Telemetry Data?

[KeithYoung]

Does anyone have data that shows G's attained by F1 cars through a speed range? Even if it's and old F1 car. Preferably braking. Traction circles at a range of velocities would be absolutely ideal.

Also, some of you may play rFactor, which I understand may have the capacity to measure telemetry and output it as a .csv or something. It's much more likely someone has data from this than happening to have data from an old F1 car.

I'll be writing an article on it and will credit the source of the data I use if you want.

Thanks

[Tim.Wright]

Of course there is the famous lap of Spa that Lewis tweeted.

Otherwise Caterham post lap data in their race previews:


Obviously its not going to be so accurate, but I don't know what you want to use it for

Tim

[Blanchimont]

Some theoretical formulas that give you an idea how accelerations and speed are related. Of course this is not the whole truth, but a good start to this topic.

Cornering (where v^2/r = a_lateral):

m * v^2 / r = ( m*g + airdensity/2 * c_downforce * A * v^2 ) * frictioncoefficient_corner

Deceleration, straight line:

m * a = ( m*g + airdensity/2 * c_downforce * A * v^2 ) * frictioncoefficient_straight + airdensity/2 * c_drag * A * v^2

Acceleration, straight line, rear wheel drive:

m * a = MINIMUM of "P_engine/v - airdensity/2 * c_drag * A * v^2" and "( m_rear*g + density/2 * c_downforce_rear * A * v^2 ) * frictioncoefficient_straight - airdensity/2 * c_drag * A * v^2"

And then you could put the corner and straight line acceleration together using an appropiate model (traction circle or ellipse).

Edit: forgot the drag resistance, any other mistakes?

[KeithYoung]

Thanks! I might be able to write a computer vision code and grab each pixel. Doesn't seem to have braking G's but I may be able to approximate a tire friction coefficient from this.

[Tim.Wright]

You've got speed and distance so you can calculate longitudinal acceleration from that.

[KeithYoung]

Some theoretical formulas that give you an idea how accelerations and speed are related. Of course this is not the whole truth, but a good start to this topic.

Cornering (where v^2/r = a_lateral):

m * v^2 / r = ( m*g + airdensity/2 * c_downforce * A * v^2 ) * frictioncoefficient_corner

Deceleration, straight line:

m * a = ( m*g + airdensity/2 * c_downforce * A * v^2 ) * frictioncoefficient_straight

Acceleration, straight line, rear wheel drive:

m * a = MINIMUM of "P_engine/v" and "( m_rear*g + density/2 * c_downforce_rear * A * v^2 ) * frictioncoefficient_straight"

And then you could put the corner and straight line acceleration together using an appropiate model (traction circle or ellipse).

That's what I'm sort of doing, the problem is I have to assume downforce levels at different speeds, as well as the friction coefficient. I want to get rid of the unknowns. Once I can calculate a CL and friction coefficient I'm set. I know my article on Integrals for Braking used a really crappy aero model. I want to have something semi accurate rather than totally made up.

[KeithYoung]

You've got speed and distance so you can calculate longitudinal acceleration from that.

Yeah true, if I can get those points accurately enough with the computer vision. I'll just have to derive it, but the error from that should be acceptable for what I'm trying to do.

[Blanchimont]

c_drag can be estimated from the top speed at the end of the straight, with an engine power of 750 hp and a cross section area of 1,4m².

friction_coefficient_corner can be estimated from low speed corners, where aerodynamic forces are not important.

friction_coefficient_straight can be estimated from low speed acceleration at the start.

c_downforce then can be estimated from high speed corners taken at steady speed. Google Earth gives you a rough corner radius.

Just watch some videos at youtube and F1.com and calculate some accelerations from the data.

https://www.youtube.com/watch?v=Tzj9V78tBpE
https://www.youtube.com/watch?v=hZlMZckt2m0

[KeithYoung]

Ouch the first way I'm trying this I'm getting a somewhat messy differential equation. Ended up with a correlation factor of .997. It's going to take some time to solve this one. Already got the numerical approximation, but my DiffE is rusty.

Thanks a bunch for that data Tim!

[KeithYoung]

c_drag can be estimated from the top speed at the end of the straight, with an engine power of 750 hp and a cross section area of 1,4m².

friction_coefficient_corner can be estimated from low speed corners, where aerodynamic forces are not important.

friction_coefficient_straight can be estimated from low speed acceleration at the start.

c_downforce then can be estimated from high speed corners taken at steady speed. Google Earth gives you a rough corner radius.

Just watch some videos at youtube and F1.com and calculate some accelerations from the data.

https://www.youtube.com/watch?v=Tzj9V78tBpE
https://www.youtube.com/watch?v=hZlMZckt2m0

Thanks for the tip

Tim, I've pretty much got everything working now from that. I found images with twice the resolution on Caterham's website that will come in handy. You should check those out.

Example:

http://caterhamf1.com/previews/2013/02-MAL/index.html

Almost time to start making all the pictures for the article...

[Lycoming]

There's also the piece of telemetry that Scarbs got his hands on a while ago. It's a pretty hi-res scan of a printout made at the track, so I don't think it gets much better than that.

http://scarbsf1.com/blog1/2011/08/18/te ... roduction/

[Jersey Tom]

I want to have something semi accurate rather than totally made up.

How accurate is "semi accurate" ? Worth thinking about - what fidelity do you really need to illustrate a concept? As a corollary, if you're illustrating a concept and could get really high fidelity car or aero data - will it fundamentally change the end result?

I increasingly like using arbitrary (but sensible) values for things like this. For example if someone has a good grasp on tire and vehicle dynamics, you can illustrate a lot of in depth concepts by creating even completely arbitrary empirical tire models.

[KeithYoung]

I want to have something semi accurate rather than totally made up.

How accurate is "semi accurate" ? Worth thinking about - what fidelity do you really need to illustrate a concept? As a corollary, if you're illustrating a concept and could get really high fidelity car or aero data - will it fundamentally change the end result?

I increasingly like using arbitrary (but sensible) values for things like this. For example if someone has a good grasp on tire and vehicle dynamics, you can illustrate a lot of in depth concepts by creating even completely arbitrary empirical tire models.

That is dependent on how the whole work/accuracy curve.

In the end, I found a way that got me within 5% of the data in the picture from Tim Wright. The data were not smooth since the smallest distance step was 7 meters due to the image resolution. First I did Linear Interpolation on all position data. This still wasn't perfect obviously. It assumed the position changed linearly between each original 7m step. It gave me horrible acceleration data, so I got a good polynomial that fit the position data, as well as a polynomial to fit the velocity data. I then got some acceleration data.

From that, again, I fit a polynomial to it in order to have something to integrate. I end up with a fairly nasty looking Differential Equation after all this. I may just have to use Numerical Methods since that's getting me "reasonably" close to the actual. I'll have images to post here soon once they're uploaded to my site for my article.

So yeah, at the moment I made an "aero model" from the data that is at least a polynomial. I have two ways to do it, the more direct messy way where everything is a function of position, or the way I have within 5% which is acceleration as a function of velocity, which itself is a function of time, rather than position.

If it's confusing to you, that's OK, it's still confusing to me...

But I'm close to finished, then it's down to making all the images etc and writing the article.

[Lycoming]

I think you missed the point, which is, what do you actually need this data for? What is your end goal? What resolution/accuracy do you need to accomplish your end goal?

If I wanted to make a lapsim to evaluate, say, the difference between, lets say two different engine tunes, then I might want good dyno data. But if I just wanted to use my lapsim for a parametric sensitivity study, my engine model could also be made up completely of "fake" values that cover the range of what could be reasonably achieved from a real engine. No need for real world data.

[KeithYoung]

I want an equation that fits close enough to the data that most people will agree it's based on the data. It needs to be relatively easy to handle, by that I mean integrate the velocity to get distance traveled without having a horribly ugly integral.

I'm not trying to do anything scientific with this, just write an article that fits an undergraduate level. The main thing now is making the pictures.

I already tried this before and I want to improve on it:

Applied Math Example - Definite Integral - Distance Traveled