iRacing.com – New Shorttracks + Tire Model Blog
Following the release of several road courses, the iRacing development team is shifting their attention more towards oval tracks in the coming weeks as four new short tracks are scheduled to be released soon.
Fans of bumper to bumper short track racing will soon be able to race at Mosport Speedway, Lucas Oil Raceway, New Smyrna Raceway & Langley Speedway. This doesn’t mean that development on upcoming road courses is slowed down though as Oulton Park and Twin Ring Motegi are already well in development as well.
In related news, iRacing head honcho Dave Kaemmer has written up an extensive blog in the iRacing forums, addressing some of the criticism that has come up with the launch of the new tire model.
You can find his very detailed blog after the jump below, it’s worth a read even if you´re not an iRacing subscriber as Dave is certainly one of the most accomplished developers in simulation racing and provides some great insight into his work.
It appears there is some unhappiness (but also some happiness) with the changes to the new tire model (NTM) that were introduced at the beginning of the current season. I’m going to try to explain where we are with the NTM and address (at least some of) the current concerns. This will be a long post, but bear with me, if you are interested.
It’s important to keep in mind what the new tire model is, and what it isn’t. It is a theoretical model that makes very good tire performance predictions from a few easily measurable quantities. It is not a perfect reproduction of the tires supplied last weekend in the real world for any particular series. While the data (lap times, cornering speeds) that we see on television are certainly helpful in figuring out whether we have the theoretical model predicting stuff properly, it is most important to me that the same model works as we move from a NASCAR tire to a Formula Mazda tire, and that we only have to change the numbers that really do change in the real world, such as the curvature of the tread and sidewalls, the rim and tire radii, the amount of carbon black, silica filler and/or extender oils that might be in the tread compound, the glass transition temperature of the tread rubber, etc. The most important thing is that they behave like tires that might be supplied on a given weekend for a given series under as many conditions as possible. Actually, the most important thing is that they behave like tires under as many conditions as possible. The NTM is already a very good model in that regard. It is not perfect, nor is it completely finished. I’d be the first to tell you that there are areas where it still doesn’t predict as well as I would like, but I’m working hard on those areas, and I’m pretty happy with where it is at the moment.
I’ll move on to more details, with some explanation of the tire temperature readings, wear, and a number of other issues that are being debated in this forum.
Tire temperature readings
These are a source of quite a bit of confusion. In real life, temperatures are taken as soon as possible after a car has rolled to a stop in the pits (sometimes after removal of the tire), by pushing a probe into the tread rubber at three points across the tire, typically at 25% (left), 50% (middle), and 75% (right) of the way across the tread. Sometimes a team will take temperatures at the shoulders (edge of tread) to see if they are in danger of blistering the shoulder from running too much camber. These temperatures are taken at single points laterally on the tread. I refer to these temperature readings as “carcass temperatures”, because they represent the temperature a few millimeters into the tread rubber. In telemetry output, the readings are taken by infrared sensors and so represent the “surface temperatures” of the tread, which can be quite a bit cooler (or hotter) than the carcass temps.
In the iRacing NTM, the carcass temps (which are what you see in the garage or in the tire pop-up info, as opposed to in the telemetry output) actually represent an average temperature over one third of the tire surface, instead of the temperature at a single point. This is something I’ll need to fix someday, because it makes the temps look “wrong”, but the carcass temps are useful as is, because they include the shoulder temps, in a way. What’s wrong is that this almost always results in a higher center temp than edge temp. For now, don’t worry about this. It doesn’t necessarily mean your pressure is too high. In the sim, if you have an edge temp that is higher than the center temp, you are probably running too much or too little camber, or too little pressure.
Depending on the tire pressure and the weight on the tire, there is a patch of tread (the contact patch) that is flattened against the ground. What the carcass temperatures tell you is roughly where your tread contact patch is, and how hard you are working that contact patch. With higher pressure or lower weight, the contact patch gets smaller. With more weight or lower pressure, the contact patch gets bigger. If you lean the tire to the left, the contact patch moves to the left. If you lean the tire to the right, the contact patch moves to the right. To some degree, you can tell roughly how big and where the contact patch is by looking at the temps. The center temp is usually highest because often the contact patch covers a good bit of the center third of the surface, even if it has moved left or right, so all of the center is doing work, all the time. Comparing center temps around the car is a good way to see if the car is balanced, or if one or more tires are doing too much work. The left and right third of the tread are not usually both completely in contact with the road. If they were, then you would see nearly equal carcass temps (actually they could be a little higher at the edges, since the tread is often a little thicker at the shoulders). You will usually get more grip, if this is the case, by increasing the pressure. That makes the contact patch smaller, and gets the shoulders a little off the ground, which makes the contact more efficient (less edge loading).
If the left temp is higher than the right temp, then the contact patch is on average to the left of center on the tire surface, and likewise for a higher right temp. If the split between left and right is higher, then either the contact patch has moved more, or it might be smaller. You will generally get better grip with the contact patch moved left of center for a left turn, and right of center for a right turn. How much is a matter for testing, since it can depend on pressure as well. Read on.
A lot of the carcass temperature is due to rolling drag, which is higher with a larger contact patch, and lower with a smaller contact patch. On a NASCAR oval, the contact patch gets a lot bigger in the corners than it is on the straights, because of the very high loads due to the banking. This increases rolling drag in the corners, and so the carcass will heat up more there. But in the corners, there is also a lot of heat generated at the surface of the tire from sliding. Even when you don’t think you’re sliding, part of the contact patch is, and that generates a lot of heat. On the straights, there’s very little heat generated from sliding, so the surface cools off quickly. This all gets more complicated when you consider that the tire tread band can move back and forth relative to the wheel rim. As cornering forces are generated, the tire tread band is deflected sideways and leans over relative to the rim, which changes the tread’s camber angle, which moves the contact patch sideways. This means the contact patch not only is a different size in the corners, but it moves to the right in a left hand turn, and to the left in a right hand turn. With lower pressures, the sidewalls aren’t as stiff, and the tread band will lean over more. So if you lower pressures a lot, you’ll need to adjust the camber to get the contact patch back where it works best. If you raise the pressure a lot, adjust the camber. Lower pressure generally means more static wheel camber is required, higher pressure means less static wheel camber is required.
You might think I’m just going down a rat-hole, and getting off topic. This is all related to what your tires temps mean, though, so bear with me. The carcass temps are influenced by rolling drag and the sliding work at the tire surface. The contact patch isn’t the same size when cornering as on the straight, though, and it’s not in the same place either. So on the straights, you might get temperature buildup on the left edges (NASCAR example), since the wheels are cambered for the corners, and are leaning over too much on the straights. But in the corners, the contact patch should move to the right and get bigger, which means more rolling drag. Plus now some of the contact patch is sliding, which heats everything up even more. You have to try to read the tire temps in the context of an entire race lap. For example, the left front (sticking to the NASCAR example) is usually set at a ridiculous positive camber, because in the corner, that gives the best contact patch location, and therefore the best grip. However, you will see the outside edge running pretty hot, because not only does that left front have to travel down the straights at its ridiculous positive camber, but it’s doing so at a low pressure, which gives a large contact patch on the straights, and lots of rolling drag. In the corners, low pressure works on the left side because a lot of load transfers over to the right side tires, which makes the left side contact patches smaller (and better for grip). What you see in the carcass temps is the temperature changes over entire laps, averaged over straights and corners. So it’s not easy to make sense of it. Looking at the surface temps in the telemetry output can help identify what’s going on in the corners, since the heat at the surface is closely related to sliding, which mostly happens in the corners.
A lot of energy is flowing around during this process. Heat is generated from rolling and sliding. Some of that heat is absorbed by the ground as it comes in contact with the tire surface. Some heat flows between the surface and the carcass. Some heat convects away into the air, some of it radiates away (but not much). Heat convects into the air inside the tire, flows through the sidewalls and rim (as well as the inner liner, and the air inside the inner liner, if there is one). Of all the numbers involved in this process, the convection coefficients and the thermal properties of the road surface have the largest margin of error at the moment, although I have very decent and defensible quantities for both. It is amazing how very small changes to these numbers can create a lot of controversy. That is just like real life, I guess. The line between “it’s junk”, and “it’s perfect” can be razor thin. In all seriousness, we’ve got things pretty close when the cries of “it’s junk” and “it’s perfect” are equally loud. I do agree that the NASCAR tires don’t seem to be building enough heat currently, although in the beta they were building a bit too much. Just be assured that these temperatures are amazingly close to correct, given all the complexity that goes into them.
Wear is usually recorded as the amount of tread depth remaining in real life, but we use percent tread depth remaining to make it easier to move from one car/tire to another without having to know tread depths. Passenger car treaded tires have tread depths of between 8 to 11 millimeters, typically. Racing slicks are surprisingly thin. They might have only 2 mm of tread when brand new, even on a NASCAR speedway tire. That’s because the thinner the tread, the less heat buildup from rolling drag. If a NASCAR speedway tire had 3 mm of tread, it would likely overheat. Interestingly, when a tire is worn, it generates less heat from rolling drag (although a bit more from sliding).
The tire wear readings are also given for three points on the tire, and are actually more useful for determining what’s going on on track. That’s because the wear is almost entirely due to sliding, which happens almost entirely in the corners (or under braking or hard acceleration on road courses). So the wear gives a better picture of where the contact patch is while it’s working really hard, which is what we really care about. The wear is also an average over a third of the tire in the sim, as opposed to a spot check at a single point as in real life, so you’re likely to see more wear in the center than on either edge, for the same reasons I talked about above. Again, comparing center wear depths around the car will give you a good idea of where the balance is, or which tires are working hardest in the corners. The edge wear depths give clues about where the contact patch is centered on the tire while cornering, as well as how big the contact patch is. This can be very helpful in setting pressures and cambers, although don’t be afraid to experiment—sometimes the best performance can be had from some other compromise.
As for the wear modeling, it is really close to reality, I don’t care what you say! As an example, the skid marks are really close, and they come right out of the wear calculations. Ok, how dark a few microns thick layer of rubber looks is not something I know directly (and I suspect the skid marks are a bit dim at the moment—I’ve seen some others say the same), but that is easy to change, so any subjective feedback is helpful there. Actually, to be honest there are a couple more things I think are involved in the actual thickness (and darkness) of the rubber layer laid down, having to do with a few of the tread compound properties, but that won’t change things a great deal. More extensive track surface modeling will have the biggest effect. See below.
Lap time fall off on ovals/slow out laps on road courses
This issue is all about the rate of temperature buildup in the tires, and the glass transition temperature of the tread compound. There are many types of rubber, but they are all made up of long chain-like molecules called polymers. The many types of plastics are also polymers. In fact, the only thing that makes plastic hard and glassy, and rubber soft and rubbery, is temperature. Heat a piece of polystyrene (from a plastic model car, say) up to 300 degrees Fahrenheit (150C), and it will behave just like rubber (but don’t touch it!). Or cool a rubber band down to -150 F (-100C, roughly), and it will behave like a piece of plastic (don’t touch it either!!). Most polymers have a particular temperature range (about 50C wide) over which they transition from a glassy state to a rubbery state. The bottom of that temperature range is called the glass transition temperature, for obvious reasons. Polystyrene has a Tg (shorthand for glass transition temperature) of about 100C. Polybutadiene (a particular polymer of interest) has a Tg of about -100C. Since room temperature is typically 25C, you can see that polystyrene is a rigid solid at room temperature, while polybutadiene is like rubber. Polystyrene would not make a good tire tread compound (not on planet Earth, anyway). Polybutadiene also doesn’t make a great tread compound (it’s too weak at typical tread temperatures). But mix the two together into a copolymer containing both butadiene and styrene, and they make a great tread compound, styrene-butadiene rubber, or SBR, which was discovered during World War II. You can mix different amounts together to get rubbers that have different Tg’s.
Why do we care about the Tg of a tread compound? For two reasons. One, the strength of the rubber compound, which determines its suitability as a tread rubber, is related to how far above the Tg it is operated. Two, the higher the Tg, the more heat is built up in the tire as it rolls down the road (more rolling drag). Generally, for racing, we want to use the highest Tg we can, without too much heat buildup. NASCAR stock cars heat tires up more than just about any other type of racecar, since they are heavily loaded, and traveling at very high speeds. Plus the tires are enclosed inside fenders. So NASCAR compounds typically have to have a fairly low Tg, or the tires would overheat. Formula car tires, on the other hand, are mounted on very light cars, and they are hanging out in the fast moving air, so they don’t have as much trouble with heat buildup. A qualifying gumball has such a high Tg that if it’s a little chilly outside, it can crack! But at operating temperature, it really sticks. This is one of the reasons for tire warmers.
If a tire tread is operating too close to its Tg, it’s not rubbery enough and doesn’t deform enough around the little micro-bumps in the pavement to stick well. If it’s too high above its Tg, it’s too weak to stick well. Somewhere in between is the sweet spot where the tire has the best grip. A NASCAR tire’s sweet spot is not far above typical air temperatures, so that’s why the best laps are the first few, when the tire is coldest. As the tire heats up, lap times get worse and worse. The wear has surprisingly little to do with it, although once the tire gets very worn the grip does fall off pretty fast. The lap time fall off has mostly to do with how fast the tire is gaining temperature, which of course can depend on driving style and car setup. As I said earlier, I do think the iRacing NASCAR tires need to gain more temperature than they are currently, and I’m working on some issues related to that.
A softer racing slick might have its sweet spot between 180 and 200 degrees, so when first going out on track it can be a little slippery. One thing to keep in mind is that if you are not driving the car hard enough, it may be that the temperature will start to drop below the temperature of the tire warmers, and grip will go down. F1 drivers have been known to complain that when following the safety car, they can end up with their tires so cold that they don’t have enough grip to heat them back up! I have, since the latest build went out, discovered an issue with how I was modeling some of the properties of the tread compound that affect this process. So, with any luck, we should be able to make better road compounds pretty soon.
*Disclaimer – the whole section above is a massive oversimplification. The sensitivity of rubber to temperature is so great that many of the quantities involved change by several orders of magnitude as the tire goes from ambient temperature to 200F. If the model has a term that is incorrect in its relationship to some other quantity (like this should be the cube root of that, not the square root…don’t laugh), then it would be “completely broken”. Better (in my mind) to find and fix that relationship, than to make sure a lap at Charlotte is within one tenth of a second of what was run there last year. Not that that is unimportant, but it is not a good yardstick to decide whether the model is good or bad.
Right now in the sim, although there are differing levels of grip on different surfaces, the track surfaces themselves are not modeled as accurately as I would like given what the NTM can do. The thermal properties of asphalt and concrete, for example, are modeled using the same numbers for now, which isn’t strictly correct, but they are not dramatically different and I have had more important pieces to work on (and still do). I will eventually get these surfaces modeled more accurately, which should lead to a different “feel” and differing levels of temperature buildup and wear from track to track.
Technically, the track surfaces are not part of the NTM, but rather the NSM (new surface model). Just saying.
Grip levels/lap times
Lap times depend on tire grip, to be sure. They also depend on thousands of other variables, in addition to the thousands in the tire model that affect tire grip. We work tirelessly (well, sometimes I get tired, but I love my work) to try to improve the accuracy of the sim in every respect. In the perfect world, I would get these thousands and thousands of variables to be accurate to within a tenth of a percent. It’s actually a very cool goal to work toward, because I get to learn a lot. But to actually get there? That’s a pipe dream. Most of these numbers have never been measured in the real world. All we get to see are a few measurements that involve lots of effects all mixed together. Sometimes we get measurements from the world of racing. Sometimes these measurements appear in journal articles about the “viscoelastic master curves of filled SBR compounds”. For a theory to be useful (and the tire model is, at the end of the day, a theory about tire behavior), it has to be able to predict unknown things given some known things. It is good if it can predict things well, without having to measure lots of things that are hard to measure. If you had a theory that could start with the bond dissociation energy of a carbon-carbon single bond (the weakest link in most polymers), which is about 0.000000000000000000578 Joules (and while it isn’t easy to measure, it’s been measured, and it doesn’t change much), and you could (after assigning similarly reasonable basic quantities to many other things) end up predicting lap times at Charlotte in a Cup car to within a few seconds, you’d be frickin’ pumped. There are so many things right with this model I could go on all day. But I do need to get its lap time predictions to within a few tenths. And we’re working on it.
At the moment, we find that our sim predicts lap times that differ from track to track by a bit. Generally, we might be fast at Charlotte, and slow at Texas, for example. We do try to get things close before a major release, which is a good exercise, since it helps us close in on numbers in the model that shouldn’t have to change. These differences are also clues to other effects that we either are not taking into account yet, or for which we might have incorrect numbers. Maybe the asphalt surface at the real Charlotte has a lower thermal conductivity than at Texas. Its density might be different because it has a different type of aggregate in it. Maybe the sun was shining while those qualifying times we’re using were laid down. Maybe a cloud came over. Maybe this compound should have less carbon black in it (we’ll never know, because tire compound engineers are not allowed to talk on the telephone). The more clues we get, the better able we are to separate different effects out and identify what’s really going on. As you can see, this is not a process that has an end date. We figured it was far enough along that you would all prefer the NTM to the OTM, even with a few things yet unknown. I sure do, anyway.
As regards the grip on grass, I’ll leave you with this to ponder: it turns out that there is some data on the tear strength of different varieties of grasses, because apparently the weight gain of cattle is negatively correlated with the tear strength of the grass variety they are grazing on (in other words, cows don’t like grass that’s too chewy), so there have been some scientific studies. And while I don’t have this data in the sim yet (NSM), it’s sitting on my hard drive, just waiting for when I need it. You never know when some measurement will come in handy…
Changes from beta
There is a lot of discussion about what has changed in the NTM between the beta release on the NW car and the latest season’s release. A number of things having to do with temperature build up changed—one example is that in the beta release I was using a constant for the heat capacity of the tread compound, whereas it turns out that the heat capacity is a fairly strong function of temperature for these long-chain hydrocarbons we call rubber. I also changed the thermal properties of the track surface a bit, since the heat buildup seemed too severe in the beta, and I was able to find some better numbers for the thermal conductivity of asphalt aggregates. Now it appears that there’s not enough heat buildup, so I need to correct something else (probably the convection coefficients, but it could be a few other things).
Another thing which changed the feel quite a bit was a change to the sidewall stiffness calculations, especially regarding torsional stiffness. There were a couple of deformation modes I hadn’t taken into account, so in the torsional case, the stiffness’s were too high, and the model was under-predicting the fall off in cornering stiffness with increasing load. I’m still not 100% happy with that, I’ve already found some fixes that will improve things.
Generally, it is almost always the case that things you are unhappy about, we are unhappy about. Sometimes we get some really good data that we know is correct, and it leads us to an unhappy place. The only way out is to figure out what’s wrong with other data that we don’t know so well. We for sure are not going to change the model back to being more wrong just because we like it better that way. Then we’d never learn anything, and we wouldn’t make any progress.
Unfortunately, as we improve things, often car setups need to change. Nowhere is the line between “it’s junk” and “it’s perfect” more razor thin than with car setup. We do our best to come up with baseline setups that work well, and that are good starting points for your own experimentation. Look at it this way: in the real world, if Goodyear shows up with a new tire construction, your car setup is up to you! And that happens all the time. As do rule changes, etc. Coping with change is part of what makes racing interesting. Just wait until we have variable weather and changing track surfaces.
A few other things
What’s still not done? Accurate wear modeling during brake lockup needs work (flat spots, essentially). There currently is no advantage to driving with the brakes locked up, so I thought this was less important than some other things. Rest assured I will correct this at some point, and at that point you will really not want to be locking up your brakes. There is still not tire smoke being generated. I should be able to get that going pretty easily, although it will depend on the lockup code. The NSM.
When will this model be finished, and why is it being so hyped before it is? I started making and learning about racing simulations in 1987, when I started writing Indy 500 for EA. If I had not shipped a sim until I was completely happy with it, you would still not have that one, let alone anything I’ve worked on since. Sad to say, but I will die before I’m happy with a sim I’m creating. And you want it that way. And so do I. I think it’s best for everyone, though, if I have to ship things out from time to time, even though there are ways to improve it. I feel like I’ve made some breakthroughs since the latest build with modeling road course (i.e. softer) compounds—and we would like to get tires made for more cars as soon as we can. But we do need to build setups, test our already released cars, remake setups for those, etc. This has been a huge R&D project, and it is difficult to schedule. I hope this post helps you to understand some of what we face, why it’s worth waiting for, and how cool the model is already.
As for the iRacing 2.0 marketing launch? The reality of this market is that the media is very launch focused; it is difficult to get interest in covering iRacing when it’s already been out for three years. We need to get coverage from time to time, since our biggest problem is that nobody knows about us. I think you would have to agree that we have come a long way since our initial launch in 2008, and we deserve some coverage given how much better the sim is now than it was then. For those of you who are members, you’ve gotten to try iRacing 2.0 way before anyone else—it’s just that you saw it come out in little changes every three months instead of all at once. And you’ll get to see iRacing 3.0 before anyone else as well! But this is not an official announcement. Shhh.
I’m sure there are many questions, even if you’ve managed to read all the way to here. I will try to respond to them (maybe not with a post this long) if I can, although you probably prefer that I keep my head buried in “Physical Properties of Polymers Handbook”. Which is an outstanding book, by the way.