Computer generated image of flag

Car Performance Experiments

This page contains the results of some experimentation I did concerning car performance. The various aspects of my testing environment and procedures are presented here.
The results have been moved to separate pages, as this page was getting a bit "overweight". The results pages have also been reformatted for easier viewing and printing.

The Track
The track is a homemade one which I would describe as adequate. It's only 24 feet long, because there are only two spots in my house which will hold a 32 foot track, and neither of them is near my computer.
It is a normal shaped track. Maximum slope is at the starting gate, decreasing smoothly for about 15 feet to flat. Height of the starting gate is about 3 feet.
The characteristics of this track (like those of any track) will tend to favor certain car configurations. For example, because of my track's relatively short flat section, a car which holds its speed well on the flat might be at a disadvantage against a car which can get to the bottom of the hill very quickly.
Any conclusions you draw from the data below should be tempered somewhat by this awareness.
The starting gate of the track is "rubber band" powered. The dowel pulls down in about 0.09 seconds. I performed an analysis which shows that the dowel will be completely out of the way before the car is moving quickly enough to push against it. In other words, all starts are "clean".
The Finish Line
This is also homemade, using the circuits from the Dave Regan site and the Pete Dzanbozoff site as guides. Thank you, gentlemen!
It connects to the parallel port of a PC. This allows software to detect cars crossing the finish line by monitoring the port pins in the usual way by reading the appropriate I/O register. (For the programatically curious, examples of how to do this can be found at the two sites above.)
The software generates output in HTML format, so that it can be cut and pasted directly into a web page. The C language timing code I used is due to Stanley Sutton. Again, thanks go to Dave Regan, at whose site this code can be found.

While my finish line is not rugged enough for the excitement and commotion of a real Pinewood Derby, it is perfect for this sort of testing. As you can see, it's a thing of beauty.

The Car
My boys and I are good (but not great) car builders. Our finishes over the years range from 2nd place in the Den to 3rd place in the District.
I constructed the baseline car according to the following specifications.

Weight - 5 ounces
Length - 7 inches
Wheelbase - 6 inches
Center of Gravity - middle of car
Placement of Weight - middle of car
Wheels - sanded and polished
Axles - filed, sanded and polished
Alignment - very good
Number of Wheels Touching - 4
Aerodynamics - very good
Wheel Hub to Car Body Clearance - 1/16 inches
Hob-E-Lube (with Molybdenum) graphite lubricant
Although many Packs don't allow the extended wheelbase, I included this in the baseline car because I moved the weight around for some of the tests, and I didn't want a lot of weight outside the wheelbase.
Many experts agree that 3 wheels touching is best. I chose 4 wheels touching for the baseline car so that the wheels would lie flat no matter where the weight was.

In testing this car against other "fast cars" we've built, this one can hold its own when optimally configured. Here is a picture of it after a tough day at the office. (Well, how do you think you would look if someone squirted graphite on you every ten minutes?)

Testing Strategy
Testing took place by "turning one of the dials", then recording the results. For example, the first test compared an unlubricated car with a lubricated car.
This, hopefully, allows us to evaluate each aspect of car performance individually.
Testing Tactics
Each test consisted of at least two sets of runs.
The first sets were test sets in which one of the performance aspects of the baseline car was altered.
The final set was a control set to recalibrate the average time for the baseline car. As the testing took place over several days, the track was taken down and reassembled several times, which obviously might have caused significant changes in times.
Before each set (except the no-lube set) the car was lubricated and "warmed up".
Each set consisted of eight runs. The two fastest and two slowest runs were thrown out, and the remaining four times were averaged. The hope is that this produced a typical time for the car.
Since my track has imperfections, there were occasionally runs which which were visibly awful. These were disregarded and did not contribute toward the runs in a set.
The test car was carefully staged in all cases, except for the "plop down" test. The car was centered over the center guide rail, with center axis parallel to the center guide rail, and with wheels as wide as possible.
Mental Calibration
Although this is second nature to the PWD experts and others out there, it isn't obvious to many people until they think about it for a second:
I measured the speed of the baseline car as it crossed the finish line at about 11.5 feet per second.
Doing the arithmetic, we find that:

1/10 of a second = 14 inches approx.
1/100 of a second = 1.4 inches
1/1000 of a second = 0.14 inches

Note that 5/100 of a second is about a car length! Keep this in mind when viewing the results below, and drawing your conclusions.
Test Runs
The test results can be found on the following pages. Each test shows one or more test sets in blue, then a control set in yellow. The numbers not in bold are the fringe values that were discarded. The averages are the large, bold numbers to the right. Beneath the averages, their difference is displayed, with parentheses if the test average was slower, without them if it was not.
All times were recorded to six decimal places (1 microsecond). Additionally, all six decimal places were used in the calculations of the averages. The heat times which appear below are rounded to three decimal places for display purposes only.
Note that on any given day of the testing, the averages for the control sets are usually within 1/100 second of each other. The exceptions are days when the track was adjusted several times during the day, usually because of intervention by small children.
Note also that the control set averages can vary greatly from day to day, by more than 1/10 of a second, because of incidental differences in track setup. This is why we run control sets with the test sets.
For the first test, the average time for the unlubed car is over 0.377 seconds slower than the average time for the baseline car. This equals over 5-1/2 feet of distance at the finish line!
Given the accuracy of the test environment, I'll go out on a limb and conclude that a lubricated car has a considerable advantage over an unlubricated car.
That having been said, I'll let you draw your own conclusions for the rest of the results, which can be found on subsequent pages.
Interpreting the Results
This section contains editorial comments. I hadn't planned on having it, because I wanted to keep this page mostly free of opinions. However, now I feel the need to include it.

Some results may not apply to your race environment.
For example, in Test 2, back-weighting the car improves performance by over 0.03 seconds (4-1/2 inches), which is a very significant difference. However, this test was run on my track, and I'll bet you dollars to donuts that your track is not exactly the same.
Some tracks have constant slope. Some tracks are S-shaped. Some tracks have a constant slope drop, then a short transition, then the flat. On any of these tracks, and perhaps others, be warned that back-weighting may be the wrong choice. The stability gained from center-weighting might give you a faster car.
Don't let the close calls dictate your actions.
In Test 11, raw axles are about 0.003 faster than deburred and polished axles. What does this mean?
Does it mean you shouldn't treat your axles? No, I think not. I've seen bad burrs actually cause a wheel to bind on the axle and not spin at all. I wouldn't want to risk that.
Further, I could easily run two consecutive control sets and get a difference of greater than 0.003. The difference in Test 11 could just as easily be attributed to the insensitivity of the testing environment as to the difference in the cars.
What is most likely to be true is that axle treatment is not as important as most people originally thought, myself included. We're still going to do it in my house, but we probably will spend more time on things like wheels and alignment, as they appear to be much more important.

The "List"
Here are additional tests which I hope to eventually perform. Many of them involve small weight changes to areas of the car, which must be balanced by equal and opposite changes to other areas of the car. These will probably wait until I have a better scale.

White teflon lube
Lighter wheels, same diameter
Lighter wheels, smaller diameter
Hub caps to create lube resevoir
Smaller diameter axles
Larger diameter axles
Coned wheel hubs
LBW shimmed alignment
High car/dowel contact position and "slow" starting gate
Lube attrition
Probably some others that are stored away in emails, but which I don't recall off the top of my head.

Anecdotal Comments

I didn't log any data on this, but I'm convinced that a "cold" car will run faster in its 2nd race than in its 1st race, usually about 0.01-0.02 seconds. To make sure I could get consistent times, I usually ran at least two warm-ups before I started logging the heats in a set.
It seems likely that this observation is related to the results of Test 27. My testing procedure is closer to what was done in Test 28, since I am always dealing with a "warm" car.
Could it be that the graphite lubricates better when it's warmer? Does the granularity change because of use, or lack of use? Perhaps someone with a chemistry background could enlighten us. In the mean time, spin those wheels before your first race, folks!
Keeping your track in a constant position is a major concern when doing testing of this type. A seemingly minor shift in track position can easily affect times by 2 or 3 hundredths of a second. I had to restart several tests because the times "just went all to heck" right in the middle of a set!
Re the "canted axles" test (which might be too close to call, anyway) I feel the need to advise as to the alignment precision of the baseline car.
The axle holes were sufficiently deep so that they could meet in the middle of the car body. It turns out that I was able to push a "round" toothpick (diameter just slightly smaller than the axle diameter) all the way through the car body. I could do this at both ends of the car.
In other words, the alignment was about as good as it could be, short of using advanced techniques such as LBW shimming .
If you are unable to be this precise (for example, if your rules do not allow drilling holes or drilling out the grooves) then "canted alignment" is certainly worth considering.
If you have the opportunity to race your car before Race Day, I would suggest that you "rotate your tires" and observe the results. If your car is "hugging the center guide", a wheel rotation may be just what the doctor ordered.
Go to Jay Jenkins' famous CARS page and look at his comment about the Orange Crate (the first car he ever helped build.)
Perhaps the same could be said for these experiments. I did put a lot of thought into this ahead of time, but having been through the experience, I see some obvious areas of improvement. Here are two of them:

The track. It should be longer (I think I can fit 28 feet into my testing area) and have a longer flat. I also want to improve the quality of the track to the point where it's tip-top, like Pack 240's track, for example.
The timing system. Quality is not the shortcoming here, quantity is! I want to have multiple timing points - at least two (one at the bottom of the hill and one at the finish line.) I think this would let folks better utilize the data to form conclusions about what will work best for their track's shape.

Acknowledgments
A number of people have provided suggestions for possible tests, as well as helpful comments which have helped improved the quality of this project. Among them are Randy Worcester, Dave Regan, Michael Lastufka, Greg Metzger, Darin McGrew, and Stan Pope. To these people, I extend my appreciation.

Closing Remarks
Although this is more than I've seen published anywhere else, it is only a limited amount of test data from a single racing environment. It would be nice to have more data from multiple racing environments in order provide a larger and better pool of knowledge.
Until we have such a knowledge base, I offer the following disclaimer concerning the data above: Your mileage may vary.

Last updated on December 13, 2006, 12:00 PM
Copyright 1998-2006 © by Cory Young. All rights reserved.