Racing Games: A Sound Study


In The Beginning
Damian Kastbauer approached me sometime last August about creating what he called a “Sound Study,” which is a scholarly-like look at current-gen sounds in games to deduce what is sonically important and to infer ideas of how those sounds were implemented. Damian’s first study was on footstep & movement sounds in games, which he had posted on his blog and later presented at GDC 2011 where his presentation was very well received for being informative and interesting. Naturally, when he told me he wanted to focus his next study about racing games and wanted my help, I was all in.

Our study began by first dissecting what sound is doing in racing games in the first place. Or to word that another way, “Why sound?” It seems that the sound of the car in a racing game is inherent, but do we really know why?

After logging many hours in various racing games, I believe that there are a few absolutely crucial functions that sound plays in regards to game mechanics. The first and most important role is that sound has to fill the communicative void between real world driving and game driving that is physical movement. Only a very lucky and rare few people have access to moveable chairs that can apply the precise G-forces that a racing car experiences on track such that their bodies can feel the sensation of the car moving in each direction. For the rest of us sitting in our little rigs or on our couches or playing on our smart phones, something else has to communicate the lateral and longitudinal force information that is crucial to driving the car. That role is filled largely by sound, and largely through the simulation of the tire, which is significantly overemphasized in racing games compared to real life.

Then of course there is the more obvious connection of engine tone and character that describes the car and its power and its behavior.

Lastly, there are the sonic features that every game faces, namely environment, user interface, and non-player characters, which are dealt with perhaps much differently than in non-racing genres though the purpose is quite the same.

And So, We Study
Damian and I decided that the best way to illustrate the differences among the games of our study in the defining sonic areas was to record little videos from each racing game to present in conjunction with our findings. This way, we can talk objectively about where we’ve been, look at where we are, and maybe even extrapolate where the genre is heading acoustically and technically.

Our approach broke down racing game sound into a handful of crucial areas, the first being unique to racing games: tire sounds, surface sounds, and engine sounds. Then we focused on things that are not specific to racing games, but are important to the game experience: camera perspective sound, environment sound, non-player character (or NPC) sounds, and finally the user interface (or UI) sounds. We broke all of these into individual videos showcasing samples from every game we studied, and have made these videos available to you below. I have wrote my own thoughts on the video here, from the perspective of a player or driver. Damian wrote a sister article to this study from the perspective of the sound implementation specialist, discussing the “how” of what you hear.

Rubber, Meet Road.
Tire sounds, as I mentioned earlier, truly are the crux of racing game sound as they fill the “void” of lack of physical communication between car and driver. The actual sound of a squeal is generated by particles of the rubber vibrating against the road surface as the force of friction attempted to be applied by the driver’s inputs starts to equal the maximum friction the rubber-to-road combination can afford. The true idea of driving is balancing the weight of the car across the four tires, such that the grip of each tire is used as efficiently as possible, thus going as fast as possible. For more information about how that works, as well as a resource of terminology, the Minnesota Autosports Club has prepared an excellent instructional flash animation (go to the “driving” tab and read at least through page 4).
The sound generated by a tire will change based on a number of conditions, where the most significant is how close to that grip limit the tire is at. Thinking back to your basic physics class, a tire works as a function of static friction. much like a rubber puck on a table, when pushed it takes a certain amount of force to break that static friction and move to kinetic, or sliding friction. The key to a fast driver is balancing as close to the line between static and kinetic friction as he or she dares. (I’ll be referring to this as the “grip limit” from here on out.) When a driver, say, turns too fast, the amount of lateral static friction is exceeded and the tire begins to slide. (This is also true longitudinally where the driver accelerates too hard and does a burnout or where the driver hits the brakes too hard and locks up the wheel.) For a significantly deeper discussion of the physics behind tires in a game, be sure to check out Brian Beckman’s presentation on tire simulation.
It is therefore absolutely crucial that the driver can distinguish exactly how close to that limit he or she is. In the real world, the driver can feel his or her body (and by extension the car itself) shifting in all directions and know from that what’s going on with the tires. In the absence of physical communication, sound fills the void in an auditory language the player can learn through practice. That language mimics sounds that real tires make, but in general are exaggerated in games. As the tire approaches the grip limit, the sound starts as a lower-frequency hum or growl. This will get louder and hissier the closer to the limit the tire is. When the tire exceeds the grip level, you hear the squeal. The fundamental frequency of that squeal is determined by the material of the tire, the material of the surface, and also by how much downward pressure is on the tire, with greater pressure producing a lower pitch. Note that this really only applies to rubber tires against a paved or otherwise hard surface. When driving in loose materials such as snow or dirt, no squeal is heard, because the static friction between the tire and the particles remains in tact but the particles themselves are sliding against each other.
To demonstrate this behavior, I devised a test where I turned the wheel of the car in the game and slowly accelerated such that the transition from nothing to some exerted force to approaching and then to exceeding the grip limit can be heard.

For our examined games, this means the connection between the physics engine determining the handling of the car is going to be very closely tied to the audio engine. Listening to our samples, we hear in the very earliest of games such as Dirt 1 that there is only one squeal state, as a sort of “on/off” type sound. Either the concept of the humming wasn’t fully realized or couldn’t be fully implemented at the time. This tells the driver when they have gone too far, but does not warn the driver when they are approaching too far. It is still very possible to drive successfully on this limited language set but lacks the nuance of the full language. As technology and technique improved, there became some amount of volume automation, suggesting an approach and then passing of the grip limit. In PGR3, for example, The sound sample I hear being played approaching the grip limit is the same sample as exceeding it, but the volume approaching the limit is much quieter. The next evolution came as changes in sample for different grip levels, tire choices, etc. such as in Forza 2 where there is a distinct hum before the grip limit and squeal after. F1 2010 had the most distinct humming state which started even at a very low speed, but that state did not change sonically for a very wide range of speed so it is tough to know exactly what it is communicating. The most recent development has been 4-wheel independent sound models in Forza 3, GranTurismo 5, or Shift 2: Unleashed, where you can hear that the front left tire is squealing more than the rear right for example. When listed to on a surround playback system, the appropriate wheel’s sound is played primarily through the corresponding surround location.
All of these improvements come in the name of improving the game-to-driver communication and do so successfully as time has gone on.

You’re A Bit Off-Line.
While the sound of a hum or squeal might indicate that the tire is already at the limit on pavement, what about on surfaces that are not hard-paved? Driving in loose surfaces is significantly different than pavement as your tire begins to act more as a “paddle” and less as a tire. (I like to envision the huge wheels on the side of a steamboat for reference.) Games that feature rally (ie the Gran Turismo series, or the DiRT series) tailor quite a bit of their sound model to this sector. The sound of the surface has to communicate things like how deep the loose material is, what sort of material it’s made of, and what size the particles are. Back to the steamboat analogy, the paddles can apply more force in a more dense material (like molasses) as the particles are closer together and have more static friction between each other. It is important in a rally game for these nuances to be communicated to the driver. However, even “pavement-only” games like Forza sill feature surface changes in the off-track areas, such as grass or gravel traps. The truly important bit for the driving experience here is to let the driver know he or she is definitely not on pavement and should expect vastly different handling. The particular sound profile for each surface should make the type of surface obvious — for example, when riding through grass you might hear the sound of the grass blades brushing against the bottom side of the car, or in gravel you should hear the little stones being spit up from the tire and pinging along the bottom side of the car and the insides of the fenderwells.

In some earlier games, the surface would be an on-or-off state that would cover the whole existence of the car, for example even if the left side is on pavement and the right is on grass you only hear the grass. The other significant improvement is employing either volume-based or sample-based changes related to speed of the car, or more specifically to the wheels. For example, in Dirt 2 it doesn’t seem to matter what speed you’re going, the sample sounds the same. Forza 2, by comparison, has a great change in sound as I spun the car with the throttle wide open, where it sounds like there are extra samples being triggered for a thicker sound compared to when letting off the gas. There is also a lot of variation in the aesthetic of the sounds used. To me, Dirt 3 has the best gravel and dirt sounds, including some really impressive samples of rocks pinging against the fenders (listen as I go over the jump) as well as a great grass-sliding-against-the-door sound going through there. By comparison, the grass in Forza 4 does not sound as convincing to my ear. However, all of the titles differentiate sound across different surfaces in a way that lets the driver know what surface he or she is on and how to expect the car to behave because of it. One thing I am curious about is from NFS Shift 2U, where when coming out of the dirt I think is supposed to be the sound of debris coming off of the car but to me sounds like a horse hoof fall. It doesn’t fit in where I think it should and the sound effect doesn’t give me a good clue.

The sound of the engine is one of the most enjoyable aspects of a “cool” car. Everyone loves to hear the unique rumble of a large V8, or the high-strung whine of a fine Italian V10. The sound is what gets blood pumping and adrenaline up. But in a racing environment, the sound is a bit less emotional and more functional. Obviously, the fundamental frequency is based on the engine’s rotational speed, or RPM. A driver will often memorize the pitch of the engine when it’s time to shift, and will wait for the sound he or she hears to match with that in his or her memory, and use that as a shift indicator (since rarely will the driver be looking at the instrument panel). After a number of hours in the same car, a driver will also know much of the harmonic content of the engine’s sound when it’s running at its peak. The driver can use this to tell if the sound changes as a possible indicator of a malfunction. For example, an oil leak will make things like valve tapping much more clicky. A cylinder that is no longer firing will interrupt the fundamental frequency. A good driver can use these sonic clues to help diagnose a failure even before the car returns to the pits.
While racing games’ engine damage states are usually exaggerated significantly, that sonic aspect is still there. In games, though, what players really want is to get the visceral experience of strapping themselves into a powerful car and hearing the rush of power as they press the gas. Because of this, we can listen to the character of the sound and hear what sort of effect the game might be employing to communicate that experience. Very early games used a combination of synthesized sounds, usually employing sawtooth or square waves, to up the excitement factor. As games moved to sample-based sounds, compression was a popular way to push volume and excitement. With the advent of Forza 4, we are now seeing distortion effects such as Izotope brought to the realtime environment, to enforce the idea of huge amounts of power. When doing this study, we tried to find one car that was synonymous across all titles and use that car where possible. The Subaru Imprezza STi with its distinct turbocharged “boxer” 180° 4-cylinder was the one that most titles offered, and we chose it where possible. This way, we can compare sonic characteristics across titles more objectively without having to consider the car in use. For this test, I found the most open straight line I could find such that I could punch the throttle and shift through a couple gears. I rev the engine a few times in neutral before going because there should be a change in harmonic structure between a relaxed, or “unloaded,” engine in neutral versus the hard-working “loaded” engine running through the gears, notably the lowest harmonics should become more distinct and louder. (You can experience this in your own car on the street, too.) In braking, there is often a reduction of RPM that happens significantly more quickly than the increase of RPM in acceleration, which on older games than the scope of this test would mean the game could not process the pitch shifting and sample-changing quick enough. This would result in a “stepped” or “staircased” sound where the listener could distinctly hear the sound jumping from one sample to the other. What we’re listening for in the braking is the smoothness of the playback. In theory, this should be more relevant in cars with wider RPM ranges, and presents a significant challenge for the F1 productions in particular as those engines with an RPM range of 0-18,000 is 2-3 times wider than a normal car.

In the samples above, Dirt 2, Dirt 3, NFS Shift 2U, PGR4, GT5, Forza 3, and Forza 4 are all samples of the Subaru STi and should all share similar sonic characteristics. In my opinion they all are reminiscent of the STi in that they share the correct rhythm and basic harmonic structure, but what I find interesting is the differences among the samples. For example, Forza 3/4 have the whistle of the turbo spooling up and the “pah” of the blow-off valve allowing pressure to release when the throttle is closed (ie to shift), and both sounds are distinct. By contrast, the turbo sounds in GT5 sound like an 808 hi-hat to me. Dirt 2 had an excellent turbo “chirp” randomly added in while I was revving in neutral. NFS Shift 2U had a fantastic blow-off valve sound on shifts and PGR4 had a distinctive turbo whistle. You can also hear that both F1 games handled the extreme RPM changes smoothly and without incident nor stepped sound. I hope you also felt a significant change in perceived power between the Subaru in Forza 4 and the Pagani Zonda R clip from Forza 4 that followed it. A lot of that emotional impact I attribute to the use of distortion on the main engine sound buss.

Roll Camera B
Racing games have almost always offered at least two common views: the “Nose cam” camera mounted on the nose of the car, such that no part of the car is showing, and the “3rd Person” or “Chase cam” view that follows behind the car a set distance, showing the whole car. Nearly all games also offer some sort of “cockpit” or “simulated driver” view that shows the interior view. Other popular views are the “hood cam” showing just the leading edge of the front bumper, the “roof cam” (or “TV Pod cam” in the F1 series), and sometimes a further-distanced 3rd Person cam.
With all these changes in view, we should be able to assume changes in sound as well. Take the car you currently own, for example. Start it up, let it idle, and set your head on the hood. Listen to the sound, and memorize it. Now get in the car, and sit in the driver’s seat. It should sound like a vastly different harmonic subset. Now get back out and stand behind the car. Again we hear a completely different tone. It’s all the same car, with all the same sound generators, but the distance relationships and phase changes as well as blockages change what we hear. Thus we should assume the same effect from our racing games.
Different games seem to choose how intense the perspective change will be based on production value. What we tested for was to see if there was a change between views, how different of a change it was, if it matched with another view’s sound, and how smoothly the change was from view to view. On games that allow the user to rotate the camera, we listened for how smooth changes were as you looked around the car, as well as stereo or surround positioning. Also, I’d like to note that it is very difficult to drive fast and change views at the same time!

It sounds to me like for most games, the “nose cam” and “hood cam” sound identical. In games with multiple interiors like NFS Shift 2U, both interiors sound the same. For games with multiple 3rd person views, both views sound the same. This basically leaves all the games with 3 “mixes” to spread across all the views. The games that transitioned from one view to the next like the F1 games were much more pleasant than the hard jumps when scrolling through, though this is a feature I assume is rarely used as most gamers will find the view they like and stick with it.

Enjoy The Forest For The Trees
If you were to strap yourself into a formula 1 car with a 136dB engine right behind your head, and drive from an echoey stadium and then out to open air, chances are you would never be able to hear the change in ambiance. However, on quieter cars or in a 3rd person perspective, it becomes more appropriate that a change in environment should have a similar change in sound. We looked specifically at things like tunnels or underpasses to listen to reverb tails and how the different states transition. Things like crowd noise when near stadiums can also add to the excitement factor of the driving, giving the feeling of the player being a superstar driver. I did notice though that there was sometimes a mismatch between the visual size of the crowd in the stands and the aural size of the crowd (with the aural usually being much bigger).

There is a significant amount of YouTube videos of people showing off their exhausts in tunnels, and Jalopnik even did a 10-best of them, so there is quite a bit of social significance to this. I really think Dirt 3 took these effects to heart and made it fun to drive around in tunnels. I realize that a number of titles are missing from this sample simply because I could not find tracks that demonstrate the effect in the limited playthough I had on them. I also think we should have given damaged engines more coverage.

And Everybody Else
A key aspect of competitive racing is knowing where the other cars are around you such that you know if you need to block a pass or if the car in front is having trouble. Often times in real life it can be difficult to hear another car over the sound of the one you’re strapped to, but like environmental sounds, in external camera positions it makes sense to hear more of it. What we wanted to listen to was if the NPC featured a noticeably simplified acoustic model, and if NPC cars retained the acoustic signatures of the cars they represent. (For example, if the NPC is driving a Ferrari, it should sound like a Ferrari, not just an ambiguous car.)

I think the PGR series really shined in this regard, with each NPC car retaining its own characteristic sound. Some of the earlier games like Dirt featured the same sound for each car, which tells me what I need to know but is not interesting. One thing of note is that the volume of NPC varied widely across titles and as you see I did have to adjust my car’s engine volume a bit just to make NPC distinct for this video.

User interfaces in driving games are mainly heavily menu-driven and thus are inherently use the sound as confirmation of selections being made. However, things like the music and the tone of the beeps and boops can set the player into a particular mood, which will carry over into the driving performance.

I feel a very drastic difference, for example, between Forza Motorsport 3’s bright white, clean, lounge-y music and “modern” beeps that gave a emotionally-sterile feeling and NFS Shift 2U’s dark, brooding music and heavy beeps, which left me in an “underdog” mental state, or that I had to prove my worth. This is completely different to the excited, loud, or brash feelings of Dirt 1 and also the PGR series, where hard pounding music and even trumpet stabs signal interaction. Codemasters had a time where they tried a more “first-person” menu style with Dirt 2 and F1 2010 where now the focus is on first-person movement sounds on top of the usual effects, which does help make the career modes a bit more immersing. However, the first-person style was dropped both for Dirt 3 and F1 2011 so perhaps they felt it was unnecessarily complicated. Me personally, I prefer the menu to be as simple as possible because I’d rather be racing than navigating it!

We Hope You Enjoyed
Damian and I wish to thank all of the games’ production staffs for all their hard work to create the games we’ve enjoyed listening into. It is truly fascinating to consider the difficulties that have been overcome in regards to locating cars to record samples from, organizing and cleaning those samples, collecting them into a useable asset set, implementing them into the game engine, testing, releasing, supporting, and then getting ready to redo the whole thing with the next title release. As new techniques emerge and new consoles arrive, the sky is the limit in terms of what games can achieve and it’s very exciting to be in the midst of it. I hope the study has shed light on where our racing games are now and perhaps inspired thoughts on where they’re going.

What Happens Next?
For me, the next big thing will be taking everything I’ve learned through this study and apply it to my own recording and implementing project currently in pre-production via Kickstarter. If you liked this study and want to see or hear more, give it a look! I am also considering doing another study for mobile racing games ie. Real Racing and Asphalt, or possibly for PC sim games like rFactor and iRacing and Simraceway.

More Info?
Be sure to read Damian’s much more technical analysis over on his blog. You’ll also definitely want to check out Game Audio Podcast #18, featuring myself, Mike Caviezel, Mike DeBelle, and Tim Bartlett. As always if you have thoughts please leave a comment below!

Game Credits:
Forza Motorsport 4, Forza Motorsport 3, Forza Motorsport 2, Gran Turismo 5, Project Gotham Racing 4, Project Gotham Racing 3, Need For Speed: Shift 2 Unleashed, Need For Speed: Shift, DiRT 3, DiRT 2, DiRT 1, F1 2011, F1 2010. Please see any of the vimeo videos for full game audio credits.