Embedded Files
A race track design that combines good racing, fun driving, and rigorous testing
This endeavour began when I was thinking about the testing processes we went through on the FSAE car. Many test runs were conducted using very specific corner radii, sequences of corners, targeting specific roll velocities or frequencies, etc. The purpose of this was to more efficiently provide validation data for specific vehicle scenarios and to fill out performance envelopes (GGV, yaw moments, shock velocity/frequency histograms, etc). Usually this extra effort meant vastly increased testing efficiency by focussing on the areas of interest and repeating them until sufficient data was collected.
My idea then was to design a test track that fills out these envelopes in this same efficient way from the start, while also promoting good racing, being fun to drive, and being demanding but rewarding of good setup work. I specifically wanted the track design to target medium-high downforce cars (GT3, Prototypes, F4/F3) which I find the most fun to drive in simulation. My checklist for each of the four criteria was as follows:
Performance envelope:
Mix of tight, medium, and open corners
Opening and tightening radii corners to fill in the gaps
Fast direction changes on throttle, off throttle, and on-brakes to test stability and the interaction of yaw moment, steering angle, and lateral acceleration
Good racing:
Avoid short straights that serve no purpose
Difficult corners leading onto longer straights
Extend braking zone distances using tighter corners and/or downhill sections and/or convex elevation changes
Reward deviation from the conventional racing line
Fun to drive:
No short sharp corners that require braking
Space out low-speed corners, but also avoid breaking up flowing high-speed sections
High-speed "thread the needle" sections
No off-camber corners
Undulations and deviations to long straights
Demanding setup work:
Convex track shape into some of the braking zones, requiring rearward brake bias shift
Lateral-grip limited sections during full throttle application out of slower corners to highlight power understeer or oversteer issues
Bumpy sections leading immediately into high lateral acceleration to test damper setup
Direction change while braking to test sensitivity to roll/pitch and promote understeer
At least one high speed corner able to be taken at full throttle, but only just, punishing cars that need to lift even slightly
Plugging all of these requirements together into 5-6 km looked like this:
The easiest way to test such a track was in Assetto Corsa, where a simple Blender model could be almost directly imported and work as intended straight away. My brief time learning blender to create my concept LMH McLaren helped here. The track layout above was the final result after some fine-tuning test drives, and once I was happy with the exact shape of every corner I added some basic scenery.
Track Guide
Corner 1
After a long straight, the braking zone here is slightly downhill and curves away from the corner, promoting an overtaking move on the inside under braking. The exit slowly opens, meaning the fastest line is not straight. This requires careful throttle application, and fills in a decent chunk of the GG plot for combined lateral and positive longitudinal acceleration.
Corner 2
A simple 90-degree left-hander. Nothing particularly exciting, but such corners are present on a lot of tracks, so it's pretty useful for a test track to have one. They also provide a good opportunity for a dive-bomb overtake, partially blocking the other car.
Corner 3
A flat right-hander immediately after the previous corner. The apex sits at the peak of a small elevation rise, which pushes the car wide through the orange section as the front becomes light. It can be taken flat with the right line in most downforce-dominant cars.
Corner 4
A sharp right, with a slight curve and crest int he braking zone, requiring a rearward brake bias to avoid understeer, and testing the vehicle's stability under braking. Particularly difficult in high downforce cars, where braking needs to be applied heavily at first, reduced over the crest, and then applied again, all while managing the reducing aerodynamic load as speed falls. Not a conventional overtaking spot, but the difficult braking zone could provide opportunities.
Corners 5 - 6
The fastest corners on the track, designed to sort out aero balance and roll rates, particularly in the direction change. The first corner requires a slight lift in a hypercar, while the second is flat but still requires precision. The radius and speed differences between the two mean the peak lateral accelerations achieved are similar despite being achieved in different circumstances, providing comparison potential for the differing aero loads, steering angles, and slip angles.
Corner 7
Continuing the high-speed section, the left-hand turn 7 requires a sharp turn-in, keeping the car in the centre at the exit to open up the following chicane. A small dab on the brakes is required, letting the car coast around the corner to slowly tighten the line as the speed drops. A steady platform under braking is essential to avoid inducing understeer on the sudden direction change; Trail braking doesn't really work here because the tightest radii is the initial turn-in point, so coming off the brakes is just as violent as getting on them. From here, the chicane (8-9) is flat out and rather easy to navigate assuming there were no issues in turn 7.
Corners 8 - 9
Immediately after the chicane, there's a close wall on the outside. The fastest line through the chicane is over the curbs, so being able to drive over them without upsetting the car here is critical. Well setup dampers will allow cars to stay flat for longer and take a more open line without the risk of the car bouncing wide into the wall. Turn 10, the following left-han
Corners 10 - 12
Another sequence of corners designed to upset poorly set up cars. The exact shape was adjusted through testing to create a unique case where the fastest line through the whole complex is to brake before the left hander of turn 10, and continue braking into the first part of the sweeping right hander of turn 11. This direction change under braking will test low speed damping, fill in a decent chunk of combined lateral/negative longitudinal acceleration in the GG plot, and run a sweep of significant yaw moment across the neutral steering angle in a grip-limited scenario. This leads to turn 11 becoming a double-apex (shown in orange), with the braking point and apex positioning being dependant on getting on-throttle as early as possible for turn 12 (short left hander) while keeping the angle shallow. The road drops off after the apex of 12, effectively making the turn tighter due to the effects of understeer to further favour this treatment.
In the onboard video below, turn 12 is taken with a lift, but it is possible to take flat if the perfect line is achieved. Clearly I need more practise at my own track.
Corner 13
Not dissimilar to corner 4, but everything is "more". More speed before the braking zone, more braking zone, more elevation change, more radii reduction. This corner doesn't add anything to the testing capability of the track, and is more for fun and good racing. Inspired by turn 10 at Bahrain, it has a long braking zone and gradually tightening radius, requiring gentle feathering of the brake to slow down in time while keeping good rotation. Keeping the car on the edge all the way in and down the slope before finally "landing" at the apex and flooring it is very satisfying. The slow increase of rotation also provides a chance to analyse lateral grip under a more sustained yaw moment and without significant aero influence, rather than the typical spike of a more aggressive turn-in.
Corners 14-15
After the tight right-hander of 13, there's a short straight followed by a section rather like Blanchimont into the bus stop if it were shrunk a little. By limiting the distance between the high-speed turn 14 and the chicane of 15-16, it means the racing line has to quickly return to the left-hand side. This leaves opportunity for a trailing car to open up the corner, carry a little more speed and create a slightly shorter route to the turn 15 apex, meaning between these two corners it's faster to take the overtaking line. Following it with a chicane instead of a single corner means the off-line overtaking car has a chance to gather itself up to avoid compromising the run through the final corner.
Corners 15-17
Coming out of the chicane, there's a quick blast of full throttle required to get up to minimum apex speed before turning into the final turn 17 right-hander, rewarding cars able to get the power down quickly while they can. Those that can't need to continue accelerating up to minimum apex speed while turning into the corner (to avoid the time loss continuing onto the main straight), while trying to avoid power understeer or oversteer. Accelerating through this tricky corner also provides another good chance to fill out the combined lateral/positive longitudinal areas of a GG plot, and provide a sustained (albeit low) yaw moment.
To see if the design really is good as a test track, I drove a flying lap at each of Project 05, Silverstone, Spa, and Monza. Silverstone and Spa for their variety of corners, and Monza for it's frequent use as a test track for many race car series.
Without diving into actual performance analysis and breaking down the car's behaviour through every corner, bump, and direction change, the simplest way I can think of measuring the efficiency and completeness of the track for testing purposes is a GG or GGV plot. The results from the 4 tracks are shown below, sampled at 50 Hz.
Project 05
For all of these plots, I'm going to assume symmetry applies. Different tyre temperatures and certainly different suspension setup (camber, toe) left to right would invalidate this, but for the sake of testing I'll ignore this. Personally I'd be setting up a symmetric car for testing purposes, as then the amount of data produced is effectively doubled.
This particular plot is nicely filled out around all edges. The lateral grip limits are well-defined and are encountered across a large range of longitudinal accelerations from -0.8 to +0.4. The spread of data points is quite even, suggesting efficient collection over a single lap. Lacking slightly are areas of both high positive and negative longitudinal acceleration with low-moderate lateral acceleration. The braking "stem" at the bottom stands out, but this is present on all plots and likely from a combination of my sub-par driving and the lack of high speed, highly curved braking zones on any track. The latter probably comes down to a combination of good racing and safety.
I'm happy with the spread of velocity here, especially when considering symmetry. The left side in particular has an almost perfect spread of well-defined lateral grip limit data between 50 km/h up to 250 km/h.
Silverstone
There is a lot of data here at the lateral grip limit, although it's not as well filled out as the Project 05 plot was. There's peak data missing around the -0.4 g to +0.2 g longitudinal area, and the velocity definition is lacking at higher speeds. A well-defined lateral grip perimeter is only present from km/h up to around 175 km/h; Above that and there's very limited longitudinal spread.
There's also a lot of noise wherever significant negative longitudinal acceleration occurs. Some of this will be down to track bumps, as the bumps on my tracks turned out to be smoother than is probably realistic (this is seen everywhere in general in the Project 05 plot in the form of cohesive "traces" of points, particularly towards the lateral grip limits). In general though, these regions are significantly less defined. Silverstone also suffers from limited resolution in the combined high longitudinal/low lateral cases.
So far, both tracks have a high concentration of data points around the perimeter, minimising less useful data in non grip-limited cases.
Monza
Monza provides probably the most well-defined shape so far, and solves the issue of minimal combined high longitudinal/low lateral data. Beyond these two benefits, everything else is done far better by the other tracks tested. Peak lateral g is low all around, missing both low and high velocities entirely. Nearly all of the grip perimeter is mapped only at velocities between 100 km/h and 175 km/h. Even the combined acceleration cases are not as diverse nor as extensive as any of the other tracks. On top of all that, it's pretty inefficient with the majority of points situated inside the grip limit; time spent at full throttle in a straight line at high speed. Not surprising for a track called the "Temple of Speed", but it seems like a poor choice to test cars on for anything other than engine power or durability.
Spa
Spa is possibly the best one here. Like Project 05, it has very good velocity mapping at the lateral grip limits, extending into the combined acceleration areas. The perimeter is very similar in terms of shape and total area, but achieves higher peak lateral g as achieved in Pouhon and Blanchimont (two red "spikes" on the right side behind the legend).
Eau Rouge also achieves high lateral g, assisted by the vertical acceleration of the compression. Project 05 is missing this elevation effect, mostly as a consequence of steep localised hills being difficult to model smoothly in Blender with my limited experience. I wouldn't consider this essential to have on a test track though, as not many other tracks have such extreme elevation change at key grip-limited locations.
Some of the improvements in definition here could be down to Spa being 25% longer than Project 05, and thus having room for 25% more corners, and longer corners too.
I'm not a fan of a long track for testing purposes though. If there's a specific corner or point on the track that I'm interested in, I don't want to have to wait several minutes for the car to get back there again for more data the next lap. This was an advantage of a completely private test track back in FSAE; we could run the car through a section of track, then immediately turn it around and run back through it again, ignoring the rest of the lap entirely. This way we could get data from the same corner 20 times in just a few minutes.
Combined perimeter overlay
As well as velocity, COG acceleration values are also influenced by other factors affecting individual tyre grip, such as steering angle, slip angle, slip ratio, and yaw moment forces. Plots of various combinations of these factors revealed that for all tested tracks (apart from Monza in some cases), the frequency and distribution of various combinations of these variables were all very similar. It became apparent that to improve the diversity further (to improve testing time/distance efficiency), tracks would have to start sacrificing straights for more corners instead, which is undesirable for both data acquisition and good racing. At least I didn't fall short of the existing tracks here.
One specific area where Project 05 did appear to fall short was in the variety of yaw moments experienced in a lap. The distribution was much more concentrated around 0 than Silverstone and Spa, and looked similar to that of a lap at Monza. I thought I was simply missing some variety in high speed corners and sequences, with the higher downforce able to provide higher yaw moments, but stratifying by velocity revealed a similar trend independent of downforce. I suspect this must come down to the track surface again. Clearly, most moments at corner entry and exit lie within the 50000 range, up to 100000 for faster, sharper instances (I've left these values unitless as I couldn't find an accurate inertia value for a prototype car anywhere). That some corners see a value of up to and beyond 500000 is inconceivable, and is only explained by bumps on the track creating noise and allowing for instantaneous extreme moments to be generated.
Yaw moment distributions
Based on the results, I'd say the track is certainly fun to drive, promises good racing (theoretically at least), and does a pretty decent job of being a test track to the point of being at least on par with some pretty famous existing circuits.
For it to be the perfect test track, I'd want to add some elements from Spa. Firstly it needs more, longer, and more complex high speed curves to extend peak lateral accelerations and the low lateral/positive longitudinal combination cases, as well as some more significant elevation changes to test stability in atypical suspension bump positions. Some more data in the lateral/negative longitudinal area would be nice too, but it's already matching if not outperforming all three other tracks here. The sort of corners that lead to this aren't good for promoting overtaking either, so it's a compromise. It's just going to be a case of needing more laps to fill in the data better.
Google Sites
Report abuse