Paul Jeffrey

Premium
ACC Blog Post 3.jpg

Aris has been at it again, revealing his third Assetto Corsa Competizione blog post - this time looking at the world of aerodynamics in ACC...


Informative blog post number 3 has been released! @Aristotelis has been at it again, finding some spare time to put down in writing more information about the inner workings of the upcoming Assetto Corsa Competizione racing simulation from Kunos Simulazioni.

In this latest post, Aris speaks in detail about the aerodynamics featured in ACC, and how they work and what effect they might have on the player when loading up the soon to be released new title.

If you missed them earlier, Aris has also posted about physics and tyres from the new title, informative reads and well worth a look as we kill time prior to the big release...

You can read the full blog post below:

Wings? Where we going we don't need any wings...
In original AC we could simulate as many wings a we wanted. So you wanted to simulate the body aerodynamics? Put a wing and apply proper values. You had a splitter? Add a wing with different values. Rear spoiler? Add a wing. Diffuser? Vertical fins? Side boards? Buffets. Add a wing. Each wing could have a specific area, lift coefficients, drag coefficients, sensitivity to ride height, yaw sensitivity… even active telemetry inputs that could modify the angle in real time and simulate all those active wings moving around… pretty impressive stuff, many industry firsts and so on. We were quite proud of the aero system in AC…

Great, scratch that and let’s build something completely new.

The aerodynamic model for ACC is completely rewritten from Stefano and Fernando. It doesn’t use an unlimited amount of wings anymore, but takes aeromap data from wind tunnels or CFD and uses them in way that everything in the car body, influences everything.

Ok let’s take a step back and analyze what is happening in real life and why such a development was needed. In real life an object, a racing car in our case, travels through air which is a fluid. Flow and turbulence is created and the shape, size and speed of them is relative to the shape of the object. Now comes the “obvious” but interesting part, a change in the shape of the object, changes the flow and the turbulence. A racing car is a big object and features on the body of the car, change the flow and create turbulences that affect a different feature in a different place of the body. A different front splitter, will not only change the aerodynamic characteristics on the front of the car but will also change the flow at the rear wing and rear diffuser, changing the aerodynamic forces of those features. A diffuser of a different shape or working in a different angle, will also change the flow behind of the rear wing and so the air approaching it, again changing the forces of the wing. A different rear wing shape or angle, can create pressures that change the flow from the front of the car, actively modifying the forces at the front.

Nothing of the above would be possible to simulate with the old system, at least not without some nasty workarounds and side effects that could be taken advantage by players in order to produce unrealistic end results.

The new aerodynamic modelling does exactly what we need. The rear wing can affect the front downforce, the front splitter can affect the rear wing and diffuser, but most importantly the ride heights now become the most important part of car setup. Modern cars that generate big amounts of downforce by the use of splitter and diffusers, are very, let me repeat that, VERY ride height and pitch sensitive.

The difference between front and rear ride height is of massive importance. Changing the front e rear ride height, moves forwards or backwards the aerodynamic pressure point, or to say it in a more simple way, how much of downforce goes to the front or to the rear of the car. Why is this so important? 1mm of ride height change at the front, can move the aerodynamic balance forward or backwards by 0.5% to over 2% depending on the car and rear wing setting. Let me say that again. 1mm of ride height, equals up and over to 2% of aero balance shift. You think it’s not much? 10mm can mean way over 20% aero shift. Good drivers will easily understand changes of 1%. Average drivers might not, but they will have big instability problems with a change of over 5%. Remember, these cars generate way above 500kg of downforce at medium speeds, so that’s a lot of force shifting around.

Furthermore we are talking about static balance. While driving, the car also pitch and rolls. Lifting off from the accelerator at a high speed turn or braking hard while turning in, can have catastrophic results in car stability if the aero “platform” is not setup properly.

It’s not over yet. With speed, the pressure created in front and under the front splitter might change the flow to the rear diffuser or make the front splitter more efficient, thus moving the aero shift even more forwards. This means the car might change balance going faster.
It is clear that splitters and diffuser are very sensitive to ride height. Finding the correct ride height and trying to maintain it in a acceptable range under all conditions is what we call “maintaining the aero platform stable”. One big aid comes from the rear wing. The wing doesn’t travel close to the ground and the flow it receives is relatively clean so its aerodynamic characteristics remain stable. The more rear wing you use, the more you average the pitch and roll sensitivity and you stabilize the car in various conditions. The balance shifts less forward and backwards under pitch and squat. Unfortunately the rear wing adds a lot of drag and of course shifts the whole balance to the rear which brings understeer. So again it’s a compromise but a much more complex one than before. You can use less rear wing angle, but then the car becomes unstable. You can use more rear wing angle but the car becomes understeery as balance, but gains grip because the whole downforce raises.

We’re not done yet. Normally in sims, a change in a wings angle, roll, speed will instantly translate in an equivalent change in aerodynamic effect. For example let’s say we brake hard for a turn and this translates in a pitch of an extra 1° of angle. All the aerodynamic wings will instantly start to generate downforce and drag for an extra 1° of angle. But that’s not what happens in reality. The air can be compressed, the flow can get slower or faster, or can create “pumps” from pockets of air that get compressed and uncompressed. All of this takes tiny bit of time. It acts as if there’s some small amount of lag in the results and, yes you guessed it, it is now simulated in ACC. Sometimes this effect can be your friend, like when you start braking from high speeds it might take a fraction of a second before the balance shifts way forward and by that time your speed is slower. Sometimes it can caught you off guard, like when you’re jumping in a kerb and you think you cleared the first part, only for the aero instability to hit you hard in the second part!

By now it is pretty clear that the aero model is advanced and complex, but how we control the aero platform in an efficient way, except from adding tons of rear right wing? That was the same question I asked myself when for the first time I configured the values and went for a first ride. I’m famous for my self control so my exact words have been: “WTF what have we done?!?!!?”.

Right. It was evident that we needed a much more stiff front suspension to limit pitch. Again remember, we’re talking that we are trying to avoid mere millimetres of ride height change, let alone macroscopic movements! So we’re gonna use bumpstops right? But nobody likes bumpstops, they make you car understeery and they make your car jump on kerbs.. bliah! So we simulated variable stiffness for bumpstops. And more precise bumpstop range. And better control of their placement when you change ride height. And then we had to control the suspension from this all, so in goes a more advanced damping system… and then we realized that we had to do something to help people setup this whole thing… Remember? In AC you can change a bit of ride height, or fuel load and then you need to change again all the alignment to bring it back to your original chosen values. Nightmare. But that needs a different post, something regarding suspensions and setup…

I told you we worked a lot. I’m starting to realize it from the amount of stuff I need to write for every damn post!

More to follow in the coming days...

Assetto Corsa Competizione will be available to purchase on Steam Early Access from September 12th 2018.

Check out the Assetto Corsa Competizione here at RaceDepartment for the latest news and discussions regarding this exciting upcoming sim. We intend to host some quality League and Club Racing events as well as hosting some great community created mods (we hope!). Join in the discussion today.

Like what you see here at RaceDepartment? Don't forget to like, subscribe and follow us on social media!

 HYPED! What do you think of the latest blog post on the development of ACC? Do you like the sound of the direction the new game is going? Let us know your thoughts in the comments section below!
 
Last edited:
They are correct. But they are going to make their lives (if not already) very painful when adjusting the aero data to real world data.

I don’t agree that aeromaps are the holy grail of aerodynamics modeling. Actually far from it.

And I do believe they should have started from scratch. But regarding to the aero model, there are much clever ways of achieving the necessary accuracy (or “simulation”).

Better yet, it’s much more easier to match the aero model without an aeromap.
 
They are correct. But they are going to make their lives (if not already) very painful when adjusting the aero data to real world data.

I don’t agree that aeromaps are the holy grail of aerodynamics modeling. Actually far from it.

And I do believe they should have started from scratch. But regarding to the aero model, there are much clever ways of achieving the necessary accuracy (or “simulation”).

Better yet, it’s much more easier to match the aero model without an aeromap.

I would think they have the full aeromaps from manufactures/racing teams.

Also from ac everything ran with lut (lookup table) anyways.I bet that will continue with ACC.

Doubtful they would guess/estimate what downforce/drag etc car is doing with this load/wing angle/pitch/yaw etc without proper data.

In another post they said apps like wings we have in AC now will not happen in ACC as they are under NDA.
 
I would think they have the full aeromaps from manufactures/racing teams.

Also from ac everything ran with lut (lookup table) anyways.I bet that will continue with ACC.

Doubtful they would guess/estimate what downforce/drag etc car is doing with this load/wing angle/pitch/yaw etc without proper data.

In another post they said apps like wings we have in AC now will not happen in ACC as they are under NDA.

But that’s exactly my point. Even when we get wind tunnel data that is representative, for instance on aircrafts, of the flight Reynolds there are effects that the WTT does not show you. And those are quite expensive tests that are the most precise available today.

Aeromaps, on the other hand, does not work like this. They are, somewhat, less precise. Because you’re assuming that your simulation is correctly calculating, for instance, drag in turbulent flow. Currently there is no numerical method that can estimate/calculate aero data with such complexity as in a turbulent flow in real time, or anything close to it. It simply DOES NOT work this way. It’s a limitation from our existing knowledge.

What I’m concluding here is that a lot of people have strong opinions without the proper knowledge of how things work.
 
I would like to see some real data. Here are three quick tests and results that would impress me very much:

Since they have close relationship with a series, and presumably access to all kinds of information, you could gather a shitload of data. If not, basic instrumentation not available from the existing telemetry could easily be installed.

1) Longitudinal acceleration:
From a steady condition, given throttle position, all other relevant parameters set to 0, apply a step input on throttle. Plot pitch, roll, ax (acceleration in longitudinal body axis), ay, az, speed.

Repeat for at least 3 different speeds and 3 levels of downforce

Repeat for at least 3 different brake inputs

2) Lateral acceleration
From a steady condition, given throttle position, all other relevant parameters set to 0, apply a doublet input on wheel. Plot pitch, roll, yaw, ay, az, speed.

Repeat for at least 3 different speeds and 3 levels of downforce.

Change step input to a controles turn input.


3) combined lateral/longitudinal
Rinse and repeat, but with larger coupled inputs.

Since there are some heavy proprietary data, I would accept data range erased from the y-axis plot
 
But that’s exactly my point. Even when we get wind tunnel data that is representative, for instance on aircrafts, of the flight Reynolds there are effects that the WTT does not show you. And those are quite expensive tests that are the most precise available today.

Aeromaps, on the other hand, does not work like this. They are, somewhat, less precise. Because you’re assuming that your simulation is correctly calculating, for instance, drag in turbulent flow. Currently there is no numerical method that can estimate/calculate aero data with such complexity as in a turbulent flow in real time, or anything close to it. It simply DOES NOT work this way. It’s a limitation from our existing knowledge.

What I’m concluding here is that a lot of people have strong opinions without the proper knowledge of how things work.

No.

If you knew anything how AC system was simplified you would realize that this is improvement.

Just go take any kunos road car it was defined by two wings at best.


You have made 4-5 mentions how you work in aircraft industry simulation we get it great.

What does that help with 30$ video game.
 
And I forgot to mention. Do this with 2 aero cars. One with normal configuration and another with a round plate of radius 15cm in front of one of the rear wingtips.

That would be awesome.

I would assume that they’re using the same approach of x-plane modeling.
 
No.

If you knew anything how AC system was simplified you would realize that this is improvement.

Just go take any kunos road car it was defined by two wings at best.


You have made 4-5 mentions how you work in aircraft industry simulation we get it great.

What does that help with 30$ video game.
Because simulation is simulation. Either it a space rocket, airplane, car or a banana boat. Simulations are based on modeling. The art of simulation is modeling. And race cars works exactly like an aircraft. We do have to model ground handling. That includes modeling all the suspension system, tyre deflection and all that **** that some dudes like to praise. And regarding aerodynamics, aero industry is the state of the art technology.

Please don’t take me wrong, I do think they do an amazing job.

What you guys should understand is that, a simplified model is not less simulation than a “omg let’s try to reproduce all the effects with bizarre approach”. And often, a simplified model is a better approach.


I’ll try to give a very basic example here. Remember when you open excel, got some data and tried to fit it? Remember you were tempted to increase the polynomial order just to see it fit better? But you could fit the data with a smaller order polynomial? That’s my point.

I’m really amazed by KS job. I just don’t like the tone used on some blog posts that are not strictly technically correct. If they claim that they can solve aero data in real time just based on the geometry... dude, we have some pretty amazing thing happening that the whole science industry is not aware of.

I just want a fun game. Today is 12!
 
What you want is hilarious to be honest.

This is a 30$ game.

I bet you anything that no team would give Kunos or any racing game that kind of data.

That's not what they claim. All sim companies claim to have access to real world data.

But I'll assume you're correct. Let's say that you don't have access. How can you PROVE that your SIMULATION is more SIMULATION when you have nothing to compare to? After all, we're talking about a simulator company (that includes reiza, studio 397, etc), right? Specially ones that sells professional products.

AFAIK, KS is based near a racetrack. They could validate their data with basic instrumentation that you can buy off the shelf. Basic accelerometers (that any modern car have), an inertial reference unit and gps data. A few strain gauges and lvdts to measure pressure and position of input controls (apart from what the ECU should give you). Take it one week to install into a regular modern car.

The tests could be run in less than 2 hours with a skilled driver. Another day to analyze and plot the data.
 
@Augusto Dufloth I think you’re misinterpreting things. The new system is actually somewhat less complex than the old one. The main input is a 3D lookup table with ride heights as its x and y axes. Nothing particularly complex. And the tests you ask for would tell much more about the tires than the aero. As you (should) know, tire grip coefficients are not constant with load, and actually changes quite drastically, so aero effects cannot very effectively be deduced from longitudinal/lateral acceleration. Force sensors on the suspension, yes. That’s how validation of wind tunnel/cfd data is done in real life.
 
Last edited:
@Augusto Dufloth I think you’re misinterpreting things. The new system is actually somewhat less complex than the old one. The main input is a 3D lookup table with ride heights as its x and y axes. Nothing particularly complex. And the tests you ask for would tell much more about the tires than the aero. As you (should) know, tire grip is not constant with load, and actually changes quite drastically, so aero effects cannot really effectively be deduced from longitudinal acceleration. Force sensors on the suspension, yes. That’s how validation of wind tunnel/cfd data is done in real life.

Yeah I think so too, we have to understand that now they are marketing the game and they need to make it sound exciting. It's also good if they made it simplier (I don't know the details but I believe you), as it's possible to get similar results with simpler models/methods.
 
@Augusto Dufloth I think you’re misinterpreting things. The new system is actually somewhat less complex than the old one. The main input is a 3D lookup table with ride heights as its x and y axes. Nothing particularly complex. And the tests you ask for would tell much more about the tires than the aero. As you (should) know, tire grip coefficients are not constant with load, and actually changes quite drastically, so aero effects cannot very effectively be deduced from longitudinal/lateral acceleration. Force sensors on the suspension, yes. That’s how validation of wind tunnel/cfd data is done in real life.
I totally agree with you. If they’re using some LUT, awesome. That’s not what ari said in his post. He claimed something totally different.

But your assumption is: the tyre model is not correct.

My assumption: the tyre model should be very good by now. Which usually is (and a lot of people like to praise that). Tyre model can be adjusted doing slalom, eight curves, coast (in different kinds of loads and sideslips. And btw, that’s what I’m modeling/matching in the past two weeks). That’s how you match tyre data after you get manufacturer data.

So, the delta response would be accounted on the aero, not the tyre.

I’m questioning the claim that they can simulate aeromaps based on geometry during the simulation.

Anyway, it’s here! Let’s enjoy
 
Last edited:
I totally agree with you. If they’re using some LUT, awesome. That’s not what ari said in his post. He claimed something totally different.

But your assumption is: the tyre model is not correct.

My assumption: the tyre model should be very good by now. Which usually is (and a lot of people like to praise that). Tyre model can be adjusted doing slalom, eight curves, coast (in different kinds of loads and sideslips. And btw, that’s what I’m modeling/matching in the past two weeks). That’s how you match tyre data after you get manufacturer data.

So, the delta response would be accounted on the aero, not the tyre.

I’m questioning the claim that they can simulate aeromaps based on geometry during the simulation.

Anyway, it’s here! Let’s enjoy
There’s no claim that they base it on geometry; it’s just a LUT from what aris said.

And assuming tire model accuracy is going down a dark path...far too many variables to verify accuracy, which can be seen by the fact that AC and ACC drive quite differently.
 
So, we are converging on a solution! :roflmao: If it's just LUT, then it's fine.

and come on, ground handling it's not that difficult to show some plots :sneaky:

Haven't tried yet. It was 1am when it launched... now a whole day of work ahead. Oh boy. I want the whole experience to be awesome
 

Latest News

Are you buying car setups?

  • Yes

  • No


Results are only viewable after voting.
Back
Top