- Doug Arnao
GPL's physics model is a brilliant achievement. It's a complex, sophisticated mathematical model of real-world race car dynamics and tire characteristics. I tend to think of it as a distillation of the textbook of race car dynamics, Milliken and Milliken's Race Car Vehicle Dynamics (see References), into a very efficient computer program.
However, the physics model in GPL isn't complete. As far as I can tell, it doesn't model bump steer, Ackermann effect, and several other relatively small suspension characteristics.
GPL also seems to use simplistic plots of tire slip ratio and slip angle. The sensation of having a torque converter between the engine and the wheels when you ramp up the power suggests that the slip ratio "curve" is actually a straight line from zero to peak. I suspect the same is true of the slip angle "curve".
Also, GPL's tracks fail to model the small, high-frequency bumps that exist in real life. This allowed the use of unrealistically low setups in GPL 1.0 because such bumps, which would have caused the cars to weave and dart uncontrollably when running way down on very hard bump rubbers in real life, simply weren't there in GPL.
Engineering improved bump maps was beyond the scope of a patch, so this was addressed in the GPL 1.1 and 1.2 patches by restricting the ride height to 2.5 inches. This may be a bit of a kludge, but it's effective.
Arguably the most significant omission in GPL's vehicle behavior implementation is the lack of an audible signal when the suspension bottoms on the bump rubbers. In real life, when the suspension reaches the limit of its travel, and contacts the bump rubbers, there is a loud thump which can be heard and felt throughout the car. But GPL is silent when this happens.
Without audible warning of suspension bottoming, everyone (including the developers of the default setups shipped with GPL) developed setups that were so low and soft that the suspension bottoms on the bump rubbers at numerous points around almost every circuit. Influenced by modern-day setup practices, where ground effect aerodynamics demand that the chassis run low to the ground, everyone figured, "lower is better".
But we were wrong. Without ground effects, the equation is drastically different. Lower is not necessarily better. There are many factors which influence the overall grip available, and low ride height (which in GPL has only the benefit of reduced weight transfer) is only one of them.
The consequence of bottoming the suspension on the bump rubbers is sudden weight transfer away from the inside tire and to the outside tire. When this happens on the front, the result is sudden, massive understeer. When it happens at the rear, the result is snap oversteer - an unstable condition which renders the car uncontrollable for all but the most talented of drivers.
I am convinced that much of GPL's undeserved reputation for being "too hard" is rooted in the the absence of this audible warning of bottoming of the suspension, and the unstable setups which resulted.
In 1967, cars had no appreciable aerodynamic downforce. Maximum cornering speed comes down entirely to mechanical grip. Mechanical grip revolves around optimizing the shape of the tire contact patch, balancing the relative load on each of the tires, and keeping the tires in contact with the track surface as much of the time as possible.
Optimizing the shape of the tire contact patch and the loads within it falls mostly to tire pressures and camber, discussed elsewhere. Keeping the tires in contact with the track surface as much as possible comes down to dampers, also discussed elsewhere, and spring rates (also known as wheel rates).
Several sources from 1967 (see Ricardo Nunnini's GPL Foolishness) make it clear that in 1967, the cars did not run on the bump rubbers. In later years, as aerodynamic developments forced the designers to find ways of coping with very high downloads at high speeds, it became common practice to run very long, relatively soft bump rubbers, and the car was in contact with these for much of any given lap.
But in 1967, the cars ran high enough so that they did not contact the bump rubbers, except in special cases like the Nurburgring, or under momentary high vertical loads at points such as the entry to 5A at Mosport or Eau Rouge at Spa. Setup technique was to select the proper springs for the circuit, lower the car until it was just bottoming on the bump rubbers, and then raise it up so that it didn't bottom, except in the special-case areas where keeping the car off the bump rbbers compromised cornering too much everywhere else.
There's a very good reason why the cars weren't designed to run on the bump rubbers. The softer the suspension (i.e, the lower the wheel rate), the better the wheel and tire can follow the bumps, and therefore the more grip it makes. When the suspension contacts the bump rubber, the wheel rate goes up dramatically. This has two detrimental effects: it reduces the available grip, and, unless the suspension is bottoming at each end at precisely the same instant and the same amount (something which rarely happens), it upsets the balance of the car .
The upset in balance is most noticeable under power at corner exit. When the suspension touches the bump rubber as the car squats under power, the rear end will suddenly go "greasy", and will start to slide out. Snap oversteer. This is bad: it's giving away grip, and it makes the car unstable and harder to drive.
The balance can also be upset by contacting the rear bump stops in mid-corner and even under braking, especially if the rear wheel rates are too soft relative to the front (more on this below).
Running on the bump rubbers also exacerbates the effects of the track's "crown", which in GPL is more like a peak than a crown. Real life roads and race courses are often crowned; that is, they are convex so that water runs off the center down to the sides, where it can drain away. GPL apparently can't have a track surface that's curved laterally, so it models this by having a peak at the center of the track (or in some cases two peaks, one near each edge) with the track surface sloping away from this peak.
The result is that twice each corner (in the majority of corners in GPL's tracks) the car must transition this peak. A car running very close to the bump rubbers will contact the bump rubbers on two corners - typically opposite corners - as it transitions this crown. This has a drastically unsettling effect on the car.
GPL has another subtle flaw. It permits front spring rates of 100 lb./in., and rear spring rates of only 120 lb./in. Because of this, almost everyone blindly went ahead and created setups with a proportion of front to rear spring rates roughly in proportion to 100/120. This is flat out wrong. It took me almost two years to realize the fallacy of this, and only then through reading Ricardo Nunnini's GPL Foolishness and reviewing his spring rate table. (A more sophisticated version of this table has essentially been built into GRE as the Fixed Wheel Rate Proportion feature).
GPL's cars all have roughly 60% of their weight on the rear wheels. This means that to have balanced spring rates - to properly hold up the car - if we used 120 lb./in. rear wheel rates, we'd need about 80 lb./in. on the front wheels. If we run anything stiffer than 80 on the front, in GPL we can't get the rear springs stiff enough to keep things in proportion.
The problem with running springs that are stiffer at the front than at the rear with relation to their relative load is that the front end then becomes a fulcrum for the center of gravity of the car when it's subjected to high vertical loads. In other words, when the car loads up on a banked corner or at the bottom of a dip, the stiffer front springs are going to make the chassis droop more at the rear - thus increasing the chances of the rear suspension bottoming. (See How Do I Know when It's Bottoming? for a detailed discussion about this crucial issue.)
After a lot of experimenting, I've concluded that it's essential to keep the front and rear spring rates in proportion to the load they carry. We can go a little bit one way or the other, but not much without screwing things up.
An important consideration is the type of circuit we're driving on. The relevant factors when considering spring rates and ride height are bumpiness and vertical load.
In terms of bumpiness, road and street circuits fall into three categories of roughness: almost absolutely smooth (Long Beach, Vancouver), moderately bumpy (Monza, Kyalami, Silverstone), and very bumpy (Zandvoort, Mid-Ohio).
In terms of vertical load, there are low-G circuits (Long Beach, Vancouver), moderate-G circuits (Monza, Silverstone), and high-G circuits (Watkins Glen, Kyalami, Mosport, Spa, the Nurburgring).
Why do I consider Monza and Silverstone to be moderate-G circuits? After all, they are nearly flat, aren't they? But they aren't. There are subtle upslopes and bankings that load up the suspension at important times.
At Monza, there is a significant upslope followed by a dropoff and then a crowned, banked turn at Curve Grande. This can easily bottom the suspension of a low, softly sprung car just after you've turned in and are trying to make the car take a set for the corner. There are also upslopes followed by dropoffs at the entrance to the First Lesmo and the exit of the Second Lesmo, and a gentle upslope and dropoff (and crown transition) at the entrance of Parabolica. Both of the Lesmo dropoffs can bottom the suspension and cause the car to snap sideways.
Silverstone has even more G's. There are dropoffs at the entrances of Copse, Becketts, Stowe, Club, and Woodcote, and steeply banked corners at Maggots, Chapel, Abbey, and Woodcote with a nasty bump at the exit of Abbey that further tends to load the suspension.
Kyalami is fairly smooth but has a high G corner at Jukskei and a very high G load at the entrance to the right-hand part of the esses. It also has a nasty bump at the entrance to the esses, a dropoff at the entrance to Club, a crest at the exit of Club, and a serious rise and crest with crown transition at the exit of Leeukop.
Spa has a very high G area at the lower part of Eau Rouge, but it also has a very high G area in the first part of Masta (it's slightly banked and very fast), followed by a crown transition. If the car bottoms here - and the GPL default setups and most other setups bottom the rear suspension here - the car becomes very unstable at a very high speed, a bad combination. Most setups also bottom in Stavelot, which is a long, very fast inside vertical curve which generates high G's, and in Malmady, at the transition from the downhill to the flat in the right-hand part. The penalty of bottoming in Stavelot is snap oversteer in the early phase of the corner, at very high speed - an extremely dangerous condition, and the penalty at Malmady is at best slow exit speed onto the long straight, and at worst a trip into the hedges.
The Glen has an obvious high G area at the exit of the Loop, but it has a more insidious small rise and crown transition right before the switch from left to right in the Esses. Get this wrong and you're slow down the straight or taking a trip into the guardrail. Almost every setup I've ever driven bottoms here, and as a result the car becomes unstable just where you need stability most.
It's very important to analyze the circuit carefully, particularly for its vertical loads. Unless the car is running so low that the car doesn't react over them, bumps are generally easy to identify. However, areas of high vertical loads can be very tricky to identify. A good clue is if you keep messing up a corner for reasons you don't understand. When the car suddenly gets greasy at the rear, or snaps into a spin for no reason, chances are you've hit a bump, a high G zone, or a bump in the middle of a high G zone.
Also, keep in mind that a dropoff - a negative bump, if you will - can be just as critical as a bump that rises out of the surface. When you go off the dropoff, the car gets light, allowing the tires to lose some grip. Then, after you go off the dropoff, and the car starts dropping, in a moment you're going to hit bottom. That will load up the suspension, possibly bottoming it just when the car is already starting to slide. The entries to Stowe and Woodcote at Silverstone are good examples of dropoffs that bottom the suspension and cause snap oversteer when braking and turning in with a car that's set up too low and/or too soft.
Also, I've found a good Force Feedback wheel to be very helpful in identifying high G areas because the forces at the wheel become heavier when the download goes up. My wheel also tracks bumps and crown transitions well; I can actually feel the wheel respond to them.
See the Circuits page for detailed discussions of problem areas at specific circuits. See How Do I Know when It's Bottoming? for help identifying when the car is having difficulties with such areas.
We need to find a compromise among ride height and spring rates which gives us as much grip as possible while keeping the car off the bump rubbers. Our options are high and soft, or low and stiff, or somewhere in between.
We'd like to make the suspension as soft as possible for more grip, especially on bumpy circuits. But when we do this, we've got to raise the ride height to keep the suspension off the bump rubbers. (We can also increase the damper stiffness in bump, which will help keep the car up over bumps, but there's a diminishing return to this; go too stiff on the dampers, and grip and traction will be seriously compromised and stability will degrade as the car dances all over the road.)
Going high, however, causes more weight transfer. We can see the effects of this in cornering, as the outside tire tempuratures go up, and in braking, as the car becomes more unstable because more weight is being transferred away from the rear, bringing the rears closer to lockup. We can compensate for this by increasing camber to even the tire temps (implying that we've gotten the tire contact patch flat against the road again) and by moving the brake balance more to the front. However, both of these measures reduce overall braking power - increasing braking distances - and increased camber also hurts traction.
We can also adjust tire temperatures by increasing anti-roll bar stiffness; making the car stiffer in roll will tend to reduce roll, keeping the tires flatter against the track and helping avoid the need to increase camber.
Indeed, this is just what was done in real life in the 60's. The cars tended to be relatively softly spring, but stiff in roll.
Just exactly what spring rates and ride height are optimal is a matter of taste and experimentation. I don't like a lot of roll; to me it makes the car feel like it's wallowing. (Maybe it's my karting background!) Obviously the best values at a flat, smooth circuit like Long Beach may be very different from the best values at a bumpy place like Zandvoort or a high G circuit like Spa. At a flat, smooth circuit you don't need as much suspension movement so you can run softer and/or lower than at a more vertically demanding circuit.
Incidentally, although GPL's setup menu permits bump rubbers no shorter than 1 inch, the physics module will accept .5 inch bump rubbers. I've begun to use .5 inch bump rubbers almost everywhere now. This allows me to run a bit lower, reducing weight transfer, while not having to stiffen the suspension more. (See the discussion of Bump Rubbers under GRE's Extended Ranges features for an explanation of why I feel this is the right thing to do.) The only exception is the Nurburgring, where it's impossible to keep the car off the bump rubbers because the vertical loads are so high.
In general, on road courses, for the lighter cars (the Lotus, Brabham, Ferrari, and Eagle), I have settled on front spring rates in the 70 to 75 lb./in. range, and rear spring rates in the 100 to 120 lb./in. range. I adjust ride height so that the car doesn't bottom on the suspension anywhere except very high G places, like the entrance to 5A at Mosport or the first part of Eau Rouge. At moderately bumpy, moderate-G circuits, this means 3.5 inch ride height; at most circuits it's around 3.75 to 4.0 inches, and at The Ring I'm up near the upper limit of 5 inches.
Spring rates at the shallowly banked ovals can be very similar to those for road courses, with some asymmetry to help keep the right front from overheating. Generally I run the left front spring softer than the right front, while keeping the rears nearly the same. Overall proportion may put a little higher stiffness at the rear, again to help the right front stay alive.
At the steeper banked ovals, I will go as stiff as necessary to keep the car up off the rear bump rubbers. (GRE's extended ranges allow me much greater freedom with this; without it, I had to use long bump stops as psuedo springs.) Viewing replays from the rear suspension view can be very helpful here, as you can see the right rear suspension as it works. Some serious observation will help you identify the point where the suspension begins to bottom.
Also the dreaded F10 view, when driving or in replays, can help; when the driveshaft angles start getting to be more than a few degrees, chances are the right rear is starting to bottom.
I like to set up the car so that the driveshafts are pretty much level (ride height about .75 to 4.0 inches) at rest, and stiffen the springs until the rear suspension doesn't bottom. At Bristol, I'm using rear spring rates over 300 lb./in.!
On the steeper ovals like Michigan and California, running the left front very softly sprung gives more grip at the front and will help keep the right front tire temperature down in a reasonable range. Generally if I can get the right front and both rears to be roughly equal, with the left front considerably lower in temperature, I figure I'm in the ball park.