Most racecar drivers have some sort of racing-related hobby that to keep themselves occupied between races. For some, that hobby is a keen interest in sim racing. For others, it's cycling, running, or some sort of sport that keeps their bodies honed for the season.
For yours truly, that hobby is the fine art of racecar engineering. It's the main reason why there are so many chassis and suspension related articles on our Racing Secrets blog on StudioVRM.net. I love learning about the intricate mechanics that make racecars work and making the fine adjustments that they need to go faster.
So you can imagine my excitement when we found out that Australian racecar engineering company ChassisSim was running a contest for racecar engineers. And it wasn't just any run of the mill theoretical contest for college students. It was a competition to see who could produce the best setup for a real SRO GT3 car using their pro-grade simulation software. Anyone of any level could enter from anywhere in the world, and there was actual prize money for the winners.
For me, this was a chance to test my hard-learned skills against the best and brightest racecar engineers from across the globe. The 2020 edition of the ChassisSim competition attracted 150 competitors, many of whom were professionals or automotive engineering students.
It would be a tough ask for a self-taught race engineer like myself to stand a chance against a field of hardened pros and pros-to-be. So I set some realistic goals - If we didn't come in last place, we would treat it like a win.
With our sights set firmly towards level ground, we put down the entry fee and dove head-first into the world of professional racecar engineering.
Table of Contents
Because this story ended up being so long, I created a table of contents so you can quickly jump to the different chapters. Just click on the chapter headings to skip to each section.
What is ChassisSim?
Those of us who have ever done any setup work on a racecar (whether in sim racing or in real life), you know that your car's setup is only as good as your test driver. If your test driver is fast, sensitive, and honest, you can trial-and-error your way to a fast setup. But if your test driver isn't a perfect machine capable of turning out metronomic laps, you may be left guessing as to whether your changes actually worked. Unfortunately, not all of us can afford to hire top-tier test drivers to set up our cars.
That's where ChassisSim comes in. ChassisSim is a software simulation suite that lets you test upgrades and setup changes to racecars on a regular home PC. You do need to take some fairly precise measurements off of your car in order for it to work. But once you have that, you can make any changes that you would want to test on your car, and a virtual test driver will take it around the track of your choice and show you exactly what differences it made.
And when I say exactly, I mean EXACTLY. ChassisSim simulates the cars and tracks down to its smallest details, from the torque curve of your engine to the smallest imperfections on the track surface. Each run produces a full data log, revealing everything from engine rpm to steering angle to the exact ride height of the car at any given point in time. It even has its own built-in metric called the Index of Stability which tells you how stable or twitchy your car is through the corners. If that isn't enough, ChassisSim has a built-in 7 post rig for testing suspension systems and a Driver-in-the-Loop simulator so you can actually drive your newly set up racecar in the virtual world.
As you can imagine, this is a huge time and money saver for race teams. Engineers can see what an alignment change or a power-adding engine upgrade will do before they lay a finger on the car. If an upgrade package works out, great. Put it on the car. If not, that's ok. Just go back to the last setup and try something else.
What's even more incredible is how well the results correlate to the real world. A few random queries on LinkedIn (of all places) revealed that the laps they simulated in ChassisSim were within a hundredth of a second of what they achieved in real life.
Needless to say, it's an incredible piece of software. All it needs is a skilled race engineer to get the best out of it. And that's the challenging bit.
Sizing Up the Challenge
The rules for the 2021 ChassisSim Engineering Competition were simple: Set up a race car so that it laps the Mount Panorama Circuit at Bathurst in the fastest time possible, regardless of whether it's driven by a computer or a human driver.
The organizers would give us a model files for a car (a mid-engined, RWD GT3 car), model files for the track, and 100 simulation runs on ChassisSim Online so we could test out our changes. We were allowed to make as many setup changes as we wanted, so long as we stayed within that allocation of 100 simulated laps.
The final submissions would first be put to the test by ChassisSim's simulated driver, which would punch out the ideal fast lap for each setup. Then, a professional racecar driver would take the wheel in a Driver in the Loop simulator and do a flying lap with the same setup. Contestants would be ranked based on a combination of the two results.
Sounds easy, right?
It would be if we could upgrade the car like a college student playing Forza Motorsport.
In order to keep the cars realistic, the contest runners added a few extra rules to this year's competition. These extra rules meant:
No changes to the weight or weight distribution of the car
No changes to the engine or gearbox
No changes the wheelbase or the front/rear track
A minimum dynamic ride height of 20mm on all four corners
A spec tyre with no adjustments allowed
Limited adjustability on the wing
No automagic active suspension or computerized differentials
No adding a hybrid system, converting the car to AWD, or changing the engine layout
Limits on how far you could move the inboard and outboard suspension pickup points (+/- 50mm on the inboard pickup points, +/- 10mm on the outboard pickup points)
In other words, we couldn't make the car go faster by slapping on a big turbo or by taking a ton of weight out of the car. We would have to earn our lap time the good old-fashioned way - with careful adjustments to the suspension system, brakes, diff, and all of the other settings that you can normally adjust on a real-life racecar. It would be hard work, to put it mildly.
But the blood sweat and tears would be well worth it. The contest offered generous cash prizes for the Top 3 finishers (in US Dollars, the most expensive of dollars) and a big pile of free ChassisSim simulation runs for all contestants who finished in the Top 10.
Game on then.
Selecting the Tools
ChassisSim is a tool designed for simulating a car's behavior. While you don't need a particularly powerful computer to run it, you do need a few other tools to get the most out of it. For this contest, we would need a good data analysis suite that would help us analyze the data that ChassisSim produced, as well as an organized way of keeping track of the setup changes that we were planning to make.
Because I had no particular loyalties to any data analysis platform, I chose the most user-friendly of data analysis suites: Motec's i2 suite. Even the (free) standard version of i2 is powerful enough to give you all of the graphs and graphics you need to turn the data logs from ChassisSim into visually friendly graphs that you can configure to your heart's content. The aesthetically pleasing UI is as intuitive as it can get - If you know how to use the data logger feature in Gran Turismo or Assetto Corsa, you already know how to use Motec i2.
As for my digital notepad, I chose Microsoft's venerable (and free) OneNote. In addition to supporting support text, images, videos, and voice notes, it has a built-in calculator that parses out mathematical equations - a very useful feature for those few occasions when you need to plug some numbers into equations.
Strategy Session
Now that we had our tools on the table, it was time to figure out how we would use them to tackle this challenge. 100 simulation laps sounds like a lot, but you will burn through them very quickly if you spend all of your time taking shots in the dark.
I quickly realized that the only workable approach would be to set this car up the same way that I set up real racing cars:
Take a series of baseline measurements and understand the car's fundamental behaviors
Change springs, dampers, and ride heights to glue the car to the track
Stabilize the car with the alignment, differential, anti-roll bars, and aero
Step back and look at the car
Make more intrusive changes to items like suspension pickup points
Test out any other ideas that come to mind
Of course, I knew there was no guarantee that a 500hp mid-engined GT3 car would respond to setup changes the same way as the 200hp club racing sedans that I was used to working with. But I wanted to see how my approach stacked up against the practiced hands of the pros. So with that settled, I fired up ChassisSim and loaded up the contest car for the very first time.
Setup Step 1- Understanding the Beast
First order of business - Load the car into ChassisSim and find out what we were working with.
The car model file had no make or model in its description, so it was anyone's guess as to what real-life car it was based on. What we did know was that the mid-mounted engine made 509 hp at 7500 rpm and the whole car weighed 2866 lbs. Those numbers sounded vaguely like the specs for a GT3 Lamborghini, so I pretended it was a green and yellow Huracán GT3.
The specifications of the default suspension setup read like that of a Front Wheel Drive touring car, but in reverse. A 44/66 front-to-rear weight distribution meant that most of the car's weight was supported by the rear wheels. ~1300 lb-f/in front springs and 1500 lb-f/in rear springs held the car up at all four corners, with a 1066 lb-f/in front anti-roll bar paired with a relatively tame 550 lb-f/in rear anti-roll bar.
Fortunately, there was nothing in the car that we weren't familiar with. No funny setup trickery, no third springs / dampers, no weird F1-style front-rear interconnected suspension. Even the differential was a standard adjustable clutch-pack setup. It was reassuring to see that there were no surprises.
Reassured by the familiarity of the car and full of confidence, I loaded up the provided track file and ran my first-ever simulated lap to get a baseline lap time. The software scrolled through a flurry of numbers before displaying a surprisingly fast lap time - a 2:01.487 (??). This was the first of many surprises in this contest. My baseline lap time was a full tenth of a second faster than Shane van Gisbergen's real-life GT3 lap record around Mount Panorama Circuit. How on earth were we supposed to make the car any faster than that?
Something had to be wrong. Thankfully, I had the presence of mind to email the contest organizers and validate the baseline lap. This turned out to be a good move. The organizers confirmed that this lap time was not what they expected either.
It turns out I had somehow loaded the track file given to us for the contest with the built-in Bathurst circuit file that came with the ChassisSim software. The resulting mis-mosh of a track was missing many of the track's trademark bumps and elevation changes, making it faster than it should have been. It was a classic case of operator error, caused by a user who didn't know how to use the software.
And so, the second order of business became to understand the software itself. Make no mistake - ChassisSim is a pro-grade simulation tool. It has so many features, options, and dialogs that you'll never be able to figure it out by clicking around. Fortunately, the creators of ChassisSim have spent the last 10 years building up a series of beginner-friendly video tutorials on their YouTube channel that show exactly how to use every single option and screen.
So for the first week, yours truly spent his commute listening to ChassisSim Director Danny Nowlan passionately explain the ins and outs of ChassisSim in a cheery tone of a man who clearly loves his job. Yes, that was a solid 3 hours of every day watching videos, but it was time well spent. The ChassisSim tutorials use real-life examples, and even offers suggestions on what settings to use for what situation. Some of them, like this one on the Damper Workbook, even have formulas in them that show you exactly what numbers to put into the software to get the best results. They are excellent chassis setup videos, and I would recommend watching them if you have any serious interest in learning how to set up a racecar.
Armed with a newfound understanding of the software, I loaded the car and track into ChassisSim and re-ran the baseline lap. This time, the software spit out a lap time of 2:04.175 - exactly what the organizers said it should be. Success. We were finally at the starting line. Time to get to work.
Setup Step 2 - Giving the Car Legs
Despite never having set foot in Australia, I have always loved Mount Panorama Circuit at Bathurst. There is no other track like it in the world. The long, fast straights leading up to a narrow concrete-lined climb up a literal mountain, the bumpy winding road across the summit of Mount Panorama, and that hair-raising descent that opens into the highlands of New South Wales all come together to make the track a legend amongst motorsports enthusiasts and sim racers worldwide.
Finding a fast lap would involve setting the car up so it would be compliant enough to absorb the track's many bumps while being stiff enough to serve as a stable platform for the car's aero. We had no idea whether the default suspension was stiffer or softer than it should be, but we didn't want to make any assumptions.
I started by increasing the spring rates by a substantial 15% front and rear, to see what it would do. Unsurprisingly, this hurt lap times - a 2:04.895. So we tried the opposite, reducing the spring rate by 10% front and rear. Somehow, this was worse. The car was even slower, returning a lap time of 2:05.035. What was going on?
We opened up the logged data in Motec i2 and looked at the data traces for suspension damper travel. The car appeared to be skipping over some of the smaller bumps on the flat sections of the track. And softening the springs made the skipping worse. This didn't make any sense. I must have screwed something up while editing the setup dialogs.
One of the advantages of using ChassisSim is that you can undo changes very easily. So that's exactly what I did. I loaded up the original baseline setup that the organizers had sent and lowered the spring rates 10% from the original setup. This worked. Just by swapping in a softer set of springs, we had gotten the lap time down to a 2:03.622 - A full 0.5 seconds faster than the baseline setup. Nice.
In the next few runs, we tried reducing the spring rates until our GT3 car rode like an American luxo-barge. While this did help soak up the bumps, it also caused the car to pitch and sway so badly that the aero balance of the car would change as soon as you hit the throttle or the brakes.
Our solution was to install tender springs on top of the main springs on each corner. This would effectively make the car behave as if it was on super-soft 700lb-f/in front and 820 lb-f/in rear springs for the first 25mm of its suspension travel. When you compressed the suspension further (e.g. under hard braking or when the car would squat down due to aerodynamic loads), the car would behave as if it was riding on 1150 lb-f/in front and 1350 lb-f/in rear springs. This setup yielded a 2:03.605 - Only marginally faster than the single spring setup. But the steering and throttle traces in Motec i2 showed that this made the car much easier for the driver to handle.
We then briefly flirted with the idea of using a third spring to keep the car flat under hard braking. That idea went out the window when we realized that even the softest of third springs would cause the front wheels to skip under braking. Instead, we changed our focus to the dampers.
Damper tuning is often considered to be a fine art rather than a science. That's probably why most people look so surprised when they find out that there is a set of well-known mathematical formulas that will tell you what damping rates to use based on the mass of the car, the spring rates, and the motion ratio of the suspension system. And thanks to ChassisSim's Danny Nowlan, there is a step by step guide on how to use the formula in one of ChassisSim's tutorial videos. Which means that I didn't even need to go look for my old suspension setup textbooks.
I punched the numbers from the latest setup into Microsoft OneNote, accidentally discovering that OneNote solves math equations for you if you put an equals sign at the end. How handy. I then put the resulting numbers back into the Damper settings screen in ChassisSim to get my newly revalved suspension dampers:
The magic formula worked. The new damper settings took another huge chunk of time out of our lap time. The car was now running a 2:03.272, almost a full second faster than it was originally. And we were just getting started.
Setup Step 3 - Stabilizing the Platform
I messed around with the dampers some more in an attempt to eke out a tiny bit more performance. But my usual method of playing with the high-speed rebound and compression adjusters didn't seem to make a tangible difference in lap time with the AI driver behind the wheel.
So we turned our attention to other matters. The wheel speed traces in i2 showed that we were getting asymmetric wheelspin across the rear wheels through the uphill section of the track. The limited slip differential wasn't working hard enough under full throttle.
Admittedly, I have very little experience in tuning a clutch pack differential. That inexperience was made painfully obvious as I fumbled around with the max diff wheel spin and locking ratio numbers in an almost-random way in the hopes that it might result in a faster lap. My best attempt resulted in a disappointing 2:03.615.
Fortunately, ChassisSim offers a very simple and effective option for people for the differentially challenged - A locked diff.
Locking the rear diff is a surprisingly common option for high-powered RWD race cars. In addition to being brutally effective in putting down power, it makes cars extremely stable under hard acceleration and hard braking. Not to say that it isn't without its downsides. A locked diff also makes RWD cars less willing to turn into corners and can result in some snappy handling if you don't manage it carefully. But given my personal inexperience and the limited number of runs allowed by the competition, we chose to take those risks and lock the rear differential.
The locked diff hurt lap time slightly, pegging us back to a 2:03.570. But it was worth it. Acceleration down the track's long straights had improved dramatically. We would just have to make up time elsewhere.
In an effort to help the car turn in, I dialed out the front toe-in that the car came with and applied a tiny bit of toe-in to the rear wheels. I also experimented with some of the higher downforce wing settings, lowered the ride height, turned the brake bias back to 52/48, and tried a few different anti-sway bar settings. Lap times incrementally dropped to 2:03.235. We were slowly going faster, a few hundredths of a second at a time.
I closed my eyes and daydreamed while ChassisSim crunched through the numbers on my latest setup. If only there was something out there that would make us a few tenths... or maybe even a whole second, faster...
Then it hit me. I missed a trick with the third spring. Yes, the whole third spring experiment was a bit of a disaster. But what if, instead of installing a third spring, we installed a third damper with no spring attached to it? With enough high-speed bump damping, a third damper would the sudden nose-diving that we were getting under hard braking.
Yes, it was a bit of a gimmick, but how bad could it be? I took an educated guess at the damping rates, attached a third damper to the front suspension of then GT3 car, and excitedly punched the Simulate button to run the car through a lap. My jaw dropped at the result - a 2:02.215 (!!) - over a full SECOND faster than our fastest setup so far.
Of course, this was far from a slam dunk setup change. The Stability Index trace showed that the change had made the car much harder to drive, as it wanted to dance under hard braking through turn-in. But it was impossible to ignore the impact of the change. We were on to something. All we had to do was to refine it.
Softening the front anti-roll bar made a massive improvement to drivability, at the expense of a few tenths of a second per lap. The car was back to running a 2:03.032. But we knew we could unlock more speed. The data showed that the car's camber was changing dramatically as it went around the track, so we started making adjustments to tackle that instead. Take out some static camber, adjust the front springs so they ramp up to full rate faster, raise the front roll center, lower the rear roll center...
The drivability was improving with every iteration, but every setup change was making the car slower and slower. By the 30th simulation, the car slowed down to a 2:03.825. Not good.
Worse yet, we were out of ideas.
Setup Step 4 - Breaking through the Wall
We had hit a brick wall in our setup, and I had already used up a third of the 100 simulations that we were allowed to use. No amount of staring at the data would get us past this roadblock. I needed outside inspiration. And something to help me regain my confidence. The gradual decline in performance was making me start to question whether I had any idea what I was doing.
Surprisingly, validating my own sanity was the easier of the two to-dos. One of the default car models that comes with ChassisSim is a Lamborghini LP560 built for the same GT3 class as our contest car. It was lighter and has slightly different specifications to the car we were using for the contest. But it was close enough that we could at least see if our car was in the right ballpark.
We loaded up the Lambo into ChassisSim, put it on the same track that we were using for the contest, and hit the Simulate button. The car turned a scorching fast 2:01.972.
This unexpected result caused a brief moment of panic before I realized that the default Lambo had something the contest car didn't - a Super Diff. Apparently, the designers of ChassisSim were fully aware of the frustration that comes with tuning a racecar with an adjustable limited slip differential. To combat this, they created a setting that would allow engineers to temporarily eliminate the differential as a variable. This setting, called "Super Diff", delivers 100% of the car's power to the ground at all times. That explains why the car was so unbelievably quick.
I quickly replaced the Super Diff with a locked rear diff, and raised the weight of the car to the same 2866lbs as the contest car. This resulted in a 2:03.990 - Slightly faster than the baseline settings for the contest car, but almost a full second slower than what I had managed after 25 simulation runs. Seeing this was a huge relief. I wasn't completely out in the weeds. My setup strategy was indeed working. I just needed to take it in the right direction.
Finding the right inspiration would be slightly harder. Aimlessly flipping through motorsport magazines and technical journals was probably not going to work. I needed something more tangible. And more solid.
Fortunately for me, I had something very tangible and very solid sitting right next door - The StudioVRM Honda Prelude. So I threw on my coveralls and sauntered out to the garage to seek inspiration from the car that literally and figuratively carried me for the last 10 years of my racing career.
I opened the hood, put the Prelude onto jack stands, and stared at the array of mechanical metal bits from the underside of the car. I closed my eyes tried to visualize how the GT3 car was running, based on the steering, throttle, and brake traces in the data logs. "If the Prelude was handling like that in real life, what would I do to fix it?"
I went to bed that night without any answers. The following night, I went back out to the garage and did the same thing. The Prelude needed new brake pads anyway, so I had a good excuse to wrench on the car at midnight. That's when I realized that the Prelude and the ChassisSim contest car had something in common - Both cars carried most of their weight over their driven wheels.
Being a front wheel drive race car, the StudioVRM Prelude has a front / rear weight distribution of around 65% / 35%. It's a very front-heavy car. As a result, we have to run our brake bias all the way forward to prevent the rear wheels from locking up under hard braking. The GT3 car that we were setting up for the contest was similar, but in reverse: Its weight distribution was 44%/56%. What if we turned the brake bias rearward and made the rear wheels do most of the braking?
The idea itself isn't particularly unique. In fact, it's somewhat common for older mid and rear-engine production cars to be have their braking systems biased towards the rear of the car. We would just be taking it to the extreme. The locked rear differential and rear toe-in already made the car stable under braking, so the driver would be able to tolerate a fair amount of rear brake bias.
In preparation for this experiment, I rolled the car back to an earlier setup and made some alignment changes to help stabilize the rear end. Front toe was now at 0.05 degrees out, rear toe was now 0.15 degrees in. This preparatory alignment tweak yielded a few hundredths of a seconds of performance on its own, lowering the simulated lap time to a 2:02.962.
I started tentatively, turning the brake bias rearwards, so that the front brakes handled only 48% of the braking. For the first time since the start of the contest, the rear wheels of the car was handling the majority of the car's braking. This yielded some small gains - a 2:02.892. The rear brake bias did make the car slightly less stable under braking, but it also saved a chunk of time under braking. So what would happen if we turned it back further?
So we tentatively turned the brake bias back some more to 42% front. We expected this setting to be somewhat scary over the swoopy, hilly section at the top of the mountain. But to my surprise, the AI driver had to do surprisingly few corrections to keep the car on the fast line. And it did so while turning a 2:02.740 thanks to the fact that the driver could hit the brakes harder and stop the car faster. ...what if we took it even further?
38% front brake bias seemed almost reckless, but the mad scientist inside me had to know what would happen. So I tried it. And boy, was I happy that I did. ChassisSim's virtual driver hammered out its fastest lap yet: A 2:02.682. Despite being slightly slower through the top of Mount Panorama, the car was now more stable than it was before. It turned out that the increased brake bias was getting the tyres hotter early in the braking zone. That extra tyre heat would get the tyres nice and sticky for the entire duration of the corner, resulting in more grip through the corner and faster acceleration while tracking out onto the straights.
It also lessened the amount of nose-diving that the car did under hard braking, which meant that we could lower the static ride height even more. The final result of the brake bias experiment was a 2:02.585. The experiment was a resounding success.
That night, I went back out to the garage with a renewed vigor and a bottle of waterless car wash. I then spent the last few minutes of my evening cleaning off the dirt that had collected on the Prelude's paint during its last track outing. It was the least I could do, after it had given me the inspiration I needed to break through the my mental roadblock.
Setup Step 5 - Super Fine Refinements
By this point, the car was a full 1.5 seconds faster than it was when I started. While this doesn't sound like very much to your average club-level racer, 1.5 seconds is huge in the world of manufacturer-built grand touring cars. It was time to stop making big changes and start focusing on the small, incremental refinements that separate the winners from the backmarkers.
I experimented with the different rear wing settings and adjusted the ride height to match. The car proved surprisingly sensitive to aero adjustments. Too much wing, and the car would lose speed through Bathurst's long straights. Too little wing and the car would want to launch itself into the air through the hilly mountain section.
We also ended up turning the brake bias back even more. We were now at just 35% front brake bias. Bizarrely, the car responded well to this change. Not only did this make the car faster, it also this reduced the amount of steering corrections the AI driver had to do through the corners. A few anti-roll bar adjustments later, and we were down to a 2:02.005. The car was really coming together now.
It seemed that the car was happiest when I made the rear end work harder. Naturally, I wanted to see what would happen if I really kicked its butt. Bringing the rear roll center closer to the car's center of gravity would do exactly that. A few small adjustments to the inboard and outboard suspension pick-up points would do the trick. While I was in there, I added a small percentage of anti-squat to keep the car level under hard acceleration.
I chuckled slightly as ChassisSim churned through the numbers. The simulation software made testing these highly intrusive suspension modifications so easy that it literally made me laugh. Adjusting the suspension pick-up points on a touring car isn't supposed to be this easy. The changes I was making should take weeks to test and costs hundreds of thousands of dollars due to the sheer number of experimental suspension parts that you have to fabricate. And at the end of the test, you always end up with a giant pile of expensive, unusable suspension parts that you would have to chuck in the trash.
Because I had access to ChassisSim, I wouldn't need to do any of that. I could just play with the numbers as much as I wanted in the simulation, figure out exactly what measurements I needed, and build production-ready parts using the numbers from ChassisSim. Imagine how much money that would save. $1000? $10,000? For a pro race team it would probably be closer to $100,000. The numbers were absolutely mind-boggling.
All I knew was that it got the job done. Our GT3 car was now lapping Mount Panorama Circuit at a blisteringly fast 2:01.705. Our car was now faster than a lighter, more powerful Lamborghini LP560 equipped with its magic Super Diff. And we had only used 50 of our 100 simulations so far.
I continued to make adjustments to the front and rear suspension geometry, shifting the pickup points to raise and lower the roll centers and adjust anti-dive and anti-squat. I also made some small adjustments to the dampers, which I ended up undoing after finding out that they yielded little to no performance benefits. Bit of a shame. They looked so promising on the damper histogram.
ChassisSim supports a Mercedes F1-style Front-Rear Interconnected (FRIC) suspension setup, and I would have been remiss to ignore the opportunity to play with that option. I installed a small FRIC spring, ran the simulation, and immediately crashed the ChassisSim software. A friendly email conversation with Danny Nowlan revealed that I had forgotten to enable some necessary parameters to get the FRIC setup to work, and the lack of required settings were causing the software to chuck a wobbly. Correcting the issue revealed that FRIC offered very little benefit for a GT3 car around Mount Panorama Circuit. It wasn't all wasted effort though: I learned a bit of Australian slang.
Another pass through the alignment settings revealed that I was going the wrong way with the car's camber settings. Dialing in some additional negative camber eliminated some understeer and brought the lap time down to a 2:01.660, or about 2.5 seconds faster than where the car had started. It was the hardest I had ever had to work to make any racecar go 2.5 seconds faster through a 2-minute lap. And it was all worth it.
Setup Step 6 - Panic at the 11th Hour
By this point, it had been three full weeks since the contest started. The deadline for the contest was over 6 weeks away. But at this point I was well and truly running out of ideas.
I made some small tweaks to pass the time, changing anti-roll bar rates and tried to pull a few cheeky tricks with the bumpsteer curve. Nothing seemed to make the car any faster or easier to drive. I considered putting this car aside and making a second, separate setup off of the baseline car to see if I could get even better results. But curiosity got the better of me. I wanted to see how well this first attempt would hold up against a practiced group of pro engineers.
I scrolled through all of the 66 setups I had tested, grabbed the fastest one, and submitted it to the contest organizers. I would have to force myself to think about other things for a while. After all, the contest deadline still a full month away. It would be a long while before I heard back about the results.
Or so I thought. A few hours later, my phone buzzed with an email from ChassisSim Director Danny Nowlan. He had taken a look at our car model, and found a problem with our submission.
Uh oh.
It turned out that the rear suspension pickup points and the front static ride height were outside of the allowed spec. Somewhere along the line, I had mixed up the limits on how far I could move the inboard pickup points and moved the outboard points too far from their stock location. And to make matters worse, the car file I had sent in still had the Super Diff activated from an earlier experiment.
Technically, this could have been grounds for an automatic disqualification. Thankfully, Danny was as magnanimous as he is masterful. He offered to let me fix our clumsy mistake so we would have a legal entry for the contest.
I immediately fired up ChassisSim, hastily put the suspension pickup points back to where they were originally, and put the locked diff back in the car. The patch fixed car ran a 2:02.482 - Significantly slower than our best attempt to date. And it was easy to see why. The aero balance of the patched-up car would fluctuate wildly over the course of a lap, and resulted in poor stability through every braking zone. It's amazing what happens to a car when you move the suspension pickup points by a few milimeters. This was no time to marvel at the mysteries of automotive suspension systems. We had to fix the car.
The idea behind moving the outboard pickup points was to raise the rear roll center and bring it closer to the car's center of gravity. We had lowered the outboard pickup points of the rear lower control arms by a substantial amount to achieve this. Maybe we could achieve a similar result by raising the inboard pickup points of the lower control arms instead? There was no time to wonder. We had to try it.
I raised the lower control arm pickup points by a full 40mm and ran the simulation to see what it would do. The wild shifts in aero balance had subsided, and as a result, the car was now significantly faster over the high-speed sections over the top of the mountain. More importantly, the lap times were back down to a 2:01.745. Not bad. But not quite good enough.
If we were going to make a serious attempt at this contest, we would need to push the limits. So I made the call to go all-in on performance. I moved the inboard pickup points for the rear lower control arms upwards by another 9mm, right up to the limit of the rules. I also lowered the upper control arm mounting points a bit to make sure it would fit within the limits of the rules. I triple checked all of the settings to make sure they complied with the rules, and re-ran the simulation. The bug-fixed setup did a 2:01.475 - The fastest the car has ever been during the entire duration of the competition.
Submitting this setup would be a gamble. The index of stability trace on the debugged car showed big spikes that weren't there before - Indicative of a car that was much less stable through the corners compared to our original submission.
There was a good chance that the human driver would absolutely hate the car. But time was ticking, and we couldn't keep Danny waiting. So I packaged up the latest car model file and submitted it with a note thanking him for giving us a second chance.
The Big Reveal
Our second submission was thankfully accepted, and there was little to do except wait for the results. Yours truly tried his best to keep the contest out of mind, instead spending the time working out the bugs in the StudioVRM Honda Prelude and giving our ProjectCRX endurance racer a proper shakedown at VIR. We knew it would take some time to get a pro racing driver to do the Driver in the Loop tests, especially as prevailing conditions were causing more lockdowns across Australia at the time.
Then, out of the blue, this message popped up in our team's Outlook inbox:
I could feel the tears of joy welling up in my sleep-deprived eyes as I re-read Danny's email over and over again. Remember, I'm not a professional race engineer. I'm a club racer who studied this stuff in his limited free time. We were the minnows in this competition. The European Minardi F1 team of the club racing world. I would have been happy to finish in the top 50% of this field.
And yet, I was getting a personal congratulations from one of the most well-respected names in the industry for narrowly missing the podium in a no-holds-barred racecar engineering competition. I was so happy that I didn't know how to react.
I sent an overly excited thank you note to Danny and immediately hopped onto my social media accounts to tell my friends... all while forgetting that the official results hadn't been made public yet. Oops.
Thankfully, the official results came out on the ChassisSim YouTube channel a few short days later:
With it, we received our prize - A license for 25 simulation runs on the ChassisSim Online simulator, so we could use it on our real-life racecars.
The All-Important Debrief
As soon as I was off of the high of scoring a top-5 finish, I asked myself the big question: "What went well and what could we have done better?"
While I don't have any detailed feedback from the judges at ChassiSim, it was pretty obvious that our entry was one of the "qualifying special" setups - Blisteringly fast over one lap; Extraordinarily difficult to drive over a full 60-minute race.
The last-minute suspension adjustments played a big part in why the car was so difficult to drive. It would have helped to spend more time studying the angles of the rear control arms to see if there was a way to achieve the same speed without making the car so peaky in the corners.
It was also clear that I didn't spend enough time tuning the suspension dampers. The fundamental strategy of using soft springs and copious amounts of high-speed damping worked ok, but the reality is that we missed a few tricks by not experimenting with different approaches there. I had enough simulations left over to play with the shocks some more. I just didn't do it.
There was a good chance that the car would have been slower over one lap regardless of what I did. But it would have been faster when the human driver got behind the wheel in the Driver in the Loop segment of the competition. I'll take that tradeoff next time.
The good news is that there were lots of positives to take away.
The first was that we nailed our choice of ancillary software. I had seen Motec i2 in the past, but never used it for myself. What a great piece of kit. It's so intuitive that you don't even need the instruction manual. Because I didn't have to spend any time learning how to use i2, I was able to spend all of my free time focusing on setting up the car in ChassisSim.
The second was the realization that the methods that we club racers (and sim racers) use when we tune our cars translates to real pro-level racecars. Sure, the cars are more sensitive to changes and the margins tend to be finer. But the basic concepts still apply. So if you are a student struggling to understand the fine art of automotive engineering, rest assured - Your long hours buried in the books will pay off one day.
But the best thing, by far, is the fact that we learned a tremendous amount about racecar setup and tuning. The virtual driver in ChassisSim is literally the ideal test driver. It's lightning fast. It doesn't make mistakes. It doesn't guess. And it produces the same results whether it's lap 1 or lap 100. It always gives honest feedback. If you make an adjustment that makes the car faster or slower, you know that it was whatever you did that made the difference.
Because I could get honest, accurate results for every single change, the two months I spent working in ChassisSim made me a better engineer. And as a result, our team's real-life racecars will go much faster in the future.
How do YOU enter?
Want to try your hand at ChassisSim? You could wait until the next competition. But if you are serious about racecar engineering (or just really enjoy making fast cars go faster), I wouldn't wait. A starter pack of simulations on ChassisSim Online costs less than a set of racing brake pads. My recommendation is to buy a bunch of simulations, download the software, and try it for yourself.
Here's their info, in case you have any questions:
ChassisSim Technologies
(Australia): +61 425 219 375
(US): 678.671.6615
Who knows? Maybe your name will be immortalized in the winner's circle next time.
In any case, that's all for today.
Thank you very much for reading.
I will see you at the track.
Disclosure:
Roger Maeda and StudioVRM.Racing are not affiliated with or supported by ChassisSim Technologies or by any of their partners or vendors. All entry fees and expenses were paid at full price out of the team budget, which currently comes out of Roger's own pocket.
コメント