Our Experiments

Ball Speed:

If you have ever seen slow motion replays of tennis or other sports on television, you have probably noticed that when the replays freeze the action, the pictures are blurred. In tennis for example, the clear image of the ball and the racket at contact are impossible to see because they are simply moving too fast for the camera.

This is because regular video, including everything from broadcast television to home camcorders records images at 30 frames per second. When we watch replay of these 30 frames or 30 pictures a second we see motion.

The first problem for the project team was finding a way to freeze the action to study the motion of a tennis ball or tennis racket. To do this the team came up with two solutions.

The first solution was to use a camera that has something called "a high speed shutter." Many video cameras and even home camcorders have these shutters built in and they range in speed from 1/100 of a second all the way up to as high as 1/10,000 of a second.

What the shutter does is allow the camera to freeze motion. The camera is still recording 30 frames per second, but with the shutter each of those 30 frames or 30 pictures is taken at the shutter speed. For example, at a shutter speed of 1/1000 of a second, the camera takes 30 pictures of the subject per second, but each of these pictures has a duration of only 1/1000 of a second.

Our project manager and co-investigator, John Yandell, has had extensive experience using these kinds of high speed video cameras at his tennis school in San Francisco. Tennis teachers can use a camera with a shutter to show a particular stroke to a student, and also to show them the instructor performing the stroke correctly.

Yandell worked with co-investigator Nasif Iskander of San Francisco University High School, experimenting with 8mm camcorders that had built in high speed shutters. The found that with a shutter speed of 1/1000 of a second they could freeze the flight of the ball sufficiently to mark its position. This technique became the basis for recording the data on the speed of the ball as described below.

Ball Spin and Biomechanics:

In addition to the speed of the ball, the project team wanted to investigate how the ball spins in tennis and how fast. The also wanted to studying the stroke patterns or biomechanics of how the player struck the ball to make it "fly" in the first place.

The radar guns tell us something about the initial speed of the ball, but nothing was known about how the ball spins in pro tennis. As the background research showed, there was one pioneering study of ball spin using high speed film done by Dr. Jack Groppel, and also computer modeling of spin in the work of Howard Brody. But the various types of spin: topspin and underspin had not been recorded with the top players in the world.

Determining how to record spin was a more difficult problem. The camcorders with the high speed shutters the team chose to study the ball speed were not capable of seeing the ball spin. This was because the camcorders, even with the shutters running, still only took one picture, or frame, every 1/30 of a second. In a 1/30 of a second the ball moves up to several feet through the air in tennis. During this 1/30 of a second, the ball could rotate many times, particularly with the heavy spin hit by many top players. To really see how the ball was spinning the researchers needed to see multiple pictures of the ball during a single revolution.

To study the spin, the research team turned to a new kind of recording technology, high speed digital recording cameras.

The value of these cameras was that they could record much higher frame rates. Instead of recording 30 frames a second like conventional video cameras, these digital cameras could record 250 or 500 frames a second. Other digital cameras were faster still and could record 1,000 or 2,000 frames a second. In addition to the frame rates, these cameras also came equipped with high speed shutters, with speeds up to 1/10,000 of a second like the conventional camcorders. Conceivably, then, the team could take 2,000 pictures a second, each with a duration of 1/10,000 of a second!

These cameras were already in use in pro tennis during television coverage showing replays of line calls and whether certain balls were in or out on a given point. Co-investigator John Yandell suggested that they could also be used to study spin, and also to study the biomechanics of the strokes.

The question was which camera was best suited for this different application? To answer this the project team tested different high speed camera systems developed by Kodak.

To find out how the cameras would perform recording tournament matches, the team arranged on court tests with teaching pros and other top players. Principal Investigator Dr. Jani Macari Pallis, Project Engineer Eric Chattot, and Project Co-investigators John Yandell and Nasif Iskander all participated in various phases of the testing.

Because of the tremendous amount of data these digital cameras recorded and how they recorded it, the project team found there were limitations and advantages and disadvantages to the various options.

For example, the fastest camera tested, the HRC 1000, could record 1,000 full frames a second, or 2,000 half frames--a picture half the size of the full frame picture, with shutter speeds of 1/1000 and higher. In the tests it gave superior data for counting ball revolutions and calculating spin. The team was excited to count ball spin on all the shots for the first time. The also concluded that the higher digital frame rates paired with the shutters would yield more and better data on the strokes of the top players that had ever been recorded.

Two problems, however, made this fastest system an unacceptable choice for the project team.

First, this fastest system recording the images digitally on a processor that had a storage capacity of only a few seconds. To save any data required downloading this data from the processor to a tape system. This made it very difficult to try and record specific ball hits--the great shot at the end of a rally, for example. It was also impossible to record even a two ball exchange between the players. The downloading process was also time consuming.

The system simply could not store enough data to be useful in recording a large number of events during competitive matches. Furthermore the team judged that downloading from the processor to tape would present insuperable logistical difficulties at a pro tennis event. The equipment would require too much space and the interaction between the team members that would be required would be too disruptive in a hushed tennis stadium environment.

The second problem with the faster system was its light sensitivity. To run at 1,000 frames per second with a shutter of 1/1000 or faster required bright sunlight. This eliminated the possibility of filming night matches or on overcast days.

The team finally settled on high speed system that was slower in terms of frame rate, but one that could record directly to a special high speed Super VHS video recorder, the HSV 500C3. This was a new state of the art tape based system that had been introduced in 1997. It could record at 250 full frames per second or 500 half frames per second (with half size pictures). It had a high speed shutter of up to 1/10,000.

This meant it could record only half the data of the faster system. However, the on court tests proved that this frame rate was sufficient to see multiple pictures of the tennis ball during one revolution, and therefore to count the spin rates accurately. It would also yield ground-breaking biomechanical data on strokes. For example, at a frame rate of 250 frames per second, it recorded over 8 times more data than the conventional 30 frame video recorders. With the use of the high speed shutters, it could freeze the images within each frame so that they were very clear. The team believed that this would make a new understanding of stroke production possible.

The system chosen was significantly better than the high speed broadcast systems television networks sometimes used during sports events. These tape based systems could record only 90 frames per second, and were typically used with a shutter setting of 1/250 of a second.

The system had other features that were essential for the success of the project. First it recorded data directly to videotape. A 2-hour SVHS tape would yield 15 minutes of high speed video footage that could be viewed, studied and edited like any regular video. It was compact enough and simple enough to use unobtrusively in a tennis stadium environment. Third it had greatly increased light sensitivity. The team members were hopeful that this would allow them to record even night matches at fast frame rates with a high speed shutter. This hope proved to be well founded during the team's trip to the U.S. Open, the number world event in professional tennis.