"Single-shot compressed ultrafast photography at one hundred billion frames per second" - http://www.cnet.com/news/worlds-fastest-2d-camera-can-captur...

Here is a pretty good video at a trillion frames per second: https://www.ted.com/talks/ramesh_raskar_a_camera_that_takes_...

The cameras doing a billion+ frames/sec do not really function the way traditional cameras do -- i.e. you could never use one to take a picture of lightning (which is a pity). They work by firing off billions of extremely short bursts of laser light, and capturing images at progressively longer offsets from the release of a pulse. That strobing light is an integral part of how those cameras work, so they can only be used in situations where you can control the illumination of the object.

(Which does not make that technology, mind you, any less cool. But it's different enough from traditional photography that I thought it was worth mentioning.)

Electricity travels at about one third the speed of light. That would make a hard limit. Both shutter and data transfer speed.

But that would just be for one recording device. If you could use multiple devices, perhaps with small POV differences that could be corrected, I would expect no limit in that "lossy" case.

But if you surrounded an event with a sphere of sensors, there is still an upper limit. The number of sensors is a function of sphere radius, and whether they are limited to the surface of the sphere (which is reasonable since tightly packed sensors will, in general, get in the way of other sensors).

A sphere 1km across has a surface are of 4pi1km^2 which is roughly 10^7m^2. If you had 1cm^2 sensors then you get 10^4 sensors per m^2, which means you get 10^11 sensors total. At 11k fps (10^4) that's a total of 10^15 fps.

Note that Planck time is 10^-44 so such an arrangement probably wouldn't help for detecting quantum gravity because you'd run out of matter in the universe before you could construct the sensors.

It is not clear how one would synchronize the cameras on such fine time scales either.

Amount of total photons emitted or hitting the Planck limit, whichever comes first.

POV differences could be eliminated by a simple lens & mirror setup.

Ultimately, sure, you run out of photons.

Let's start off with a baseline image resolution of 1 megapixel.

1 Joule is about 6E18 visible light photons, give or take. So that's then 6E12 photons per pixel.

Full sunlight on Earth has an intensity of around 1400 Watts (Joules/second) per square meter, or 0.14 Watts per square cm.

So achieving a frame rate in the trillions of fps would likely run into some limits relative to the size of your optics, the intensity of light coming from the scene, or the bulk quantum efficiency of the sensor.

And, of course, if you try to achieve quintillions of fps then you start dumping in nearly an entire Watt in photons into every single pixel on your sensor, which rapidly becomes a very big problem re: everything melting.

Additionally, modern electronics has trouble operating above the low GHz range, beyond that you'd need to move to superconductors or some such to keep up with the data.

The number of captured photons is a good hard limit. dalke's answer for ultrafast photography is kind of "cheating" in the sense that they record the same event periodically so they can gather enough light. That technique can't record single unique events that fast AFAIK.

No, it's not "cheating."

The title of my link says 'single-shot'. The underlying Nature article at http://www.nature.com/nature/journal/v516/n7529/full/nature1... says "Here we demonstrate a two-dimensional dynamic imaging technique, compressed ultrafast photography (CUP), which can capture non-repetitive time-evolving events at up to 10^11 frames per second"

Not that it isn't awesome how fast cameras are, just keep in mind, the clip is 25ms. So at 11,000 fps, that would be 275 images (11 frames per ms). Not sure if we actually have a camera that can take 11,000 images in 1 second. (I guess that depends on the resolution)

We do have those, but it does indeed depend on the resolution.

These can do 25k fps at 1280x800 pixels (and 1 million fps at 128x32 pixels): http://www.visionresearch.com/Products/High-Speed-Cameras/UH...

Yeah we do. The company I work for designs sensors that can pull that off.