32 replies to this topic

[Neverhouse]

Hi there,

I have a few questions about telemetry data analysis. I frequently play a racing simulator and already had a look at some car setup guides, but there are a few things I just can't answer...

1) Ride height: It's pretty obvious that this shouldn't drop to zero, but how do I detect "too much RideHeight"? If it goes up to 120 or 130mm, is that ok or should I stiffen the springs?

2) Packers: When I notice that my ride height is too low.. when should I increase ride height? When should I increase packers instead?

3) Damper velocity: Is there something like an "optimum minimum and maximum value"? My last logfile shows me that the damper velocity is always between -70 and +70 (mm/s)... what does that tell me? Are the damper settings ok, or should they be lower / higher?

4) How do I know if I should stiffen/soften the springs or the dampers?

Hope someone can answer me that questions, before they're keeping me awake at night.

Cheers,
Neverhouse

[gruntguru]

Hi there,

I have a few questions about telemetry data analysis. I frequently play a racing simulator and already had a look at some car setup guides, but there are a few things I just can't answer...

1) Ride height: It's pretty obvious that this shouldn't drop to zero, but how do I detect "too much RideHeight"? If it goes up to 120 or 130mm, is that ok or should I stiffen the springs?

2) Packers: When I notice that my ride height is too low.. when should I increase ride height? When should I increase packers instead?

3) Damper velocity: Is there something like an "optimum minimum and maximum value"? My last logfile shows me that the damper velocity is always between -70 and +70 (mm/s)... what does that tell me? Are the damper settings ok, or should they be lower / higher?

4) How do I know if I should stiffen/soften the springs or the dampers?

Hope someone can answer me that questions, before they're keeping me awake at night.

Cheers,
Neverhouse

Somebody may reply with some rule-of-thumb answers for you, but why not try proven track testing techniques? You will learn a lot more that way.
1.

[gruntguru]

I have a few questions about telemetry data analysis. I frequently play a racing simulator and already had a look at some car setup guides, but there are a few things I just can't answer...

Hope someone can answer me that questions, before they're keeping me awake at night.

Somebody may reply with some rule-of-thumb answers for you, but why not try proven track testing techniques? You will learn a lot more that way.
1. Find a set-up that produces good control and consistent lap-times.
2. Make one small change (perhaps stiffen springs all-round by 10%) and observe the effect on lap-times (as well as all the telemetry channels but don't be tempted to chase other stuff you see there).
3. Optimise that variable (ie if stiffer springs improved the lap-times, stiffen them a bit more etc)
4. After optimising spring rate, you will need to optimise damping rate.

Edited by gruntguru, 08 November 2009 - 03:46.

[Greg Locock]

70 mm/s is very slow speed for a shock, perhaps your springs are too stiff. OTOH 120mm of suspension travel sounds like a lot. Is this an aero car?

4) Yes that is a very good question.

[Neverhouse]

It's a Porsche 911 GT3 Cup.

So this is more a "trial-and-error" process? Ok, the driver has to find an adequate setup in training, but aren't there some general "Do's and Dont's"?

[Greg Locock]

It's a Porsche 911 GT3 Cup.

So this is more a "trial-and-error" process? Ok, the driver has to find an adequate setup in training, but aren't there some general "Do's and Dont's"?

If you want to approach it as trial and error, yes it is. If you want a structured approach use that instead. Milliken has quite nice summary of interactions.

[DaveW]

It's a Porsche 911 GT3 Cup.

So it's not a "serious" aero vehicle. Try not to use bump rubbers. A 911 is likely to carry around 40% of the static weight on the front wheels, so the rear spring rates should be higher than the fronts (at the wheels). You might like to start with around 120 N/mm front & 180 N/mm rear springs (at the wheels). Lower rates (with suitable damper adjustments) should improve "grip", but will cause the platform to move around more during manoeuvres (&, if the simulation is realistic, geometry changes might push the vehicle to over-steer at the limit on lower rate rear springs). Adjust lateral imbalance with bar changes (increasing front &/or reducing rear moves balance towards under-steer, & vv). If your model has the ability to simulate compound springs, don't.

If the simulation is running on a "billiard" table (no vertical features), then I suspect your stated peak damper velocities may be a little high (suggesting that your model may be under-damped). Otherwise they are too low. In reality, expect maximum damper velocities of around 400 mm/sec on European circuits. US circuits are likely to generate higher damper velocities - perhaps as high as 1 m/sec.

Lower static ride heights should yield higher cornering speeds, hence better lap times, provided you don't run out of wheel travel.

Not sure how detailed your model dampers are. Start with symmetrical dampers matched to the springs with damping coefficients of, say, around 5 N/mm/sec front & 8 N/mm/sec rear (at the wheels). Compression-biased damping should improve grip directly & should permit a lowered static ride height. Rebound-biased damping should lower the average c.g. height when running (useful if a minimum static ride height rule applies). Digressive compression damping at the front axle should improve (reduce) the steering time constant. More linear damping at the rear should improve "traction". Compression blow-off should improve "kerbing" (assuming kerbs exist in your simulation).

Multiply the rates suggested above by 5.71 to convert to lb/in (in case)....

Have fun, & experiment.

[Neverhouse]

Milliken has quite nice summary of interactions.

I assume you mean the book "Race Car Vehicle Dynamics"?

DaveW, thanks for you detailed explanation.

In reality, expect maximum damper velocities of around 400 mm/sec on European circuits. US circuits are likely to generate higher damper velocities - perhaps as high as 1 m/sec.

Why is that so? Are US circuits more "bumpier" that the ones in Europe?

Mainly I'm using the 2 popular simulators "rFactor" (with PCC 07 ) and "GTR2". I've never driven a real race-car, so I can't tell how realistic they are.
I live in a pretty small town, but we have a racing team here that ranked #4 in the german Porsche Carrera Cup 2009. But I bet it wouldn't make much sense to walk in there and ask them if I could test-drive one of their cars.

[DaveW]

Why is that so? Are US circuits more "bumpier" that the ones in Europe?

By & large, yes.

Oh. & I'm sure your last paragraph is correct....

Edited by DaveW, 09 November 2009 - 23:00.

[ferruccio]

It's a Porsche 911 GT3 Cup.

So this is more a "trial-and-error" process? Ok, the driver has to find an adequate setup in training, but aren't there some general "Do's and Dont's"?

Generally a professional race team, in real life, will want to eliminate as much 'trial and error' approach to set up. In other words they want to avoid 'guesswork'. It's highly inefficient to rely on guesswork and usually results in being at the back of the grid or inconsistent performance throughout the season. Worse is when the engineers don't understand why the set up is quick. They try to understand it after the fact. They don't understand the car.

A good team spends a great deal of time preparing. They maximize their track time in testing sessions. They actually run 'sweeps'. Like changing one parameter through it's range while everything else remain constant to understand how the car generally responds and behaves. They have a well structured testing plan and are methodical in their approach. With all the quality data collected they can minimize trial and error approaches on race day. Trial and error are still used sometimes when things don't go to plan but its best to avoid it.

You can apply this approach to a racing game as well i'm sure. Some teams actually do this on special simulators for race set up and it really helps on race day.

[gruntguru]

Generally a professional race team, in real life, will want to eliminate as much 'trial and error' approach to set up. In other words they want to avoid 'guesswork'. It's highly inefficient to rely on guesswork and usually results in being at the back of the grid or inconsistent performance throughout the season. Worse is when the engineers don't understand why the set up is quick. They try to understand it after the fact. They don't understand the car.

A good team spends a great deal of time preparing. They maximize their track time in testing sessions. They actually run 'sweeps'. Like changing one parameter through it's range while everything else remain constant to understand how the car generally responds and behaves. They have a well structured testing plan and are methodical in their approach. With all the quality data collected they can minimize trial and error approaches on race day. Trial and error are still used sometimes when things don't go to plan but its best to avoid it.

You can apply this approach to a racing game as well i'm sure. Some teams actually do this on special simulators for race set up and it really helps on race day.

I think a "Sweep" is the kind of testing that many people think to be trial-and-error.

[DaveW]

They actually run 'sweeps'. Like changing one parameter through it's range while everything else remain constant to understand how the car generally responds and behaves.

aka as a "matrix search". Apologies, but I can't think that a matrix search is the most efficient way of arriving at an optimal suspension set-up. If, for example, you try just three settings for each of 2 (pairs of) dampers & springs, that is a total of 81 runs (I think), & the optimal set-up will be found only if the selections include the optimal value for each parameter. That would be an expensive exercise to carry out at a track, even in the unlikely event that everything else (temperatures, tyre characteristics, etc.) remained constant for the duration of the exercise. There are more efficient methods, I think.

[Greg Locock]

I suspect most people these days do a real life 2 level taguchi screening experiment for first order effects only, record the results, use them to calibrate their chosen analytical tool and then run three level third order interaction analytical models to explore the entire solution space.

Leastways that's what I do for limit handling work. 4 cpus running flat chat for weeks at a time.

Of course defining your appropriate factor levels is half the fun.

[DaveW]

I suspect most people these days do a real life 2 level taguchi screening experiment for first order effects only, record the results, use them to calibrate their chosen analytical tool and then run three level third order interaction analytical models to explore the entire solution space.

Something like that, Greg... Arguably, the efficiency of a "blind" DOE phase can be further improved by incorporating a priori knowledge. A pernicious feature of any blind statistical approach is that, whilst it may determine what to do, it is unlikely to reveal why - so a deficiency affecting performance may remain undiscovered. Perhaps that is not a major issue for road vehicles, but it can be so in a racing environment.

Edited by DaveW, 11 November 2009 - 07:23.

[MattPete]

A good team spends a great deal of time preparing. They maximize their track time in testing sessions. They actually run 'sweeps'. Like changing one parameter through it's range while everything else remain constant to understand how the car generally responds and behaves.

That method is going to miss all of the interactions between the variables, which is where the cool stuff lies.

[Fat Boy]

That method is going to miss all of the interactions between the variables, which is where the cool stuff lies.

This method also educates your driver, which is where all the speed lies.

[ferruccio]

Well the trick is to plan what sweeps to do. You can't do everything. Good teams prepare well beforehand on which tests to do to best maximize their track time during test sessions. The purpose of the test days is just to collect as much useful data as possible. They of course also test for scenarios, qualifying, race, wet if possible etc etc. just to establish baseline behaviour in those conditions

[DaveW]

This method also educates your driver, which is where all the speed lies.

Without question, real (as opposed to simulator) cockpit time is essential even for experienced drivers, as demonstrated by Ferrari F1 (& others) this year. That aside, however (&, perhaps, returning to topic), a cockpit is not the best place to assess suspension set-up, not exclusively, anyway. The fact is a driver has few natural cues about the effect of suspension set-up on corner exit speeds, & it is those that have most effect on lap time. Recorded measurements (telemetry, if you will), supported by the results of hardware in the loop tests ("rig" tests), are useful, even essential, tools to achieve a good suspension set-up compromise.

[murpia]

How about analysing the effect of changes to individual components (spring rates, bar rates, wing angles, pushrod lengths, inboard pickup point positions etc.) into their effect on vehicle dynamics parameters such as aero balance, roll distribution, heave stiffness, total downforce etc. Do this ahead of the track session using simulation tools.

Then at the track, instead of sweeping say front spring rate, sweep +/- 1% roll distribution. That educates the driver as to the effect of the change, and the engineer can decide in future to achieve the change by either springs, bars or geometry so as to take into account secondary effects.

Regards, Ian

[Fat Boy]

Getting OT here.

In terms of ultimate suspension setup, I agree, the driver is not the best analysis tool. Hell, the racetrack isn't the best place to do it, either. There are simply too many variables with either to evaluate the setup alone. So you end up doing what you can on a shaker or better yet, on a sim. I think it would be hard to argue against this.

However, this thread excepted, we don't race on a computer track and we don't do it with computer drivers. When you have 8 minutes left in qualifying, you're 0.2 sec off pole and sitting in 5th place, you don't need to run a DOE. You need to have a driver who knows what he needs to go faster. It's often pretty small and they can often call it to you as they roll toward you in pit lane. "I just need a little more platform across the front of the car, like when we did XXX" or "We need to get the rear to plant just a tick better when I go to throttle, like when we did YYY". This means that you have to do sweep testing if you want the driver to be able to help you in these situations. So often engineers get stuck in computer world and neglect that there is a human being in the car. It makes a difference.

The problem with sweeping just roll distribution is that springs and bars act differently. If you were to sweep roll couple 1% using springs, you'd change not only the roll characteristics, but also pitch characteristics. If you were going to do it strictly on the bars (a lot of the time the drivers have the ability to change things more than this on the ARB adjuster), then this acts in a different manner than springs or geometry. More often than not, changes that we make in a static sense actually make a bigger influence in a second order dynamic sense. It will change the timing or progression of an issue and that will be a much bigger deal than the first order influence that we initially were trying to effect.

Regardless of what your grip number or Sim says, the ultimate truth is in the stopwatch.

[DaveW]

Regardless of what your grip number or Sim says, the ultimate truth is in the stopwatch.

Absolutely, & I fully understand that a race engineer's task is (at least) as much psychological as it is technical.

However, it is also true that drivers can push set-ups into a corner of the envelope from which it is difficult to recover, either from lack of experience (the first FFord I saw had rebound travel preloaded out - along with the rest of the pit lane, I was told), or to try & work around a deficiency (I once saw a vehicle with rear springs 3 times stronger than they should have been - caused by rear compliance steer).

I suppose I was trying to make the point that a complaining driver is usually right but his interpretation of the cause is not necessarily so, & that set-ups he is happier with might not improve lap times. I tried to explain why that can be so.

BTW the rear compliance steer example was, by the time the driver had finished, 0.8 secs slower round Magny Cours compared with a more conventional set-up & scrutineers pulled apart the engine of the FFord example several times after I saw it because they were convinced it had been "fixed".

[Greg Locock]

The result of the 2000 run experiment can often be boiled down into a spreadsheet, or a table, or even a rule of thumb. For instance, off topic, lots of modelling of solar car racing gives you the simple rule of thumb that you should slow down by half the speed of a headwind. I can tell the driver that, I can tell the whole team that, they don't need to read it off the computer. Similarly, if there are clouds overhead we should speed up, tho I prefer to do so by a slightly complex rule. During the race we don't run any simulations, I have one spreadhseet, and can run the race on my calculator if need be. In fact one sheet of graph paper and some mental arithmetic will do.

So far as the way that different things behave in different ways, here's a study I did on the difference between sta bar balance, roll steer, and tire cornering stiffness balance on high speed stability. Sta bars behave differently to the other two, not very much to my surprise. http://greglocock.we...cledynamics.htm

[Fat Boy]

However, it is also true that drivers can push set-ups into a corner of the envelope from which it is difficult to recover, either from lack of experience (the first FFord I saw had rebound travel preloaded out - along with the rest of the pit lane, I was told), or to try & work around a deficiency (I once saw a vehicle with rear springs 3 times stronger than they should have been - caused by rear compliance steer).

I suppose I was trying to make the point that a complaining driver is usually right but his interpretation of the cause is not necessarily so, & that set-ups he is happier with might not improve lap times. I tried to explain why that can be so.

I completely agree that a driver, especially an uneducated one, can paint you into a corner if you follow him blindly. That's where the rig testing and homework really comes into play. You have to know what a 'reasonable' setup is before you get to the racetrack. There is two over-riding approaches I've seen in racecar engineering. The first is 'Here is the perfect car you must drive it' and the second is 'Help me make a car that works for you'. The more pure science oriented an engineer is, the more likely he is to choose the first. While there are times that everyone has to say, "Trust me this is right" to the driver, it almost always goes wrong to take that approach with the entire car.

--------------------------------------------------------

And what the hell is it with FFord guys winding in turn after turn of preload? I've never understood that either, but it's almost a universal trait.

[Ross Stonefeld]

However, this thread excepted, we don't race on a computer track and we don't do it with computer drivers.

Regardless of what your grip number or Sim says, the ultimate truth is in the stopwatch.

This would be considered blasphemy in a Grand Prix paddock. Which is why I adore you

[DaveW]

And what the hell is it with FFord guys winding in turn after turn of preload? I've never understood that either, but it's almost a universal trait.

I suspect it is because, for most drivers, FFord is their first encounter with a suspended race car & they pay for seat time, so call the shots.

I know a successful F3 team owner who takes the time at a track test to educate his drivers on how to "put together" a corner. The fact that he feels the need to do that is interesting, I think, as is the fact that a surprisingly high proportion of his drivers "graduate" to high profile racing careers.

[murpia]

The problem with sweeping just roll distribution is that springs and bars act differently. If you were to sweep roll couple 1% using springs, you'd change not only the roll characteristics, but also pitch characteristics.

That's exactly my point. Get the driver to understand what e.g. +/- 1% of roll distribution feels like, but let the engineer decide whether to achieve it by springs, bars or geometry etc.

That way the engineer can account for the secondary effects the way that best suits the car & track (mainly aero effects and vertical stiffness effects).

If the driver insists on a spring change, but the engineer knows it's bad for aero, he has a much better chance of convincing the educated driver that a bar or geometry change will be a better compromise.

Regards, Ian

[Fat Boy]

I suspect it is because, for most drivers, FFord is their first encounter with a suspended race car & they pay for seat time, so call the shots.

I know a successful F3 team owner who takes the time at a track test to educate his drivers on how to "put together" a corner. The fact that he feels the need to do that is interesting, I think, as is the fact that a surprisingly high proportion of his drivers "graduate" to high profile racing careers.

I always figured it comes down to the FFord 'tuners'. Most are mechanics with a layman's understanding of what is happening. They equate stiff with fast and see increasing preload as 'stiffening the spring' (I don't know how many times this has been said to me). It's an easy change to make, it's something that their driver can probably feel. If it's done later in a session (as the driver is learning the track) the lap time usually falls after the change is done. Kind of a comedy of errors.

It doesn't surprise me at all that educating an F3 driver on something that seems a simple topic produces good results. We often expect way, way too much from junior formula drivers. We wouldn't expect a mechanic that has only worked on karts to rebuild the entire car, but we expect a driver that is in his first year out of karts to know what a racecar needs. This seems more than just a little contradictory to me. We've been over this subject before, but Tiger Woods and Roger Federer have coaches who help them with form and technique, why is it such a stretch to think that a driver might need one?

Edited by Fat Boy, 12 November 2009 - 16:55.

[Fat Boy]

This would be considered blasphemy in a Grand Prix paddock. Which is why I adore you

Yup, you and PII.

[Fat Boy]

That's exactly my point. Get the driver to understand what e.g. +/- 1% of roll distribution feels like, but let the engineer decide whether to achieve it by springs, bars or geometry etc.

That way the engineer can account for the secondary effects the way that best suits the car & track (mainly aero effects and vertical stiffness effects).

If the driver insists on a spring change, but the engineer knows it's bad for aero, he has a much better chance of convincing the educated driver that a bar or geometry change will be a better compromise.

Regards, Ian

I think we're saying the same thing. I didn't understand what you meant, initially.

[Ross Stonefeld]

Yup, you and PII.

Where is that dude, has he gone deep undercover?

[maxay1]

I have to echo a comment I believe made by Fat Boy, with regard to the OP not returning to tell us of the results of his newly gained knowledge....Even given a sim, it would be interesting to know....

[Fat Boy]

Where is that dude, has he gone deep undercover?

I just naturally assumed he's setting up a sniper's nest on the top of the Rio Casino in Vegas to take out his arch nemesis (who is not a very good racecar driver).

[desmo]

I think his head exploded when he discovered he had to join the CCP to make his traitorous job exportation scheme fly