# Macro the 1/f rule and flash duration

Started Sep 8, 2008 | Discussions thread
re: several things to consider/clarify

This is one of the shots I'm looking at:

Note that there are two complete cycles in 50 microseconds. That works out to 25 microseconds per cycle, which is 40 KHz.

Then we've got this one:

I recon about 18 microseconds between peaks here. That gives me 55.5KHz.

But it looks as if the system takes a while to stabilize at a constant frequency, and this latter shot is triggered nearer to the start of the burst, before things "settle down". So I'd tend to put the actual frequency somewhere nearer to 40KHz.

the question is than how much (if any) this residual glow is able to
illuminate a scene highly enough to be "recognized" by a sensor?

I think the idea is that over 1/8000th of a second, the illumination is averaged by the photodiodes of the camera's sensor. At 40 KHz, we get five "cycles" to illuminate any given point on the sensor. For longer shutter speeds, we get even more. Thus, for the purposes of normal shooting, the modulated light from the flash simulates a continuous level of illumination.

But for your purposes, with a very fast-moving subject, we may see some variation in the brightness, leaving a "somewhat-strobed" look. But the darkest parts will have received 0.5 to 0.6 X the "peak" brightness. So we're talking about a difference of only one stop to distinguish the pulsations. I suspect it'll be hard to see.

I think we'd just see a blur, and not be able to easily see those 1-stop variations. But maybe they'd be more obvious than I think.

It would be much better if the strobing was complete, with the light output dropping to zero between pulses.

this is a valid concern, and [if we take the 1/8000 example] during it's
travel, a shutter slit will traverse (in time) figuratively speaking
either
3.25 or 6.5 micropulses, depending of what we agree upon is THE
frequency of the modulation; so indeed each pulse might be a light
source to a slice of cells on a sensor;

At 40KHz, we'll get 5 complete micropulses for every location on the sensor when shooting at 1/8000th. I think that would make it difficult to analyze what we'd see in the shots.

Any time you shoot very fast-moving objects with a focal plane shutter at high shutter speeds, you get bizarre distortions of the shapes of the objects.

I remember a thread a few years ago on here where we were discussing whether or not a baseball bat really did bend as far as it looked to be bending, or if we were just seeing that odd, curved representation of the bat due to the vertical travel of the slit over the focal plane capturing different bat positions at different places in the frame. And there was a shot of a helicopter with grossly distorted blades that was very amusing too

But you could use much slower shutter speeds, such as shooting at the X-synch, or just a bit faster (1/320th?) so that you'd only see the strobing effect, and not have the traveling slit to worry about. That might be the preferred technique.

some engineering of course would take place to have it as linear as
possible, but perhaps the actual traversing speed might be different at
the beginning and the end of each shot;

That's true. And that would add to the complexity of analyzing the shots

but then... how fast is FAST? we are not talking about a bullet hitting
an apple, are we?

I think it becomes visible in things like swinging baseball bats. So it's certainly a concern at bee-wing speeds

I know, the 550ex does the same (and also from 2 to 199 strobes),
which is better than nothing, but even at 200 Hz it is very useful to
e.g. determine the real angular velocity of that revolving fan in my
tests, but not really useful to capture strobe effects of IIFs;

That's my thinking too.

my only hope would be if there is some source to offer it commercially,
but then it might be rather frightfully expensive, although all
componetnt
parts are quite off-the-shelves and cheap[ish].

Depending on how bright you need it to be, things can be very inexpensive. As you require greater brightness, you begin to need arrays of the very powerful LEDs, and quite high-current systems to drive them. So they could be somewhat expensive. But I have no idea what it would take to simulate the kind of brightness that we get from a "real" xenon tube system.

Some experimentation is in order

All this pushes are further and further towards a set of questions
concerning a real latency times of PCBs in our optical computers of the
Eos series, latency and reaction times of sensors, but also of the Digic
FW and HW as well - any sources of information on these? And I'm well
past the time when my knowledge here was approaching any kind of
reliability - out of my depth indeed, but sensing that these are of some
importance too,

I think that for the purposes of doing what you're after, we can figure that the sensors in the cameras will behave just the same as film would in the same situation. But maybe not! Maybe the photodiodes of the sensor would end up being noticeably slower to respond than we'd like. Hmmmm

If we can build a flash illuminator unit which triggers the same way as any other external flash would trigger, but which would emit a burst of very rapid pulses instead of a single pulse, the rest of the problem would take care of itself automatically - if the sensor diodes are fast enough.

I could build such a flash unit, but it would need to operate in a "manual" mode instead of ETTL of any sort.

So you'd need to determine exposure for any given ISO/aperture/working-distance by trial and error (just like we do now with these flashes in their manual modes), and then make your shots.

ETTL would not be possible (at least for me to devise, since I just don't know how that communication works at all).

-- hide signature --

Jim H.

Complain
Post ()
Keyboard shortcuts: