be a positive application of the gear lever. We’re using the QBE61G-specific Quaife gear lever, designed for the 996 and I am told it is another improvement over the standard mechanism. Ergonomically, I have little doubt and the actual process of changing gear becomes natural within the first lap. The clutch feels light, probably too light, but as this test car approaches 30-hours of running, it can be forgiven.

Under heavy acceleration (and maximum load) a small breathe of throttle is all that is required to provoke a seamless and clutchless up shift. There is no significant transmission jolt and the speed of engagement means there is only a small transfer of weight and therefore wasted energy.

A drivers unfamiliarity with a car is normally most prevalent on corner-entry. The initial cornering phase involves more thought than can be consciously achieved given the available time. But such is the precision of the QBE61G that I find myself underwhelmed by the complexity of down shifting five gears in a bumpy and difficult braking area. Within no more than a lap and half I am confident that my down shifting technique, synchronised with a positive blip of throttle, is learnt and already I can become more receptive to reaching the threshold of the Porsche’s ceramic brakes.

Each and every push of the gear leaver emphasizes the elegance of this gearbox. Even a total mismatch of revs on a downshift does little to provoke the kind transmission or differential oversteer you might expect from this type of car. In fact, I would even surmise that drivers without a good heel-and-toe technique would be more than capable of driving the car at moderate to high-speed.

Perhaps, ultimately, the point is this: despite its reputation and its undeniably intimating demeanor, my first experience of a Porsche GT3 was not the learning curve I had expected. I might even go as far as to say that it is a more comfortable experience than say, a Duratec Formula Ford. If you consider the reputation it has, maybe this is difficult to accept or even understand. However, what key component was different about my first experience to that of those before me? Gearbox. And what does that tell you?

Related posts:

1. Articles
2. Publications
3. Porsche 996 GT3 Test (Racecar Engineering)

This article appeared in Racecar Engineering, The International Journal of Motorsport Technology (RCE VN18 N7).

Fundamentally, motorsport is the management of risk. On the track, it is the drivers responsibility to take the risks he (or she) feels necessary, within their environment, to maximize performance and yield the best possible result. But off the track, this is an engineers job and by calculating the risks associated with set-up, electronics and engines (to name but a few), only then can the driver focus on the ultimate goal: to win the race.

As the technology on-board modern racing cars continues to increase, so does the complexity of managing it effectively. Traditional wiring looms have become cumbersome and increasingly within modern racing cars they are being replaced with CAN-bus systems. Migration will typically yield a substantial reduction in wires: the minimal configuration can contain just two (data and power) whilst potential still connecting an arbitrary number of sensors, actuators, gauges and ECUs.

CAN (or Controller Area Network) quickly became a standard in the automotive industry after its inception by Robert Bosch GmbH in 1988. However, its introduction into motorsport has proven more sedate – it featured on the first WTCC BMW (320si) in only 2006. Last year, the Power Control Module (PCM) from Ole Buhr Racing received full FIA homologation and its use is set to expand further into World Touring Cars and complement its already established position in sports and prototype racing.

What is it and how does it work?

A CAN-bus system consists of a collection of actuators and sensors (known as nodes), connected via a single twisted-pair (network) cable. CAN itself is the computer network protocol that allows these nodes to communicate without the use of a host computer. Essentially, CAN-bus connects the electronic devices on-board the car in a similar fashion to a traditional computer network.

All nodes on the network are able to transmit and receive messages. Messages carry a maximum payload of 8-bytes and are protected by a cyclic redundancy check (CRC). Within most applications (networks within 40m), transmission occurs at 1Mbit/s.

Unusually for a computer network, individual nodes are not assigned addresses. Instead, messages are broadcast to every node on the network and as such, simultaneous transmission is not possible. A scheme of bit-wise arbitration is used to determine which nodes can transmit based upon the priority of the message and the bus is always available to a node with a higher priority (or dominant) message, even if a lower priority (recessive) message is already in transmission. In which case, once the dominant message is complete and the bus becomes idle, the recessive message is transmitted again.

This bit-wise arbitration is particularly valuable in real-time environments where transmission efficiency is key and where a more common message-wise arbitration approach (such as CSMA/CD used in Ethernet) is not particularly well suited. In CAN, no bandwidth is used for any non-relevant information.

Error detection and prevention

Perhaps the most compelling features of CAN are its sophisticated mechanisms for preventing and detecting errors.

CAN is a differential serial bus which means it particularly good at reducing noise caused by electromagnetic interference (EMI). As the engine-bay of a racing car is a particularly hostile environment for EMI, this design increases the reliability of transmission and reduces the likelihood of unpredictable behavior – such as the loss of engine ignition (from the ECU).

Internally, complementary signals are sent over two separate wires and as a result, it doubles the noise immunity of the signal. This can be best expressed in mathematics.

The voltage difference of the high state is said to be , where the represents one wire, and represents the other. If these are exchanged on the low state (), then the total combined difference is expressed as and is therefore double the noise immunity of a single wire serial bus.

CAN hardware also retransmits faulty messages automatically by detecting errors at both a message and bit level.

At a bit level, two different techniques are used: bit-monitoring and bit-stuffing. Bit-monitoring is where a CAN node continuously reads back a transmission from the bus, comparing what was transmitted to what was received. If a difference exists, the message is considered faulty.

Similarly, bit-stuffing is where after five consecutive bits of the same level are transmitted, an additional sixth bit, of the opposite level, is added (but is not interpreted by the receivers). The actual purpose of this is to avoid excessive DC components but it serves as an additional means for detecting errors.

At the message level three further techniques are used: a frame check, an acknowledgement check and a CRC.
The form of a message (known as the frame check) is validated by comparing the expected sequence of bits (specified in the protocol) against the actual transmission.
Each receiving node adds an acknowledgment to the message and if this cannot be detected by the following node, the message is again considered faulty.

Finally, a basic Cyclic Redundancy Check is applied to the message bits in order to protect the integrity of the message payload.

The CAN specification also includes a slower fault-tolerance mode (below 125kb/s) which enables the bus to continue functioning in the event of the cable being damaged or even partially cut and is particularly relevant in the case of an accident or minor contact over the duration of a race. Ultimately, if a node or the cable is damaged, the network will continuing functioning (albeit at a slower speed).

Why use CAN?

It goes without saying that CAN is not suited to every environment. Although already small (197×107x46mm), Ole Buhr’s current PCM is still somewhat oversized for single-seaters. However, in LMP, sports and touring cars where space and weight is at a relative premium, its introduction is supported by good pedigree: ALMS Champions with Porsche (RS Sypder), LMP with Creation and BTCC Champions with VX Racing represent highlights from 2007.

And neither is it cheap – at almost GBP 3500+VAT, Ole Buhr Racing know its market. But in professional categories where the initial cost of components is secondary to the cost of poor reliability, it’s perfect.

In such applications, fast access to diagnostics means that CAN is the only reliable option. Instead of the time-consuming process of searching for short-circuits or faulty junctions with a voltmeter, software immediately shows the status of every node on the network and because of its architecture, nodes can be readily replaced, added or removed without any implication on other components.

The PCM also features a built-in, programmable logic controller which allows for the automation of tasks when certain conditions arise. For example, the engine could be automatically started as the car is released from its jacks.

It’s here to stay

Overall, it is the absolute simplicity of CAN which makes its use so valuable. An entire network of actuators, sensors or even switches and dials can be reliably maintained with enormous reductions in analogue wiring and therefore risk. As its application within motorsport continues to become more prevalent, the overall cost, given time, will settle. In the mean time, if you’re wondering whether adding another junction to your wiring loom is such a good idea, maybe take a closer look at CAN and what it can do for you.

References

• http://www.embedded.com/columns/murphyslaw/13000304
• http://www.computer-solutions.co.uk/info/Embedded_tutorials/can_tutorial.htm?gclid=CKON_5fTpZICFQ9BMAodRGZxMA
• http://www.sensorland.com/HowPage054.html
• http://www.obr.uk.com/Handbooks/PCM%20Handbook%20ver%20301a.pdf
• http://www.embedded.com/columns/murphyslaw/13000304?_requestid=498927
• http://www.ems-wuensche.com/CAN_technical_information.html#anchor639233
• http://www.kvaser.com/can/intro/index.htm

Related posts:

1. Quaife QTEK QBE61G gearbox, Porsche 996 GT3
2. Preview: Silverstone (UK Formula Vee)