Slashdot Mirror
Scientists Build World's Fastest Camera
Hugh Pickens writes "Researchers have developed a camera that snaps images less than a half a billionth of a second long and can capture over six million images in a second continuously. Dubbed Serial Time-Encoded Amplified imaging, or Steam, the technique depends on carefully manipulating so-called 'supercontinuum' laser pulses. While other cameras used in scientific research can capture shorter-lived images, they can only capture about eight images, and have to be triggered to do so for a given event. The Steam camera, by contrast, can capture images continuously, making it ideal for random events that cannot be triggered. Keisuke Gode, lead author of the study, and his colleagues used their camera to image minute spheres flowing along a thin tube of water in a microfluidic device." (More below.)
High Pickens continues: "Using the STEAM camera they were able to image the spheres at a frame rate of 6.1 megahertz — in other words, the camera took a picture once every 163 nanoseconds. The camera could be used for studies of combustion, laser cutting and any system that changes quickly and unpredictably. One important application would be analyzing flowing blood samples. Because the imaging of individual cells in a volume of blood is impossible for current cameras, a small random sample is taken and those few cells are imaged manually with a microscope. 'But, what if you needed to detect the presence of very rare cells that, although few in number, signify early stages of a disease?,' asks Gode, citing circulating tumor cells as a perfect example of such a target. The team is working to extend the technique to 3-D imaging with the same time resolution, and to increase the effective number of pixels in a given image from 2,500 to 100,000."
But is this fast enough to photograph my wife with a closed mouth?
particle events. super hardon collider type things.
and simple things, like water drops forming, ice forming. the more detail you record, the more you learn exists.
-.no
particle events. super hardon collider type things.
Is that some new porno I'm too old to understand?
I just did.
Want to see it again?
Comment removed based on user account deletion
And what kind of storage do you need for a study that takes days or weeks?
Who the hell sifts through all these pictures after they are taken?
...image pixel area is 1x1!
Super hardon collider? Dude, if you're doing it that fast; stop. Just stop. It's really meant to be enjoyed.
Ask her to say something nice about you.
well, 1 second of footage = 6,000,000 frames / 30 fps = 200000 seconds
that would mean a 55 hour movie of a lightning strike
it would also mean a sports replay that would last well into next year
It's not the storage, it's the bandwidth. 50x50 is about 2k, so if you're only filming for a millisecond, storage isn't an issue -- 2GB is all. But you're not getting that onto a disk in a millisecond; you'd have a hard time getting it onto RAM.
On the other hand, if they're changing from 8 frames to a few hundred or thousand, that should be doable, and it's a huge leap forward.
Using FireWire, of course! Because USB would suck for this. Actually, USB sucks not just for this.
Ezekiel 23:20
Couldn't you just throw away all the pictures that look exactly the same somewhere between the camera and the storage device?
Hmm, what's the difference between a single-shot radiation hardened FILM camera built in the 1940s designed to take pictures of ENORMOUS & insanely bright things (Nuclear explosions) and a 'camera' that records interference patterns in light to film CELLS at 6 million frames per second?
Gee, I dunno, they sound pretty similar to me.
This new one only has an imaging area of 50x50 pixels - the film in the Rapatronic can surely beat that!
The opinions in this post are ficticious. Any similarity to actual opinions, real or imagined, is purely coincidental.
You missed the fact that you're a fucking moron. What the FUCK are you doing on /. if you aren't impressed by useless cool gadgets and cant see what the uses of the useful cool gadgets are?
it would also mean a sports replay that would last well into next year
Well, finally, a technical justification for how long sideline "instant" replay reviews seem to take.
Welcome to the Panopticon. Used to be a prison, now it's your home.
Comment removed based on user account deletion
Steam? sounds like vaporware to me.
My sig has been answered.
or, put differently, if the thing were light sensitive enough (which they never are, you have to bombard the scene with photons which typically causes heat issues, but that's another topic)...
6,000,000 frames per second means that each frame takes 1/6,000,000th of a second.
I know, dur, right? Here comes the awesome bit.
The speed of light is 299,792,458 meter per second.
divide one by the other (or multiply if you take the fraction): 299792458 meters per second / 6000000 frames per second = 49.9654097 meters per frame.
In other words, if you'd turn on the light at one end of a 400 meter street and start recording near that light source at that very moment, you could actually see light expand from the light source along the street to the other end in ~16 frames (the light has to travel back to the camera).
It would be a real world representation of the relativistic raytracing experiments regarding travelling light here:
http://www.anu.edu.au/Physics/Searle/
Note that 6 million frames per second is not the impressive part about this camera, though. The fastest camera reportedly does 200,000,000 frames per second; but it has lower resolution, only lets you capture a few frames, etc.
an uncompressed half hour video at 60fps eats around 10~40gb depending on resolution standard, so an uncompressed half hour video at 600,000,000 fps would take around 100~400pb? times that by 48, you would get 4.8~19.2eb per day, and 33.6~134.4eb per week. nothing a SAN can't handle... now the problem is if the camera is movable. Although... they really need to consider making movies using 120fps cameras. watching IMAX in 24fps is killing my eyes.
Where is the "Ignorant" mod tag?
Except it's not .5 billion per second. It's basically a .5 billionth of a second shutter speed, but only 6 million frames per second. It's also 2500 pixels per frame, not 2500 bits per frame. Let's just say it's 8 bits per pixel.
6,000,000 x 2500 x 8 = 120 gigabits per second, or 15 gigabytes per second.
What kind of processing power do you need to analyze six million frames per second in real time? (Honest question, I don't know, but I'd imagine it's just as ridiculous as the storage requirement raised by GP.)
I suppose you could analyze every nth frame, but if you're looking for events that occur on the microsecond scale, you run the risk of missing it entirely.
And what kind of storage do you need for a study that takes days or weeks?
According to the summary:
The team is working to extend the technique to 3-D imaging with the same time resolution, and to increase the effective number of pixels in a given image from 2,500 to 100,000.
I don't think an image with 2.5k pixels (or even 100k pixels) take that much storage.
Cosplayers.net - The Cosplayers Network
A 6-million core Beowulf cluster? ;)
Nah. At 2500 pixels (not mega! not even kilo!), you only get 6100000*2500*3 = 45.75 GB/min, or (*60) 2.745 TB per minute. ^^
Want a full two-hour movie? Well, then you only need 329.4 TB of space
What you say? Full-HD you want? Then 273.21754 petabytes you must have! All your bytes are belong to us!
Any sufficiently advanced intelligence is indistinguishable from stupidity.
or, put differently, if the thing were light sensitive enough (which they never are, you have to bombard the scene with photons which typically causes heat issues, but that's another topic)...
Light sensitivity is a big issue in some applications of fluorescent microscopy. Not heat sensitivity but photobleaching. I don't understand anything about the quantum physics involved, but the fluorophores lose their ablity to give a signal the longer they're excited, it can be rapid and it's annoying in many applications.
I look at cells on a confocal microscope, it uses a laser of one wavelength to excite fluorescent proteins, which causes it to emit light at a different wavelength. Filters can be used to see just the emitted wavelength, so I can tell which cells have the protein and where it is within the cell. Even if I turn the laser down to %1 and take fast images, there is some minimal loss of signal. Wouldn't be a problem, except that sometimes I need to make time lapse movies for up to 20 hours. That many exposures add up quickly, and I commonly see cells go dark due to photobleaching. There are plenty of tricks availiable, but if I could take pictures faster, that would be better than some of the other compromises I have to make, like turning down the resolution or resetting the contrast.
Quantum dots and some inorganic fluorphores I understand are more resistant to photobleaching, but they're not very good yet as far as I know for live cell imaging. For that we need to use fluorescent proteins, which are both dimmer and less photostable.
I don't know if this technology will actually be useful or even compatible with confocal microscopy, but if it could cut down on the exposure time required, that would really help lower loss of signal and/or help with resolution issues. TFA seems to indicate they're developing this with microscopy in mind. Hopefully they'll make a microscope with it quickly, like say a month, and it will be "cheap," like under $400K. And they'll send me one free. With a pony.
If you're only looking to capture a few seconds, just put it in RAM and write it to long-term storage later. Write speeds for high-end consumer RAM are in that neighborhood. DDR3 1800 can write just over 14GB/s. For a research project, 128 GB of RAM is certainly feasible. That will give you a full 9 seconds of video.
If you need more pixels you can line up arrays in parallel to capture several seconds from each array at the same time. They can all use the same clock so everything stays synchronized.