Hacker News new | past | comments | ask | show | jobs | submit login
Single-shot compressed photography at one hundred billion frames per second (nature.com)
115 points by mlrtime on Dec 4, 2014 | hide | past | favorite | 32 comments



Good lord. I was ready to be disappointed by another streak camera, but this is insane. If I understand it correctly, it bounces the image off a micromirror array with a pseudorandom patterns being shone on it for sub-nanosecond durations, then collects the composite image by sort of scattering it in a known way over a CCD, and the original image can be sussed out of that once the data has been pulled down.

10^7 frames per second! Anyone have any ideas what we could watch at a rate like that? You only get 350 frames, though, it seems, and the spatial resolution is probably only a handful of pixels.


It is not 10^7 FPS. It is 10^11 FPS.

And they do collect that scattered light on a streak camera - that performs shearing operation in the temporal domain (camera moves during capture). Without it it would be impossible to reproject the data into these 10^11 FPS, you would only have your regular 10^7 FPS, achievable by CCD/CMOS.

edit. The temporal / spatial resolution probably can be increased by using larger sensor :) & Anyone wants to hack a DIY version of it? Take a Lytro camera, place it on the piezo actuator, do some fancy math ;)


It's 10^11 frames per second actually -- light travels 3 millimeters between each frame.


That's pretty fascinating.

Isn't that useful beyond photography? If this detector can detect just a single bit of information every 10^-11 second, that's 100 Gbit/s.


For a single pixel!


Oops, you're right. I'd edit if i could.


Is it related to compressed sensing? http://en.wikipedia.org/wiki/Compressed_sensing

Edit: I was thinking about the single-pixel camera: http://dsp.rice.edu/cscamera


Yes it is compressed sensing. Compressed sensing is where you have an input that is randomly 'combined' and then you reconstruct the signal you want (it has to be sparse in some way) using the L1 norm.


Looks like you can watch a laser pulse bounce off a mirror or difract into a resin. http://www.nature.com/nature/journal/v516/n7529/fig_tab/natu...


These are movies of things I never thought we would have movies of. It reminds me of when the first pictures from STMs came out - we never thought we would see pictures of atoms and molecules in our lifetimes - even if they were just arrays of bumps on a surface. Praise to the team that did this work, it's a gift to the world.


Now that technology is here, I do not think it would be as much of a challenge to increase number of frames compared to original research. Also can this resolve what is going on in the cell? Would be quite a boon to protein folding research if we can see clearly how the process happens.


Nope because that is limited by the diffraction limit for white light or evescent propogation for FL. You might be able to come up with something clever with FRET but I dont see comming up with something clever being limited by fps speeds.


A nuclear explosion please.


Free read-only link: http://rdcu.be/bLy6


Not sure if it's because of HTTPSwitchboard or not, but copy/paste, print, etc. all seems to work fine. The loading time is pretty bad however. (Also, thanks!)


Here's a link to the supplemental videos for the article:

http://www.nature.com/nature/journal/v516/n7529/fig_tab/natu...


#4 is really nice; light moving at different speeds in different media. I'm also struck by #2, light reflecting off a mirror, because some of the photons apparently carry on through the mirror (he said, as if recording the main packet bouncing off it were no big deal...).

In the near term, could this technology lead to to the development of improved beam splitters, mirrors, bottles etc. that would drive the development of optical computing?


Quoting the caption from figure 3 in the article:

"The bright spot behind the mirror is attributed to imperfect resolution."


What is that "faster-than-light propagation of non-information" they mention observing?


I've read a pretty cool, although somewhat useless, example of this a while ago. Imagine towers with powerful projectors pointing up spaced out 10 miles apart all the way around equator. You configure towers such that they sequentially activate projectors for a small fraction of a second, and only one tower is active at any given time, and at least one tower is active at any given point of time, and towers must activate sequentially.

Now with this setup if you dial down activation time so that it is smaller then 53.7 microseconds [0] next tower will not be able to receive any communication from previous tower on when to activate, signal would just not travel fast enough. Activations of towers would need to be scheduled. Now consider an observer looking at Earth with a powerful telescope from a significant distance. To him there would be an object emitting light that is moving around earth faster than light.

[0] 10(miles towers apart) / 186282(light speed in miles/second in vacuum)= ~0.0000537 seconds = 53.7 microseconds


So, basically, you fake it with carefully timed "cron jobs", setup well in advance of the experiment, but there's no real communication between instances during the experiment.

They just seem to be real-time-coordinated from outside because of the way the cron jobs are firing.


It's when you (for example) sweep a beam across a field. The spot of light can move really quickly, even crossing the field faster than light itself could, but it's not conveying information from where it's been to where it's going. Nothing physical is moving at anything even approaching the speed of light.


I hadn't seen "non-information" used as a way to describe that sort of thing before, but it makes a lot of sense. Avoids a whole bunch of unproductive arguments about whether something's "real" or not.


Wow, I've never considered such a thing before. Can this be used for anything?


Sadly, no. Imagine an immense wall light-years wide and a light-year or so away from you. If you have a (really!) bright laser pointing at it, you can wiggle around the dot so that the dot moves at FTL speeds. But, there'll be at least a year's lag from your wiggle to the dot moving (and another year before you can see the dot move), so you can't use it for signaling faster-than-light or anything like that.


unless there's an alien with another laser pointer on the other side ;)


Typical example is a wave crashing on the shore at a slight angle. Wave may be going only a few mph. But the wavefront breaking on the sand will appear to move horizontally at tremendous speed.


I look forward to watching a movie of photons getting trapped in quantum wells during photosynthesis sometime soon!


The paper is clearly written, read it if you can.


Amazing! You can actually see the laser pulse bouncing off of the material!


Ramesh Raskar presents femto-photography, a new type of imaging so fast it visualizes the world one trillion frames per second, so detailed it shows light itself in motion. This technology may someday be used to build cameras that can look “around” corners or see inside the body without X-rays.

http://www.ted.com/talks/ramesh_raskar_a_camera_that_takes_o...


*jaw drops to the first floor




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: