Large - Small
(right click, 'Save As')

A video summary of the autocross and data analysis presents most of the material on this page in animated form.   Two versions are available for download: 640x480 28MB and 320x240 5.5MB.
Right click and 'Save As'.  For online viewing, try Google Video.

• Overhead path animations
• On-board Side/front-wheel camera
• On-board, forward-looking camera
• External, fixed-position camera
• Performance envelope description
• Animated acceleration path
• Potential-coded track path
• Understeer/oversteer comparison

Thanks to Shadi Krecht and Shahid Mahmood for help with photography and videography.

Figure 1:  An oblique view of the Gainesville Raceway parking lot (right) and drag strip (left) taken from a Cessna 152 at 600' AGL.

This Gainesville event was dissimilar to many autocrosses I had previously attended due to the freeform design of the course layout.  Jim Reyenga designed the course on an open section of the drag strip parking lot, affording significant freedom in turn sequences.  In comparison, Brooksville autocrosses are generally restricted to an out-and-back style where much of the course is a slalom due to the narrow runway width.  Other Gainesville events make use of the test track to link sections of the circular skidpads, variable-radius turn, chicane, and 90-degree corners with the occasional slalom for good measure.  Plots below of the ground-track through the course show just how much the course deviates from the typical restrictions.

For myself, the most notable challenge was the broad turns that afforded several line options.  I found that my interpretation of the best racing line changed between my walk-through and my first run and continued to change dramatically for each subsequent run.

The overall times seemed higher-than-expected to me, especially with regards to some of the typically-fast racers who posted somewhat-lethargic times.  Perhaps this was due to the freedom of choice given in the turns and the difficulty (at least for me) in settling on a comfortable racing line.  Many people complained of the relatively low-traction surface and difficulty in maintaining grip in both turns and straight-line accelerations.

Figure 2: Normalized race times for five runs of Project 240 vehicle of science (Nye, 1994), #240, Nissan 240SX.

My personal struggle with understanding the course layout is perhaps best emphasized by the plot to the left, which shows steady improvements in the lap times from the first two the fourth run with a total difference of nearly 5 seconds.  The run times are shown normalized by the fastest-time-of-the-day (FTD) for the touring class and for all classes.  My typical performance relative to FTD has been in the range of 1.15 to 1.20, meaning my times are 15-25% longer than the fast guys. 

The FTD was carried by Gary Mease in SM2 with a Miata.  Jeremy Warwin earned the fastest time for the Street Touring category, although the difference in between the blue and red lines in the plot represents a 3 second difference, or almost 10%.

It is interesting to note the improvements between each laps were remarkably steady despite large variations in driving style and 'incidents'

Run 1: 6-point harness sub-straps accidentally engage seat-position lever on acceleration.  Seat slides to aft stop, causing hands/feet to be far from steering wheel/pedals and harness becomes useless, slack mess of straps.

Run 2: Newly-installed Tein springs change chassis response, accidental drifting (beta>30deg) of two turns. Improper line on final hairpin and large understeer into slalom.

Run 3: Improved line, still losing time (wheelspin) on both hairpin turn-entries.

Run 4: Best time, used late-apex line through both hairpins with good results.  Downshifted to first for slow turnaround section.

Run 5: Locked front brake in preparation for first hairpin, subsequent understeer undermined racing line.

Figure 3:  Combined acceleration potential diagram showing three iso-acceleration lines and magnitude-dependent colorbars

A true, model-based estimate of the acceleration potential requires a very high-fidelity model of the tire response, engine output, and knowledge of the chassis compliance.  In lieu of such a model, a simple, elliptical estimate of the acceleration potential can be quickly generated from the orthogonal acclerometer measurements.  Figure 3 shows a crude envelope of the acceleration levels from Run 4, the fastest of the five. 

Several assumptions are made in the performance potential estimates:

• Maximum acceleration in turns, driving, and braking will be encountered at least once
• Lateral acceleration limit is symmetric
• Forward acceleration limit is measured in second gear (roughly 80% of first gear accel)
• Longitudinal acceleration limit is asymmetric, with braking greater than driving
• Combined acceleration limit is semi-ellipse with lateral acceleration on mutual horizontal axis, and driving or braking acceleration on vertical axes.

The green outer boundary represents an estimate of the 100% performance limit, where the vehicle system is operating at the maximum potential.  The inner boundaries, yellow and red, represent the 66% and 33% scalings of the outer boundary.  Colorbars are shown in the braking, right, and left-driving directions.  For those interested, the color map is based on the 30% of the hue range, starting from 0-degrees hue (red) to 108 degrees hue (green).

Figure 4: Ground track of Run 1

Figure 5: Ground track of Run 2

Figure 6: Ground track of 4th run color-coded to represent magnitude of achieved acceleration potential

Figure 4 shows the ground path of my first, and worst, run. The car position at each 1/100th second is plotted as color-coded dot using the performance potential estimate described in the previous section.  The combination of the performance estimate with the ground track provides data that, in my opinion, is more useful than the traditional green/red track plot.  In particular, the type of plot shows intuitive feedback on the quality of the driving, where an entirely green lap would represent achieved potential while a patched red/yellow/green lap shows lost potential.  It is important to note that since the color scaling is based strictly on a normalized magnitude, direction information is lost - making a maximum braking and a maximum driving section appear as the same color.

The seat-sliding incident of the first run severely hampered my ability to properly race the car (although not from a safety standpoint).  The incident is visible in the plot as the first large red area after the start.  As the seat reclined, my foot momentarily came off the accelerator and the car coasted briefly before I entered the right turn into the slalom.  Similar red regions show up elsewhere on the track in transition regions, mainly because my poor seating position did not allow me to move quickly between the different controls (throttle/brake/steering-direction). 

It is also interesting to note the path curvature in the upper-right corner of the plot near the major turn around.  Compared to the other two runs, the path is unnecessarily winding due to my inadvertent late-braking, overshoot, and slow recovery.  A limitation of the color-coding is seen here, where the path remains green despite an obviously-poor racing line.  Thus, the color shows no indication of whether the acceleration is made in the appropriate direction.

Figure 5 shows the ground path and acceleration potential of the second run, which is the slowest of the last four runs.  A cursory look at the plot suggests that red-coloring covers a broader area of the track than for run 1, despite the improved time. The discrepancy here comes from the boundary scaling, which is performed on a per-run basis.  In other words, if I achieved a much higher acceleration in run 2, then the colors from run 1 would not represent similar absolute accelerations.

Comparing the path of run 2 to run 1, the first slalom section appears to be more curved while the second slalom appears to be less curved.  The former is a result of unintentional drifting that pushed the car far off the racing line.  The latter difference comes from a late-apex, which both extended the straight and simplified the subsequent slalom.  Marginal acceleration levels on the last straight is a result of a slow-speed, second-gear acceleration outside the power-band.

It is only with the fourth run (Figure 6) that the disparate aspects of the track strategy started to become apparent.  In particular, comparing the first (top left) and second (top right) slaloms with the previous two plots shows a much straighter path.  Additionally, the course is overall much more green, indicating that my driving was more consistent at the higher acceleration levels.

Several spots of red are visible in the track data, most are apparent in straight sections just before or after turns, in transition between throttle and brake pedals.  Another low acceleration section is visible immediately after the 180-degree turn in the plot center.  In this case, I had downshifted to first before the turn, and shifted up just before the straight preceding the back sweeper.  The red areas represent the coasting period during the shift while the prior green area and later yellow show how different the acceleration levels are between gears.

Figure 7: Combined accelerations of Run 1 showing significant portions near and below 66% potential iso-acceleration line.

Figure 8: Combined accelerations of Run 4 showing improved acceleration potential (mostly above 66% line)

Acceleration plots of runs 1 and 4 are shown in Figures 7 and 8, respectively.  The three potential boundaries are left intact to help in interpreting the plot.  Figure 7 shows that much of the longitudinal acceleration in run 1 was within the 66% boundary, especially for combined accelerations.  The lateral accelerations are near the limits, but note the slight difference between the values of the two runs.  Combined right turns and braking are performed well, although the track design didn't allow any significant left turn and braking combination.

Figure 8 shows an improved acceleration distribution, particularly in the forward direction.  The performance boundaries are wider than run 1 and yet much of the acceleration takes place above the 66% threshold.  The larger driving accelerations are likely attributed to downshifting to first gear for the short straights after the turns.  Interestingly, the lateral limits appear to be less developed than run 1, particularly for right turns.  The discrepancy may be caused in part by large left accelerations widening the symmetric lateral boundary.  The maximum right lateral acceleration is actually found in a combined turning and braking section.

The Gainesville autocross was my first event in a new class (STS) and with new springs on the 240.  I am still in the engagement period with the new setup, but so far I am delighted with the chassis response.  The data presented here is also with an entirely new computational/logging setup, so I am likewise happy with the measured results.

Color-coding the track data with a performance potential estimate rather than a fore/aft acceleration provides a more useful plot.  The colors here provide visual feedback on both the quality and consistency of the run.  This feature is especially important for analysis during the race, considering that only a few minutes are available to study the data and determine a new approach for the next run.  In post-processing, the colored-ground-track provides a sound basis for qualitative comparisons between different runs.

The acceleration limit boundaries are quite useful for comparing two or more acceleration-space plots.  However, since the combined limit is based only on an elliptical fit, many of the realistic limitations of the car are ignored.  A better approach would be to estimate the parameters of a Pacejka combined-slip Magic Formula model from the measured data.  This, however, can only be done with a sufficient amount and diversity of input data, both of which may be lacking in an autocross course.

The video summary used animated segments to help illustrate the different racing conditions.  I am pleased with the overall results, although the poor on-board video quality detracts from the aesthetics.  It is also my first time narrating, so it is always a struggle to understand how to effectively present material audibly and visually.

Future autocross analyses will benefit from improved video quality and a better understanding of the track layout.  In particular, the location of key cones should be measured to allow path-planning elements to be included.

As usual, I'd love to hear suggestions for improving data presentation on this and future autocross studies.  In particular, please advise on critiques of the video production or argument structure.