In-Depth Look at the Latest Tyre Model

Back in EA 0.4 we tweaked around the root problems, which we couldn’t address at the time, but which gave a generally better, however still wrong result. We traded inherent understeer for sensitive throttle pickup and an overall more "floaty" behavior, allowing and wanting a lot of slip to produce grip.

Generally the perceivable problems trace back at least to ACC, if not even AC. So let's dive into what we were up to for this patch.

Disclaimer: All changes made apply solely to the GT slicks on cars from Mazda MX-5 Cup to Ferrari F40 LM, including all GT2, 3, 4 and Cup cars, but not F1 or road compounds. You will likely want to start over with setups. Time permitting I will also update default setups - please bear with me. This tyre model was deployed to Assetto Corsa EVO in Update 0.9

Yaw Moment

Racing a car on and around the limit in tire research is an edge case. Yet, it's almost the only case we truly care about in a simulation, because this is where all the fun and frustration happens. Does the tire communicate with you, does it behave intuitively and reliably and does it do all these things correctly?

One of the most important edge cases is on race cars with a rear limited slip differential, that ties the rear wheels together on power to prevent energy being lost through the inside, unloaded wheel and instead drives the torque through the outside wheel. This difference in torque between the inside and outside rear wheel creates a "yaw moment". Think of it as putting a vertical stick through the center of the car from roof to floor. Then you go to one side of the rear bumper (say, right rear corner of the car) and push forward. At the same time imagine someone pulling on the left rear of the car. This force difference will rotate the car around the tooth pick in the center - does that make sense? I hope so.

Now, various effects have an impact on how much friction a tire produces and how much force it will be able to exert into the ground. Load, slip, camber, speed, temperature, pressure, toe... and countless other things. For the above yaw moment to take place, A LOT of things need to work exactly right. We have basically reviewed almost every factor involved and verified their impact on this particular edge case that makes all the difference to the behavior of the car. You will especially notice this in 1st to 3rd, sometimes 4th gear corners. Above that, the effect starts to wane as downforce increases the inside rear tire's grip and due to general performance degradation due to speed.

Response

The workarounds also forced us to use tires with rather high peak slip angles, which in turn results in slow response to steering inputs, as you first have to turn quite a bit to reach the peak grip of the tire. Also the rear end has to start moving sideways before it can produce the grip at higher slip angles. While the rear starts moving, the car’s inertia might be too big for the tire to catch it again when the slip angle finally reaches its peak grip, and the rear always remains in some sort of sensitive float and slide state.

Due to various improvements done, we were able to lower the tire's slip angle needs and as a result the response is much sharper.

Grip loss

Equally important as grip is not having grip, or rather, how you end up not having grip. The previous calibration led to overly lenient tires that would keep producing more grip the harder you drove over the limit, which made TC pointless and didn't penalize too much steering. We have already worked in that area, mainly by adding dials and maths to better control the behavior of the slip/grip curves in various driving situations. This is not yet completed, but already better than before. You might still drive with TC off in Q, but in a race, I'm not sure if you would dare.

FFB

Changes to FFB really come as a side effect. But we have also made a tiny change to the scaling of the self alignment torque, which makes oversteer much more noticeable and generally provides higher FFB forces.

Camber

Camber has various effects on the tire. It softens it, it increases lateral grip, it decreases longitudinal grip, it increases the tire’s response. Apart from solving a bug that led to the inside cambered tire to produce a lot of longitudinal friction, we now have separate impacts of camber on longitudinal and lateral grip, that we can also control individually. At least in some of my testing I found that you can feel the trade off with high cambers where you might gain cornering speed, but you also lose traction and braking capacity. We can now easily tweak this further. This marks a major improvement over ACC, where maximum negative camber was always the right path without any trade-offs.

Combined Grip

A lot of work went into shaping curves. How various effects impact one another, especially in regards to the famous "traction circle". Which just means: How much grip is still there longitudinally when you already request maximum grip laterally - and vice versa. Further, the tire does not only change its capacity, it also changes where this capacity is found. Higher load on the tire shifts the peak grip to higher slip angles, the correct temperature makes the peak wider, speed makes the peak narrower and sharper, camber shifts the peak to lower slip angles. We had several effects previously that shifted the peak further and further away, basically to unattainable slip levels, which led to a "more slip = more grip" sensation. This was, simply spoken, down to the grip/slip curve having an anchor in the far distance that wasn't moving down with the rest of the curve. We’ve thrown that overboard.

To make cars rotate with the 0.4 tire we had to resort to rather low amounts of combined grip. All changes done this patch contribute to us now being able to increase the combined grip again. Thus, the tire will offer a wider range within which it allows to tap into more longitudinal and lateral grip at the same time before giving in. This reduces the overall sensitivity on corner exits, and offers a more rewarding range to play with as you approach the limit.

Contact patch

There's also a new logic that better accounts for contact patch variation in all situations. In any given situation the load spread on the contact patch has further influence on where the grip peak is found, adding to the overall dynamic of the tire. For example higher camber might make for a responsive tire, but in a low load situation the contact patch won’t be fully utilized and thus the camber offers less lateral grip than you expect. Only once load increases, the tire is deformed enough and the contact patch reaches optimal utilization allowing you to extract the most grip from the chosen camber angle. We think this might lead to tracks and car combinations where more camber will not automatically be the right choice, as the track profile does not benefit from always using maximum camber. This likely needs further tweaking and we are looking forward to hearing your experience.

Overall the work on the tire for this patch has gone very deep and touched on the very fundamentals of the model in order to better understand its emergent behavior and to better review and debug the contribution of individual factors. We are much happier with this tire than previously, but work will always continue as there are still several topics that need review, debugging, decisions and ultimately implementation and testing. Among other things, we will keep reviewing the impact of tire pressures, temperatures and how they build, tire wear and various other parameters that can be summarized as damage to the tire like graining, blistering or flat spots.

Not only on tire level are we continuing to push forward. We are already experimenting with improved differential modeling and other drivetrain components, but are also planning to make adjustments to aerodynamic behavior or slip streaming. The variety of content we model often requires bespoke solutions for individual cars (e.g. torque vectoring algorithms or magnetic dampers) which leads to an ever increasing toolkit to accurately simulate the entire variety seen in automotive engineering.

A big thanks goes out to our physics programmer `Fernando`, who had to deal with me for months and rip the code apart time and time again and never flinched once.

We hope you like the improvements and continue to provide bug reports and feedback in order to push and expand the limits of our physics engine.