The principle of the high speed capture experiments is simple:

1. Keep the object in (reasonable) darkness
2. Open the shutter
3. Trigger a flash synchronized with the event of interest
4. Close the shutter

The characteristics of the flash burst is important for the success of these experiments. Ideally, the flash burst should be extremely short and very well defined, i.e. fast rise- and fall times. The latecncy from the trigger pulse to the peak of the flash burst must also be known and predictable. By switching the flash gun to manual mode, at least on the 550EX, the output power can be set to 1/1 to 1/128 in 1/2^n steps.

One may wonder why not a modern camera with shutter times down to 1/8000 seconds could not be used instead, but two things kills this idea:

1. The shutter does not open fully and close fully at higher shutter speeds, which explains why the flash synchronization times generally are in the range of 1/125 to 1/250 seconds for SLRs. Instead, a "virtual" shorter shutter time is created by rather "painting" the sensor with a partially open shutter, just like a flatbed scanner does. The shutter opening can rather be seen as slit passing the sensor in about 1/200 second.
2. The latency from the trigger to the shutter opening is difficult to predict and is slow as several electromechanical systems needs to be engaged, i.e. mirror, shutter and aperture

I recently carried out a series of new experiments and realized it was time to properly characterize the flash burst of the Canon 550EX flash gun as well as the Elinchrom 400FX studio strobe. The test was carried out using a synchronized trigger and a digitizing oscilloscope, set to trigger on the flash trigger pulse. An ambient light sensor (Osram SFH3710) was used to characterize the shape of the flash burst. The actual "light-intensity-to-time" performance is not specified in the datasheet so there may be some non-linear behavior hidden there, but it should give a pretty good correlation.

In order to avoid saturating or "blinding" the light sensor, which would distort the shape measurement, the flash was masked more and more as the power was increased. Therefore, exact shape or the height (=actual power) of each pulse is not properly characterized with this setup, which are somewhat irrelevant parameters for the experiments carried out here.

The hypothesis that the burst pulse length is proportional to the flash output power proven to be true and at 1/128 power, the peak pulse is about 180 microseconds, which equals a hypothetical shutter time of 1/5000 s. At full power, the pulse is about 2 ms, equaling 1/500 s shutter time.

As was further assumed, the pulse characteristics of the far more powerful studio strobes are not useful for the high speed experiments.

The latency from trigger to the flash burst peak appears to be fairly constant at around 100 microseconds for all power settings.

Bottom line is that the 550EX is highly useful for high speed experiments up to 1/16 power or even higher depending on the experiment.
.