How We Test Cars and Trucks
Behind the Numbers With the Edmunds Test Team
Burnouts, powerslides, closed roads, racetracks — that's what people think the Edmunds.com test crew does all day, right?
Not always. Like most of you, we also drive our test cars mundane miles through suburban sprawl, inch forward on hopelessly jammed freeways and go on grocery runs. But behind every story, supporting and quantifying every editor's driving impressions, you'll find what we call "the numbers."
The numbers are utterly familiar. Anyone who has spent any amount of time reading car reviews will recognize them: acceleration time from zero to 60 mph; quarter-mile elapsed time and trap speed; braking distance from 60 mph to a dead stop; lateral acceleration around a skid pad; and speed through a slalom course.
Yet the process of developing data is a no-nonsense business. We do it as well as we can, and we don't skip over the hard parts, either. Getting trustworthy numbers requires slavish devotion to standardized procedures, accurate measurement equipment, controlled conditions, dedicated testing locations and skilled drivers with a knowledgeable and consistent approach.
We're all about the numbers.
Since our track tests require maneuvers at the limit of vehicle performance, our test cars go through a standard check-in procedure before they even turn a wheel. We inspect the cars for anything untoward that might compromise safety. Fluid levels are verified and, if necessary, topped off. The lug nuts on the wheels are set to the proper torque. Tire pressures are adjusted to the recommended cold inflation pressure found on the door placard and in the owner's manual. Even the track gets attention, as the courses are cleared of debris and the marking cones are set out.
There's also a weigh-in to determine the "as-tested" curb weight. The Society of Automotive Engineers (SAE) defines curb weight as the mass of a car without driver, passengers or cargo but with a full complement of fluids — including a full tank of fuel. So every vehicle we test has a full tank of the recommended fuel before it rolls onto our Longacre portable digital scales for a reading of total weight and weight distribution.
Could we get more impressive numbers with a near-empty tank? Certainly, but that would be a departure from accepted test practices and would dilute the real-world value of our results. After all, testing ain't racing. And once you start down that path, it becomes too easy to rationalize other questionable tactics to reduce weight like removing floor mats, emptying the glove compartment and ditching the jack and spare tire. We don't play the game that way. In fact, if a test vehicle comes to us with a cargo cover, cross rails for the roof rack or headphones for the rear DVD player, we leave it all in.
Our as-tested curb weights are nearly always higher than a manufacturer's published curb weight. That's because published curb weights typically represent a base model with few extras, while many of our test cars come equipped with popular options. As a result, our test weight and measured performance are closer to that of the typical examples of a vehicle you might see on a dealer's lot.
We derive our straight-line measurements from data recorded by Racelogic's VBOX III, an instrument-grade GPS-based data logger used by automotive engineers and even professional race teams because its 100-hertz sampling rate improves accuracy. This self-contained unit fits on the front passenger seat, and, in the words of Chief Road Test Editor Chris Walton, one of our two test drivers, "...talks to as many as 12 satellites in order to track a vehicle some 12,600 miles below." A head-up display for the VBOX is suction-cupped to the windshield to give instant feedback to either Walton or Josh Jacquot, senior road test editor and our other test driver. A compact flash-memory card holds our precious data until it can be processed.
B Is for Braking
Acceleration and braking tests are the first order of business once a vehicle is properly blessed. Quarter-mile and 0-60 results are the most talked about aspects of performance, so we always want to make sure they're in the bag in case the weather turns sour. But braking tests nevertheless precede acceleration runs on the day's schedule. We've found that slowing a really fast car after a series of quarter-mile speed runs can stress the brakes, so to ensure that we don't inadvertently corrupt the braking tests, we conduct them first.
Our policy is to test on a surface that complies with Federal Motor Vehicle Safety Standards (FMVSS) for brake certification testing, so the surface we use is smooth, flat asphalt. We never test on concrete and we don't use drag strip launch areas coated with sticky, exotic rubber. FMVSS-135 stipulates that brake test surfaces should have a peak friction coefficient (PFC) of 0.90. One of the two facilities we use is regularly certified to this PFC level, and we've proved to ourselves that the other facility produces stops of equivalent distance through empirical same-day tests with a control vehicle.
ABS is so prevalent that most stopping distance tests are a simple matter of accelerating to 70 mph or so, coasting until velocity drops to 65 mph and then stomping on the brake pedal to fully engage the ABS until everything comes to rest. Non-ABS cars require far more finesse to engage the brakes firmly and hold the car at the point of impending wheel lockup throughout the stop, a skill that our two designated test drivers have honed over years of experience.
Stops are performed about two minutes apart until we reach the point where the stopping distance begins to trend irrevocably longer. Some performance cars can make a dozen such stops with no degradation, while others show signs of stress after just three or four. To determine the point at which serious brake fade begins, we factor in the number of stops and the magnitude of any performance fall-off, and then take into account sensory cues like changing pedal feel and even the smell of the brakes. The stopping distance we publish comes from looking at the 60-0 and 30-0 slices of the data and selecting the shortest runs.
A Is for Acceleration
Acceleration tests follow as soon as the brakes cool. While we're waiting, let's review our fuel policy. Nominally, we use the minimum required fuel octane for our test runs, and if a manufacturer recommends a higher grade "for best performance," we'll use that. The only exception occurs when 93 octane is recommended, a grade of gasoline that isn't available here in California and many other states. Fortunately, cars that present this problem are few in number and all of them list 91 octane as the minimum requirement, a fuel we can readily obtain. If the runs come out a little slower than a manufacturer's claim, so be it. It's the manufacturer's fault if it optimizes an engine for a grade of fuel that isn't widely available.
All of the published acceleration data — the quarter-mile time and trap speed plus the 0-60 time and all of the intermediate speeds — come from the best single run. But getting that best run requires a skilled driver who can feel out the ideal launch technique that optimizes speed, yet doesn't damage the car. This is no easy task due to the wide variety of vehicle configurations, engine outputs, transmission types, gear ratios and tires in the marketplace. And then there are traction control systems, launch controls and a variety of sport settings to wade through.
Josh Jacquot explains: "When testing a car with this sort of adjustments, I'll get in, fire it up and make a pass, ignoring the electronic enhancements to establish a baseline number." After that, more aggressive settings are engaged to see if they represent any real improvement.
For vehicles with manual transmissions, Jacquot notes, "Some cars are quicker with moderate wheelspin; some are quicker with none at all. Some cars like a little clutch slip; some require instant clutch engagement and lots of drivetrain shock. Trial and error is, unfortunately, the only way to be certain which technique is best."
Vehicles equipped with automatic transmissions don't require as much imagination. Jacquot says, "If it yields a better test result, we'll brake-torque the vehicle by revving the engine at the start line while it's held in check by the brake. Often, a simple brake-off, throttle-on technique is best since the electronic stability control in many cars will no longer allow simultaneous brake/throttle application. Sport modes are always utilized after completing a default pass. Same goes for manual shifting — although this rarely helps."
The automated manual transmission is an emerging technology that requires a similar technique of brake torque, even though there's no torque converter and the transmission's clutch disc could be damaged. "The use of simultaneous brake/throttle application on this kind of transmission can, occasionally, result in quicker acceleration. It is a technique we use under the assumption that most of these cars have drivetrain protection that will keep this Neanderthal behavior from destroying the clutch pack. It's an unrealistic technique for real-world driving, but some cars are considerably slower without it," Jacquot said.
A Few Words About Rollout
The term "rollout" might not be familiar, but it comes from the drag strip. The arrangement of the timing beams for drag racing can be confusing, primarily because the 7-inch separation between the "pre-stage" and "stage" beams is not the source of rollout. The pre-stage beam, which has no effect on timing, is only there to help drivers creep up to the starting position. Rollout comes from the 1-foot separation (11.5 inches, actually) between the point where the leading edge of a front tire "rolls in" to the final staging beam — triggering the countdown to the green light that starts the race — and the point where the trailing edge of that tire "rolls out" of that same beam, the triggering event that starts the clock. A driver skilled at "shallow staging" can therefore get almost a free foot of untimed acceleration before the clock officially starts, effectively achieving a rolling-start velocity of 3-5 mph and shaving the 0.3 second it typically takes to cover that distance off his elapsed time (ET) in the process.
We believe the use of rollout for quarter-mile timed runs is appropriate, as this test is designed to represent an optimum drag strip run that a car owner can replicate at a drag strip. In the spirit of consistency, we also follow NHRA practice when calculating quarter-mile trap speed at the end of the run. So we publish the average speed over the final 66 feet of the quarter-mile run, even though our VBOX can tell us the instantaneous speed at the end of the 1,320-foot course, which is usually faster.
On the other hand, the use of rollout with 0-60 times is inappropriate in our view. For one, 0-60-mph acceleration is not a drag-racing convention. More important, it's called ZERO to 60 mph, not 3 or 4 mph to 60 mph, which is what you get when you apply rollout. While it is tempting to use rollout in order to make 0-60 acceleration look more impressive by 0.3 second, thereby hyping both the car's performance and the apparent skill of the test driver, we think it's cheating.
Nevertheless, some car magazines and some automobile manufacturers use rollout anyway — and fail to tell their customers. We've decided against this practice. We publish real 0-60 times instead. But in order to illuminate this issue and ensure we do justice to every car's real performance, we've begun publishing a clearly marked "with rollout" 0-60 time alongside the primary no-rollout 0-60 time so readers can see the effects of this bogus practice.
Correction factors are another source of controversy in vehicle testing. Because weather conditions vary from day to day, this affects an engine's horsepower output. As a consequence, acceleration times can be effectively compared only if the results are adjusted to a set of standard atmospheric conditions. The most widely recognized correction factors are those the SAE specifies within its horsepower measurement procedure.
SAE correction factors have undergone a revision or two in recent years, and it is our policy to use the one contained in the most recent horsepower measurement procedure, SAE J1349. Turbocharged engine performance is not corrected by this standard, because modern turbocharged engines with electronic controls essentially produce and optimize their own atmosphere.
The old standard, SAE J607, is now considered obsolete by the SAE, but the use of its correction factor produces quarter-mile times that are about 0.3 second quicker than those returned by J1349. Some publications still use J607, ostensibly because they don't want to lose the ability to make comparisons to their library of past data. (Sure, the 0.3-second advantage they get in quarter-mile times has nothing to do with it.)
If the outdated correction factor is combined with rollout, the results can be dramatic. The following example is based on data from a single run of our 2008 Mitsubishi Lancer GTS long-term test car. Here you can see the effects that the worst-case combination of correction factor and rollout can have on a 0-60 time.
1/4 mile (sec @ mph)
16.44 @ 83.85
16.17 @ 85.03
If you inappropriately apply rollout to 0-60 times and use the outmoded SAE J607 for weather correction, the 0-60 time appears to be 7.9 seconds. We use the more current SAE J1349 and do not use rollout for 0-60 runs, so we would report 8.6 seconds, a difference of some 0.7 second. On quarter-mile runs, where we do include rollout for reasons explained earlier, the difference comes down to correction factor alone, and in this example the difference would round out to 0.2 second and 1.1 mph.
Same car, same run, same raw data file, same ambient conditions, but different data processing — clearly, a lot of tricks can be played by massaging the raw data. And there's a strong temptation to corrupt the data in this way because acceleration times arouse such strong emotions among readers. Enthusiasts want their dream car to be super fast, so those publications that produce the lowest numbers are hailed as professionals, while anyone who gets a lesser number "doesn't know how to drive." We think it's more important to be as correct about performance as possible, so we're scrupulous about our data.
Meanwhile, the weather data we use for the correction calculations comes from a Novalynx WS-18 portable weather station we set up at the track. It records ambient temperature, wind speed and direction, barometric pressure and relative humidity at five-minute intervals throughout the day.
Our primary handling tests are conducted on a 6-by-100-foot slalom and a skid pad with a 100-foot radius. We don't use the VBOX for these tests, preferring instead to use a set of Brower wireless optical timing lights with a remote driver display. This allows us to simultaneously test handling with one driver while the other conducts straight-line tests with our VBOX.
For the slalom, seven cones stand 100-feet apart, making six 100-foot gates over a total course length of 600 feet. A driver will approach the course at ever-increasing speed and will keep making runs until he finds the limit at which he can negotiate all the gates. As with acceleration tests, different cars require different techniques to get the best time. Chris Walton explains: "Some vehicles, sports cars in particular, respond well to being driven aggressively. Others, like utility vehicles and most crossovers, prefer a gentler technique to minimize chassis upset."
Driving the slalom course really quickly and maintaining a rhythm isn't easy — it's a specialized skill unto itself. A highly proficient driver can rip a great time and sense a lot about the transitional behavior and steering response of a vehicle, and he can also sense a bad run and abort well before it builds to a spin. In our experience, stability control systems need to be turned off to get the best times and really feel what the chassis is capable of, though some vehicles don't provide a full-off mode. Walton notes, "In the end, a car not bound by electronic aids will show signs of pushy understeer or loosey-goosey oversteer when pushed to the limit."
To get the published speed, we divide the overall distance of the course by the time in seconds and convert to mph. Runs with cone strikes are eliminated and the fastest clean run is published.
Driving the skid pad to determine lateral grip is a more straightforward proposition. Our circular course uses a 100-foot radius, and the driver puts his inside tires as close as possible to the line that marks the circumference of the circle without drifting wide. The actual radius of the circle at the centerline of the car is what we use for our calculation, and that works out to 103 feet.
We measure and report the average lateral acceleration a car can sustain for a full 360-degree circuit of the course rather than the often-fleeting instantaneous lateral Gs. This makes it possible to use our timing beams again, using a single unit as both start and finish.
Few cars do well with hair-raising tail-out drifts, but we switch off stability control systems to the greatest extent possible anyway because most such systems intervene so early and often that they interfere with the ability of our drivers to feel the balance of the car.
Very few runs are required to find the limit, and that's a relief because tire degradation occurs quite rapidly if too many runs are made. To avoid the tire issue as much as possible, we alternate between clockwise and counter-clockwise and try not to make more than two circuits in each direction. The fastest clockwise and counterclockwise passes are averaged and thrown into an equation with the actual radius to calculate lateral Gs.
At the end of the day, we end up with a little less rubber, an exhausted crew, some more black marks on our poor cones and a pile of data. Back at the office, the data is given a final once-over and the reportable runs are selected. Using a time stamp, we match up our weather data with the selected (best) acceleration run, and apply the appropriate correction.
The final results are ultimately calculated and recorded along with the driver's comments and then handed over to the editor assigned to the overall story for a particular vehicle.
Whenever you read one of our stories, you'll find nuggets of information from our day at the test track. The results of the calculations made from the selected runs also show up in the Specifications and Performance page that accompanies each test article.
And of course it's all dressed up with descriptions of burnouts and powerslides, not to mention comments about just how well the iPod connection worked while stuck in traffic on the 405 freeway. But the numbers are there.
Chief Road Test Editor Chris Walton says:
It's more than a little ironic that the farther away the test equipment gets from the subject vehicles, the more precise the measurements become. Here's what I mean.
When magazines began testing cars in the 1950s, there was a second person riding inside the car beside the driver. He'd start a set of stopwatches with a master lever, then click them off one-by-one as the speedo would reach 60 mph, when the quarter-mile mark was crossed and when the speedo reached 100 mph. When it came time to measure braking distances, a .22-caliber bullet would blast a hole in a can of chalk to draw a line on the pavement to measure braking distances.
Jump ahead a decade or so, when cars still had substantial chrome bumpers, and a bicycle wheel or so-called "fifth wheel" was bolted to the rear bumper of the car to record the number of its rotations and thus time over distance. (I had a speedometer on my Schwinn Stingray that worked just like this.) Alignment, tire pressure and surface irregularities were all challenges. (And whatever you do, don't turn left — or was that right?)
Another decade passed and the radar gun made its debut. Sitting atop a tripod (or even worse, suction-cupped to the inside of the windshield), it bounced radar waves off the vehicle (or something in the distance) and, when analyzed by a computer, the data generated a bunch of neat graphs. But the system was also susceptible to pervasive interference and required error-inducing assumptions and lots of data smoothing. A couple of magazines still use this system to report acceleration and braking data. So does law enforcement, but even they are beginning to admit how inaccurate and unreliable radar is.
Finally, the U.S. government put a bunch of machines in orbit above the Earth, and our VBOX system talks to as many as 12 satellites to track a vehicle some 12,600 miles below. Our data is now more precise than it has ever been.
So test gear has gone from being inside the car, to being bolted to the outside of the car, to standing behind the car, to now sending microwaves from space to tell you how fast a car is. Man, how far we've come.