iRacing Physics (Brake Bias in Dallara P217)

This is a continuation from the Rennsport thread, so that the moderators don't get further annoyed about thread drift. Here's the original thread:


The iRacing physics discussion started from page 7 and by the end centred on exploits of unrealistic brake bias in the Dallara P217 by top-level drivers in the IMSA iRacing Series. It's not a private conversation between MD11 and me, so feel free to contribute, anyone who is interested.


@MadDriver11 wrote: "There is no need to fill your mouths with big telemetry screens where you apparently ready only what you want and ignore what you don't like"

Do you apply the same standard to yourself? It appears you completely ignored my telemetry because it doesn't align with your own opinion. Did you even look at it properly? I've shown to you telemetry indicating that at least one iRacing car does not have the fundamental problem with combined lateral and longitudinal grip that you claim.


But progress has been made. You've now given some definite SWA numbers, as opposed to "almost zero" in sim or "they actually turn the steering wheel" in RL (both clear examples of hyperbole). I have something to compare in RL (45 deg going into the chicane) and the sim (Morad etc never reach 20 deg) for the Mulsanne chicanes. So thanks for that and the RL video link. I will go study them.


How about more generally at Le Mans or is this only a problem at the chicanes? If there was the fundamental problem you claim, it would be visible in major corners also, would it not? Have you got SWA numbers for some of them, or is that work I have to do?


Also, you didn't answer my question. What do you regard as a reasonable brake bias range in RL for a Dallara P217 driven at Le Mans? IIRC you have a degree in automotive engineering, so surely this is a question you can easily answer? Do you agree with Jackson's opinion: "53-56% as far as I know" - thanks for the reply @mclarenf1papa and I see that's a substantial difference to what we currently see the sim drivers exploiting, so good to have this highlighted.

As I said in my post in the Rennsport thread, this seems more likely to indicate that iRacing have a dynamic balance problem, rather than a problem with grip/force combining. I pointed to a previous example of this in iRacing in that thread, so perhaps the problem is still there?
 
@MadDriver11 @mclarenf1papa or Moderators - if someone could add a link to this thread from the Rennsport thread please, I would appreciate it. TIA.

Moderators - What is the chance of getting a general race simulations sub-forum, for discussing race simulation issues like this generally? I think if you had one, it might reduce the level of off-topic chat seen in news threads.
 
Here are the videos:

Real life:


Daniel Morad driving the Dallara P217 at Le Mans in iRacing:


Does anyone have a RL Dallara P217 video for Le Mans? Yes, I am too lazy to search :D
 
As I said in my post in the Rennsport thread, this seems more likely to indicate that iRacing have a dynamic balance problem, rather than a problem with grip/force combining.
I would counter this by saying the dallara (at least last I drove it) definitely has too little combined grip.

The truth is that none of the available P2/DPi (or LMP in general) cars in sims are particularly accurate to their real counterparts. These cars are very complex to simulate properly and require things to be simulated that even professionals in motorsport don’t typically consider, let alone sim/game devs. Once cars start making downforce, the level of difficulty of simulation goes up exponentially (because everything matters more; tire model accuracy, suspension model accuracy, etc).
 
@MadDriver11 - I have analysed your suggested videos: RL Ferrari 488 GT3 (2019) at Le Mans and Daniel Morad driving the iRacing Dallara P217 at Le Mans.


Analysis

See attached table (PDF). It shows the peak SWA achieved for both of the Mulsanne chicanes (all three corner phases, RLR and LRL). All video estimates are to the nearest 5 deg, but analysis of the RL video is very difficult because of poor visibility of the steering wheel - it is deeply shadowed with much of the wheel outside the camera's field of view.

As you did not come back to my question about expected SWA for some of the major turns, I have also added two of them: Mulsanne (right-hander) and Indianapolis (left-hander).

I have also included video timestamps, for ease of checking each entry in the table.

Also included are peak SWAs (rounded to nearest 5 deg for consistency with the video analysis) for the fastest lap in Martin Krönke's VRS 2022 Season 2 datapack; refer back to the Rennsport thread for the telemetry plots. Replay timestamps are also included for those who have access to the VRS service. Martin is another top-level driver in iRacing and he ran brake bias 48.5% for his lap.


Conclusions

Based on this evidence, your claim of the sim racers that "they won't reach 20 degrees" on entering the chicanes at Le Mans is clearly and obviously untrue. The entry SWAs range from 30-50 deg.

It is apparent that the RL racer generally uses more steering lock than the sim racers.

However, we don't have full knowledge of the steering ratios used for each car:
  • The 488 GT3 has a fixed steering ratio, but there's no data to confirm the ratio visible in the iRacing Garage (Legacy Ferrari 488 GT3, the car driven in the RL video) and I've not been able to find confirmation of the ratio from web searching.
  • The Dallara P217 has a variable ratio: 9.5-12.0 according to the iRacing Garage.
  • We do know the exact steering ratio used by Martin (11.0), but we do not know the ratio used by Daniel.
  • So this is another source of uncertainty when comparing RL and sim SWAs in the two videos.
I'm guessing that the 488 GT3 ratio may be slower than the P217 (more SWA for a given steered angle at the tyre).


Further Questions For You

Do you have a better RL Le Mans video for study, with a clearly visible steering wheel please? Ideally for a P217, because that's what we're trying to compare RL to in iRacing!

Out of interest, do you actually use telemetry in RL and/or sim, or is your working method entirely video-based?

What is your opinion of the SWA numbers for the two major turns in my analysis?


Brake Bias

If you accept Jackson's estimate of P217 brake bias range 53-56% as being reasonable for RL use, then sim racers exploiting more rearward settings may indicate a significant problem in the simulation. Daniel was driving with 48.4% and 49.0% bias in the laps analysed and Martin with 48.5%.

However, it's worth pointing out the brake bias ranges available in iRacing:
  • Legacy Ferrari 488 GT3: 45.7-61.6%
  • Dallara P217: 42.0-65.0%
These are presumably reflective of what's available on the RL cars, given that iRacing seem quite obsessive about modelling such fine detail when they build their virtual cars.
 

Attachments

  • SWA Analysis - Le Mans - P217 v 488 GT3.pdf
    406.7 KB · Views: 73
Last edited:
The Dallara P217 has a variable ratio: 9.5-12.0 according to the iRacing Garage.
Rack is homologated in regs, so very unlikely adjustable.

These are presumably reflective of what's available on the RL cars, given that iRacing seem quite obsessive about modelling such fine detail when they build their virtual cars.
That P2 range is pretty crazy, most cars IRL don’t go more than +-5% with the knob. You can change the master cylinders around to get more range, but as far as on-track stuff goes, 23% is wild.

Also my number’s not really an estimate. Saw an onboard where 54% is more or less visible on the steering wheel, and assuming the brakes are “square” on the dallara (like the oreca, meaning same sizes for everything front and rear), the dallara’s more rearward weight distribution would put it into the range I mentioned.
 
Many thanks for the clarifications Jackson, very helpful.

Then my choice of the word "estimate" looks off and we should use something along the lines of likely to be in the range, based on onboard video evidence.

Regarding the rack, do you have access to the regs for confirmation? Also, would the 488 GT3 seen in that video be constrained to the same ratio as a P217 competing in the same series?

Sorry for the dumb questions. These are not series I follow in RL; my main interests are in open wheel/formula competition.

Regarding bias adjustability on-track, I haven't checked that. The ranges quoted are from the iRacing Garage, so presumably they are fully reflective of master cylinder options in RL.
 
Last edited:
Regarding the rack, do you have access to the regs for confirmation? Also, would the 488 GT3 seen in that video be constrained to the same ratio as a P217 competing in the same series?
It would be in the FIA homologation documents and is different per car. Oreca rack, for example, is 73mm/360deg which gives you a circa 13:1 steering ratio, +- a bit for u-joint reasons. 488’s is 75mm/360deg, but not sure what that works out to as a ratio since it’s suspension geometry dependent and I haven’t made a model for that car.
 
I would counter this by saying the dallara (at least last I drove it) definitely has too little combined grip.
I would be interested in your latest thoughts if you drive it again.

IMO there was a very significant tyre model update in 2021 Season 4, based on the cars I drive regularly. It felt very obvious to me when getting the cars severely out of shape and especially for a full sideways slide.
 
This is a g-g diagram (produced using McLaren ATLAS) for a Formula Vee driving Tsukuba Short. Not driven by Jake Burton this time, but by another top-level iRacer. It's from a few seasons back (2021 S3), so not the very latest tyre model and would not include the significant tyre model change I feel that I detected in the 2021 S4 release.

The plot clearly shows the two hairpins for this track and there is strong combined longitudinal and lateral grip for both turns during the trail braking phase.

GG Plot - 2021 S3 - Tsukuba Short - Formula Vee - crexLive.PNG
 
Last edited:
Another g-g diagram. Formula Vee at Tsukuba Short on the current tyre model. Driver is me; I am roughly a top 5% driver, based on lap pace. The plot is for all 9 laps driven in the stint (previous post plot was one lap only).

GG Plot - 2022 S2 - Tsukuba Short - Formula Vee - BT7 Driver.PNG
 
Last edited:
1654565910926.png

Not sure how representative this data is since Tsukuba has a lot of camber which really messes up g-g diagrams (for use in absolute terms, anyway). But the line I drew in red should not ideally be straight if the car doesn't have notable downforce. That would show a lack of combined grip under high longitudinal (specifically braking) force. Not possible to tell too much on the accel side since power's a bit low.

Again that may not be true for the aforementioned reason as well as driver ability and the car itself being factors.
 
Thanks Jackson. Yes, agreed, the track camber in the hairpins doesn't help, but my purpose with these is to further counter the ludicrous (IMO) claim that:

"No matter how many iterations iracing made, the issues still remain today: there is no combined longitudinal and lateral grip available. As soon as you use any grip in a given direction there is none left in the other."

The third hairpin at Tsukuba Long is notably flatter than the first two, so I will see what I can produce for that one. ATLAS has an option to join the scatter plot dots, which can make interpretation a little easier for full laps. I also have a method to constrain plots to single track sections, but it requires some work to customise the segments.
 
On reflection, I think Tsukuba Long hairpin 3 is quite cambered, like the other two hairpins.

FWIW, here's driver input and gg telemetry for the top driver zoomed to show only hairpin 3 in his fastest lap.

The data cursor is positioned at the start of braking in both plots. gg plot has the data points joined this time.


Driver Inputs - 2022 S2 - Tsukuba Long - Formula Vee - crexLive - Fastest Lap - Hairpin 3.PNG
GG Plot - 2022 S2 - Tsukuba Long - Formula Vee - crexLive - Fastest Lap - Hairpin 3.PNG
 
View attachment 572403
Not sure how representative this data is since Tsukuba has a lot of camber which really messes up g-g diagrams (for use in absolute terms, anyway). But the line I drew in red should not ideally be straight if the car doesn't have notable downforce. That would show a lack of combined grip under high longitudinal (specifically braking) force. Not possible to tell too much on the accel side since power's a bit low.

Again that may not be true for the aforementioned reason as well as driver ability and the car itself being factors.

What's the dynamical explanation of that straightening for a downforce car?

Also, do you have a link to the onboard in which you saw 54% bias more or less visible?
 
Some more plots, same top-level iRacing driver, MX-5 at Okayama Full, zoomed into turn 1. Two laps, gg diagram then driver inputs, but this time I have taken care to use the same scale factors for lateral and longitudinal G.

GG Plot - 2022 S2 - Okayama Full - MX-5 - crexLive - Lap 2 - Turn 1.PNG
Driver Inputs - 2022 S2 - Okayama Full - MX-5 - crexLive - Lap 2 - Turn 1.PNG
GG Plot - 2022 S2 - Okayama Full - MX-5 - crexLive - Lap 1 - Turn 1.PNG
Driver Inputs - 2022 S2 - Okayama Full - MX-5 - crexLive - Lap 1 - Turn 1.PNG
 
What's the dynamical explanation of that straightening for a downforce car?

Also, do you have a link to the onboard in which you saw 54% bias more or less visible?
You're not turning in the first part of the braking zone (where you have the highest decel rate due to downforce + drag) so it'll form a bit of a teardrop shape instead of the more circular shape you can see on the mx5 data you just posted.

Can't find the video again. It was spa in the wet and the bias showed on braking only. Anyway, the Oreca typically runs 55-58 in real life, so 54 for the Dallara is pretty reasonable (Dallara weight distribution is a bit more rearwards - again this is assuming the discs/pistons are square front to rear like on the Oreca).
 
Last edited:
Does anyone have an opinion on the car yaw angle (CYA) profiles for Formula Vee versus MX-5?

The profile for the MX-5 is along the lines I expected, whereas that for the Formula Vee (peaking then flattening while still turning) is not. There are a few other cars with the latter characteristic, one very extreme (or at least it was a few seasons back), but the Formula Vee example is good enough to illustrate the point.

Just to expand on my use of CYA - it's expressed in the vehicle coordinate system. In most of the vehicle dynamics literature, the vehicle sideslip/slip angle is used in preference. CYA is simply the inverse of that angle. I prefer to work with CYA as it is signed in the same direction as the vehicle turning, so you can more easily compare CYA trends with SWA trends.
 
Last edited:
It would be in the FIA homologation documents and is different per car. Oreca rack, for example, is 73mm/360deg which gives you a circa 13:1 steering ratio, +- a bit for u-joint reasons. 488’s is 75mm/360deg, but not sure what that works out to as a ratio since it’s suspension geometry dependent and I haven’t made a model for that car.

Is this the correct document?

https://www.fia.com/sites/default/files/2022_lmp2-2017_wmsc_2021.12.15_published_2021.12.16.pdf

I can't find anything related to steering racks.

Or is that specified/constrained elsewhere?
 

Latest News

Online or Offline racing?

  • 100% online racing

    Votes: 103 7.9%
  • 75% online 25% offline

    Votes: 134 10.3%
  • 50% online 50% offline

    Votes: 189 14.5%
  • 25% online 75% offline

    Votes: 366 28.0%
  • 100% offline racing

    Votes: 510 39.0%
  • Something else, explain in comment

    Votes: 5 0.4%
Back
Top