Time for me to do some mass brain picking here. Not being terribly familiar with real cars (in a hands on, mechnical sense), I'd appreciate any tips on how to get all the data needed to simulate a real car, if you had the real thing sitting infront of you, and a garage with every tool imaginable in it, and a trained monkey just waiting instructions of what to do and what to write down.

To be more specific, I have noticed the data gathering page from the VHPA website is one of the more popular pages but I've yet to write anything in it.

I'm going through and filling in some of the easier ones now but if anyone would like to submit some good descriptions of how to obtain a value of some of the trickier properties, then that would be a great help.

Linkage: http://www.vehicle-analyser.com/gdata.html

Cheers

Physical

* Mass - Weighbridge, with fuel and driver on board.
* Body Centre of Gravity - Nasty - suspend the car in various ways, or guess. Corner weights might help pin it down a bit.
* Fuel Tank Capacity and Centre of Gravity - Drain the tank, then refil it carefully. CoG position - perhaps corner weights whilst filling? Or educated guesses.
* Reference Height - Isn't that normally the underside of the car? What you take is up to you, as long as other measurements are referenced from it. Therefore arbitrary.
* Number of Passenger Seats - Count them
* Passenger CoG positions - Maths and guesswork - geometry comes in useful...

Drive-train - back to top

* Peak torque/power - Rolling road, but this is of questionable accuracy. Engine dyno much better.
* Idle point - Look at the dashboard whilst idling.
* Limit - Ask the ECU, the manufacturer, or rev it up (with load on it) and see.
* Fuel Type - Read the filler cap.

Wheels & Tyres - back to top

* Tyre dimensions - Read and/or measure the tyres
* Rotating mass - Weight them.
* Motion Ratio - remove springs. Move wheel from full droop to full bump in small increments, and measure damper lengths at each point. Draw on graph. Deduce gradient, which is the motion ratio. May vary through travel.
* Unsprung Mass - Remove from car, and weight them. Add half of damper and spring to it.
* Track Widths and Wheelbase - Measure from centrelines of axle/tyre. Track may vary with suspension travel, roll and single wheel bump.

Aerodynamics - back to top

* Frontal area - Careful measuring and a bit of fudging.
* Coefficients of drag and lift - Windtunnel really, but if you can log ride height and plot it with speed you may be able to deduce lift. Drag needs coast down tests to get approximations - Bosch Automotive Handbook covers this if you have one. If not, get one!!!
* Aerodynamic component centre of effects - Not easily!!!

Setup Data
Brakes - back to top

* Brake torque - Very difficult, but don't MoT stations test this, or at least basically this? Could it be deduced from braking accelerations and a known mass?
* Brake balance - Statically you can try torque wrenches on the wheels with a helper pressing the brake pedal a certain amount. Again, difficult to work out. Can also look at wheel cylinder/piston diameters, and master cylinder diameters, then trying to take into account pad size and effective radius etc.

Suspension - back to top

* Spring length - Free length? Take it off and measure it.
* Spring stiffness - Take it off, put a known load on it, and see how much it compresses. Spring testing machines are good at this.
* Bump/Rebound Damping - You'll need a damper dyno. No other way.
* Anti-Roll Bar Stiffness - Complex(ish) equations involving ID and OD, lengths, effective arm lengths, motion ratios if applicable.

Steering - back to top

* Toe-In - Measure with string or tracking gauge. If you are a fool invest in lasers and suffer the same problems as with string (but without the tripping hazard).
* Caster - Caster gauge - turn the wheels 20° one way and zero gauge. Turn 20° the other way and read off gauge.
* Ackermann - Not really sure, but I suppose you need to measure the angles of each wheel at full lock relative to the centreline, then compare them to the geometric ideal to get a ratio.
* Maximum Lock - Measure it - turnplates with a protractor on them.

Gearing - back to top

* Individual Gear Ratios - Count teeth on the gears.
* Final Drive Ratio - Count teeth on the crownwheel and pinion.
* Differential Type/Settings - Look at it, and compare with the various types.

Tyres - back to top

* Tyre Compound - Read tyres, or ask the manufacturer.
* Tyre Pressures - Use a tyre pressure gauge. Hot or Cold pressures? Hot is what you want, cold is what you set. This is why tyre pressures are not a simple thing!
* Camber Adjust - You mean with suspension travel? You'll need to measure it over the range of suspension travel (spring removed).

ALSO - BUMP STEER? IN WHICH CASE USE A BUMP STEER GAUGE

Downforce/Misc - back to top

* Wing Angles - Measure them. Presumably relative to the reference plane. Can use either a straight edge across the leading and trailing edge, or perhaps work out the chord angle from harder to measure data.
* Passengers - Count them. Or give them a number and get then to shout out in turn.
* Handicaps - Count the wheelchairs.


Hope this helps. How accurate do you want them?

Sitting the car on a weighbridge with individual weight measurements for each corner would help with a lot of things. This way you could calculate the overall mass of the vehicle and the CoG location in the X-Y plane, as well as the fuel tank mass and CoG if you use Tristan's drain/refill method. You could also get a friend of known mass to sit in each seat in turn to calculate the CoG of passengers.

Finding the height of the CoG is going to be tricky.

A rolling-road dyno will be required for engine testing, and you can also use it for testing the brake strengths (and, therefore, brake balance). Rolling roads are fine if used correctly...we've got one here and I've not heard of any problems with it.

Finding the rotating mass of a wheel isn't as simple as weighing it as this doesn't take into account the radial weight distribution. What you'd need to do is to apply a known torque to a wheel and measure how fast it speeds up or slows down. Then use Torque = I * angular acceleration to work out I, the moment of inertia.

In aerodynamics, the value 'CdA' can often be taken to be a single parameter, for the simple reason that 'Cd' can only be defined once you've defined an 'A'. Cd is usually measured using a coast-down test. This includes rolling resistance, obviously, but there are equations which can be used to seperate out the effects.

If the wings are infinitely thin then you can measure the wing angles by placing a ruler between the leading and trailing edges and measuring the angle. If the wing has thickness then you will have to use a ruler with a cut-out to match the profile of the wing at the leading and trailing edges so that the edge of the ruler lies along the chord line.

Advanced weighbridge stuff = okay

Rolling road - as long as you only want wheel torque/power and you don't use that horrid "run down" test to basically make up transmission losses then okay. But most people want flywheel torque/power, which is impossible to accurately know on a rolling road.

Oh, he wants I. In that case, suspending the wheel assembly on a pivoting platform (tri-wire suspension I think it's called), and simply timing it's natural rotations can lead you to quite accurate inertial figures. Been a few years since I did this, and that was with flywheels, but the principal is the same.

Coast down already mentioned. I have a Bosch book here Ben if you want to equation/process copied out of it.

Last bit covered. But if you simply after a reference angle, then ruler across both edges is fine in 99% of cases. It's not like you'd do aero calcs on the wing profile because vehicle wings always operate in a relatively large degree of turbulence.

Ah thanks guys, that's a lot of text. I'll have a read through and merge it with what I wrote on my lunch break (you covered some points I already did but I'm sure my explanations can be improved).

Regarding spring and anti-roll bar stiffnesses, I've already made a little calculator in Excel that will find it's way into the VHPA tools menu come the next release. For the rotating mass, I do mean just the mass. I already calculate the moment of inertia from the mass and the physical dimensions (although the algorithm could probably be refined slightly). I need the figure seperate from unsprung mass though since all the unsprung does not rotate. There is no consideration of brake discs in the process though, as I'm not modelling the brakes themselves at all (I'm waiting for LFS to do that before I follow suit).

Stewart - Cd is independant of A, I always thought (or at least until the object gets very small, when compared to air). Hence a perfect sphere has a Cd of 0.1 regardless of size. y/n?

Quote from Bob Smith :

Stewart - Cd is independant of A, I always thought (or at least until the object gets very small, when compared to air). Hence a perfect sphere has a Cd of 0.1 regardless of size. y/n?

Cd is a empirically-calculated constant which relates an object's properties to its drag. 'A' is simply a relevant reference area. It needn't be a frontal area, just an area. The real value of calculating Cd comes when you want to compare objects which are similar in shape but different sizes...in that case, the reference area you choose must be consistent in all cases, but it doesn't really matter which area you use.

For example, there isn't really any need to know the exact frontal area of each car. A simple width x height calculation would give you a good enough reference area to calculate a Cd in order to compare cars. The only requirement is that you are consistent in your calculation of 'area'. I've probably needlessly complicated things...sorry about that!
Assuming you choose a consistent definition of area then, yes, Cd is constant with area.

BTW...A sphere only has Cd = 0.1 at high Reynolds Numbers, once the boundary layer becomes turbulent. Cd is constant with area but it is not constant with flow velocity.

Wow - I may not know enough (or anything) to help you out but I did want to say I am staying tuned and am interested in reading/learning about this side of things!

OK I've uploaded what I've done so far: http://www.vehicle-analyser.com/gdata.html

Tristan, about what you were saying regarding brakes. Are there any equations to calculate brake torque from effective rotor diameters, friction coefficients, piston diameters, etc?

Stewart - ok, aerodynamics is more complicated than I thought. I've never really read up that much on it. I'll have to do some reading as I don't even know what the Reynolds number is. "Cd is constant with area but it is not constant with flow velocity." is interesting though, I think that is why stardard aerodynamic equations are only accurate up to a couple of hundred mph? I'm using frontal area as the side area of a car is very different and I'm only concerned with cars going in a straight line (in this program anyway).

At work I've got an excel thingy for calculating brakes (it was originally made for doing the calcs on Lancia Montecarlos, but I think I've got Reynard data [guesses] in it at the moment.

Give me a PM tomoz to remind me, as I'm crap at remembering stuff like this, and I'll email it over.

Ta very much. I'll PM you at lunch.

Quote from Bob Smith :

Stewart - ok, aerodynamics is more complicated than I thought. I've never really read up that much on it. I'll have to do some reading as I don't even know what the Reynolds number is. "Cd is constant with area but it is not constant with flow velocity." is interesting though, I think that is why stardard aerodynamic equations are only accurate up to a couple of hundred mph?

Hehe, it's OK, it took me a 4 year degree to learn what I know now

The 'standard aerodynamic equations' you mention probably assume that the fluid is incompressible (and isentropic and adiabatic, etc...). It's much simpler if you don't have to account for the change in density of a fluid as the pressure changes. A standard rule of thumb is that a density error of 5% is usually acceptable...that takes you to about Mach 0.3 (approx. 200 mph at sea level).

Reynolds Number is a non-dimensional number based on flow velocity, fluid viscosity and some reference length on the object. It can be thought of as the ratio of inertial forces to viscous forces. At low Reynolds Numbers, the viscous forces (i.e. skin friction drag) dominate the drag forces, while at high Reynolds Numbers the inertial forces dominate (i.e. pressure drag - the 'wake' behind an object).
Drag coefficients change because of changes in boundary layer behaviour at different flow velocities.

Quote :

I'm using frontal area as the side area of a car is very different and I'm only concerned with cars going in a straight line (in this program anyway).

The 'width x height' calculation should give an approximation to frontal area, not side area!

Edit - I should probably point out that these effects (probably) aren't relevant to VHPA as it stands at the moment...the usual drag equation is fine unless you want super-accurate figures at very low speeds or above 200mph.

Quote from StewartFisher :

The 'width x height' calculation should give an approximation to frontal area, not side area!

Yes, correct, since it is the frontal area that is required.

Regarding more complex aerodynamics, it would be nice for simulating top fuel dragsters and other high speed vehicles but I suspect I would need more data on the aerodynamic shape of the vehicle, which would put it out of the scope of the program. If there any general rules of thumb though, I'm interested. I have noted that it's typically high speed acceleration that is off (specifically, too high) when comparing the outputs to telemetry of the same vehicle.

Quote from Bob Smith :

Regarding more complex aerodynamics, it would be nice for simulating top fuel dragsters and other high speed vehicles but I suspect I would need more data on the aerodynamic shape of the vehicle, which would put it out of the scope of the program. If there any general rules of thumb though, I'm interested. I have noted that it's typically high speed acceleration that is off (specifically, too high) when comparing the outputs to telemetry of the same vehicle.

The only thing I can think of is called the 'Prandtl-Glauert Similarity Rule'. The analysis is, technically, limited to two-dimensional, inviscid, irrotational flows with thin aerofoils at low angles of attack and at Mach numbers well below 1. However, I'm informed that it actually works quite well for real flows.

The idea is that the compressible drag and lift of an aerofoil can be related to the incompressible drag and lift by the following equation:

Cd, compressible = (1 / SQRT(1-M^2)) * Cd, incompressible

where M is the 'free-stream' Mach number (the Mach number of the vehicle, in this case).

You can calculate the Mach number, in air, using the vehicle velocity (v, m/s) and the ambient temperature (T, °C):

M = v / SQRT(401.8 * (T + 273.15))

As I said, a lot of assumptions go into this approximation and there's no guarantee that it'll work for a vehicle simulation. However, it will provide an increase in Cd at higher Mach numbers...this might be enough to reduce the high-speed acceleration. As a guide, the Cd will increase by 5% at M = 0.3 (230 mph) and 10% at M = 0.4 (307 mph).
Or it could just give you complete nonsense

Although I doubt it will solve the extra high speed acceleration I seem to have (too subtle an effect at the speeds I'm trypically dealing with), it is interesting none-the-less. I will try adding this and see how it works in practise whenever I next start coding VHPA again.

A question about coil springs:

Using these definitions for the terms, if I have the following data on a spring:

Free length: 145mm
Pitch: 38.8mm
Turns / coils: 5.25

If I'm reading things right, then the free length and pitch give me only 3.75 turns, or 1.5 turns less than the spring actually has. What does this mean? Is 3.75 the number of active turns (so 1.5 turns are ineffective, I thought this was usually 1 turn?)???

Could it be that the last bit of the spring has closed and ground threads, and therefore isn't the same pitch as the 'main body' of the spring? Can ye poste a picturee?

That would be the case, yes. I can do better than a photo, I'll send something your way tomorrow, Tristan.

What should I use for the active number of coils then?

Quote from StewartFisher :

You can calculate the Mach number, in air, using the vehicle velocity (v, m/s) and the ambient temperature (T, °C):

M = v / SQRT(401.8 * (T + 273.15))

I take it that is a specific approximation of this equation: http://en.wikipedia.org/wiki/M ... r#Calculating_Mach_Number ?

Anyway, I've added this (your) equation into VHPA now, it takes only a couple of mph of the top speed of a very fast car, but reduced the top speed of a top fuel dragster by 15 to 20mph. Sounds quite reasonable to me.

Quote from Bob Smith :

That would be the case, yes. I can do better than a photo, I'll send something your way tomorrow, Tristan.

What should I use for the active number of coils then?

Well, the end coils are pretty damn negligable - any part of the spring that touches the spring perch won't contribute to the spring (which is about 0.5 turns at each end), whilst the lower pitch at the ends also increases the local rate (and can therefore be ignored most of the time). So number of actual coils - ~1.3 - 1.6 = number of active coils.

Is this to calculate spring rate from number of coils and wire diameter?

The other way is to measure by experiement, which isn't as hard as it sounds. Simply measure the free length of a spring, then take a known mass and suspend it through the coil (having a hole in your bench for the wire/string is useful ), so that M.g acts on the spring. Measure loaded height and do the simple maths. Double check (or use as a double check) with the theory method. Variations in metal alloy and heat treatment, as well as age, inaccurate measurements etc will add variance to the results.

Ah, I see what you mean now.

Pitch x Number of Turns > Free Length
whilst
Free Length / Number of Turns < Pitch
whilst
Free Length / Pitch < Number of Turns

It doesn't add up, does it...

http://www.eng-tips.com/threadminder.cfm?pid=800 is probably a good place to post a question, as there are some very technical people on there (but also some muppets, in true forum fashion).

However, judging by the picture that you send me (on page 104) it looks as though there is about 3.5 ACTIVE coils, and hence I'd use that number to start with, as the numbers won't be a million miles away from reality (which has component flex and tyre rates anyway, so you'll never be 'perfect').

Can you find a source for the actual rate of these springs anywhere, and compare that to your calcs based on the above data?

Quote from tristancliffe :

Can you find a source for the actual rate of these springs anywhere, and compare that to your calcs based on the above data?

If I could do that, I wouldn't have bothered trying to work it out. Getting the active coils from pitch * free length seems so to work well enough, as we're interested in wheel rates and are having to guess the motion ratio anyway, so absolute stiffnesses numbers don't need to be exact, just the ratio between the two. For the car in question, with a motion ratio of 1.4:1, it works out about 3Hz front 2.9Hz rear, which seems right in the ballpark, and almost exactly what we ended up with after Bruno Senna had tuned the suspension by feel. So from that point, all seemed to match up and we're happy.

I'm still very surprised by the damping data though, that's a lot harder than I expected (having compared those values to critical, using estimated wheelrates as explained above anyway).

When you say we, who is we? And whan you say Bruno Senna tuned the suspension, you make it sound as though he did it for you?

But if you were working for a race team that Senna does driving for, then I don't think I'm the man to help - you'll have vastly cleverer people in the room next door.

I'd use the number of visibly active coils and go from there. If you know the weight distribution, the motion ratio, the free length of springs and loaded length of springs, then can you deduce the spring rate from the amount of sag in the suspension when you put the car on the wheels (or from the data you have)?

I don't know what you know/remember of my personal life, but this thread (title is enough) combined with this preview should make things clear enough.

Ah I see. You're churning out another flawed car game

I remember you saying now that the other people on the team didn't really care about accuracy/details/realims (or something along those lines). Have you managed to convince them otherwise?

I've not worked on Ferrari Challenge myself, I'm still doing training writing my own (semi-)sim, and learning loads from it. I just happened to be asked to double check (estimate) the values independantly so a comparision could be made. I'll be getting more involved in real work in the next couple of games we've started working on, once it's been decided that I'm having too much fun writing my own dynamics engine.

I think it's safe to say all the games this company has made so far are definitely arcade, Ferrari Challenge being the most sim like so far. As far as what seperates our games from being a hardcore sim (purely from a physics/dynamics points of view), there are only 2 differences:
a) driver aids (TC, ABS & stability control, plus steering aids when using a gamepad)
b) tyre modelling (made easy to drive and simplified)

All the rest is there; engine, clutch, gearbox, differentials, brakes, suspension (with 3D movement, non-linear dampers), steering geometry, aerodynamics. It's technically a full simulation, just made easy to drive, and in places certainly not as detailed as it could be. Not as dynamic as LFS for sure (and the FFB seems crap, though I've not had a go on the company G25 yet, hearing the noises it makes when driving is enough to know not to bother) but then it doesn't intend to be.

Ultimately it's more important that cars feel right than are mathematically correct, for a console racer. I doubt we're quite on par in terms of physics with Forza but the game is convincing enough, and certainly meets its own objectives.