MIT's New Camera Can Take 1 Trillion Frames Per Second

First time accepted submitter probain writes "MIT has made a camera that can take trillion frames per second! With this high speed capability, they can actually see the movement of photons of light across a scene or object. This is just mind-boggling." ExtremeTech has a nice video of the system, too. What would you like to see slowed down to such a degree?

Comments

What

[Anonymous Coward]

they can actually see the movement of photons of light across a scene or object

...no.

Re:What

[buchner.johannes]

Well they can, just not individual photons or individual photon events.
It's exactly the same as an oscilloscope -- you also don't see the shape of an individual pulse. You under-sample, and then add the samples together assuming it was always the same pulse.

Re:What

[vlm]

they can actually see the movement of photons of light across a scene or object

...no.

Sure you can. I've seen the nuke test footage. I'm not talking about the stereotypical "wind gust front" but actual "light" output. The first few frames are classified and I have not seen them, because they show asymmetries that imply various things about internal construction, but once the fireball gets a couple feet across its pretty much perfectly spherical and that's the unclassified frames I've seen. If there were a useful way to search youtube / google / archive.org for a description like this, I'd give you a link to the actual movie. You can distinctly see the disk of light hit the ground and expand very rapidly circularly underneath the slowly growing fireball, well, slowly growing compared to the speed of light, anyway. There are not many frames to this "movie" probably synchronization of the cameras and the "bang" was harder back in the 50s. The footage is many decades old.

I believe the relevant part of the story is this might be the first "trillion fps" camera that isn't classified and is owned by "civilians"

Re:What

[vlm]

Well they can, just not individual photons or individual photon events.
It's exactly the same as an oscilloscope -- you also don't see the shape of an individual pulse. You under-sample, and then add the samples together assuming it was always the same pulse.

Only with digital scopes. With analog that's exactly how it works, you can, if you want, see literally one pulse. Not much analog scopes on professional desktops anymore... they're all on hardware hackers basement desks now, like mine. Thats why I bring it up, on average across /. readership there are probably more analog scope users than digital scope users. That would make an interesting /. poll,
1) I use an analog scope
2) I use a digital scope
3) Cowboy Neal is a my scope
4) Whats an oscilloscope?

Re:What

[GrpA]

No... The OP is correct.

This isn't new technology. It's called "Gated Image Intensifier Photography" and is used for everything from Lidar to special night vision devices that can see underwater. It is one of the few technologies that allows detection of stealth submarines by taking images of the submarine without the backscatter caused by water in front of it. It's one of the only technologies that can track supercavitating weapons underwater. It can also help see through many obscurants.

It's like a flashlight, except you only look at light reflected at a particular time after the flash ( usually a laser ) goes off. As a result, you can choose to see light that is only reflected from, say, 100m away to 101m away. Everything else looks dark and because of this, it's a good technology for seeing through trees and the likes.

If you want to understand gating of image tubes and streak tubes in particular ( what they use - an electronically steerable image intensifier that can track very high speed objects such as bullets being fired from a gun ) just look up Image Tubes by Illes P Csorba. A great book.

What they are doing here is just gating the image a little faster and repeating it often to capture very short duration repetative events in high detail. Not a new technology, just a variation on existing tech.

And you'll find many modern Gen3 NV devices are autogated, meaning they do this automatically, though it's more a way to pulse-width modulate the light coming in so that they can work under brighter conditions, such as when soldiers burst into a room and the enemy turns on the lights inside...

GrpA

Re:What

[Anonymous Coward]

Not true either -- some digital scopes (the super-high-bandwidth ones) work that way, so all you need is a wicked fast sampling circuit, instead of making the whole thing wicked fast, but most "ordinary" digital scopes are perfectly capable of one-shot captures.

And OTOH, most analog scopes I've used, when you use the fastest available horizontal sweep, the beam is dim enough I can't see a single event -- so unless you're using a storage scope or one of those polaroid capture widgets, you're forced to visually average across multiple sweeps here as well.

But yeah, I got a couple analog tektronixes for the odd electronics project. Nothing impresses non-geeks like hooking up a scope to an RS232 line and reading off the bit rate and data format...

Re:What

[Maddog Batty]

A succinct answer and basically correct.

My understanding of this is the clever part is the very short pulsed laser combined with a very short exposure camera. Each laser pulse send lots of photons together in a bunch across the field of view of the camera. Some clever camera synchronisation allows each "frame" taken by the camera to be slightly (pico seconds) later than the previous one. When run as a movie, this appears to show a light pulse as moves across the field of view.

However, it doesn't take a picture of a single photon - it takes pictures of a bunch of photons and neither does it take a movie of the same bunch of photons moving across the field of view - each frame is taken of a different laser pulse.

(for simplicity I've ignored the fact that the camera is a line scan camera rather than full frame camera)

Re:What

[vlm]

If you have a rare and expensive quad channel scope, watch the TX and RX, AND the hardware control lines and have fun telling them how fast their interrupt service routines are, this used to completely freak out OS/device driver developers (so... you mean you just look on a scope, instead of hand counting theoretical instructions?)

I will admit you are correct, if you have way too much money you can buy direct non undersampling digital scopes. Or I suppose if you're only monitoring audio speed signals or whatever.

Re:What

[vlm]

thats pretty interesting. Is a possible countermeasure against this, parking your tank behind two nearly perpendicular mirrors so the beam path between the two mirrors is like 100 miles?

Or would parking your tank behind a mirror make the terrain behind the tank look just like the terrain in front of the tank, but backwards... that might be too easy to detect after all...

Re:What

[Anonymous Coward]

The O-scope you are describing is a sampling scope. Where the sample rate is very slow compared to the signal. The trigger point is moved and a repetitive signal (PRBS type or others) can be reconstructed. This type of sampling assumes that the channel parameters are virtually static. Digital Real-Time scopes exist (and have existed for a long time) that oversample (80+ Gigasamples/s) the signal and display an accurate representation of a single pulse.

Re:What

[GrpA]

Yes, since it's "Active" illumination, you just detect from which direction the flash of light is coming from - though you need a very wideband detector since it could be anywhere on the spectrum and will almost certainly be infrared above 1000nm...

Then once you see the person aiming their "LIDAR" at you, you swivel the tank's gun and send some high-velocity non-photonic matter their way... Probably the most effective countermeasure.

Most of this technology uses very long wavelength ( around 1500nm ) light so that it's not going to be obvious what you're doing. It also tends to work over very long distances, eg, 10Km away... :) It's more used for detection and identification of enemy equipment at long range under conditions of darkness.

Even then I don't think it's all that common. Thermal is more practical for detection now and I imagine Lidar is special use only ( eg, when very high resolution images are required, when topological information is important or for underwater use ) -

GrpA

Re:What

[tzanger]

Nothing impresses non-geeks like hooking up a scope to an RS232 line and reading off the bit rate and data format...

I'm pretty sure non-techies don't have a clue what RS232 is, nor care about the bit rate and data format. :-) It does tend to impress the protogeeks and the more experienced geek squad techs, though.

Re:What

[Anonymous Coward]

I believe the relevant part of the story is this might be the first "trillion fps" camera that isn't classified and is owned by "civilians"

Streak cameras with picosecond resolution or below have been available off the shelf for decades. They were really expensive and are still pricey now, even if a lot cheaper. So not within the reach of most hobbyists, although still available to civilians with money or some salvaging luck.

I am not sure about being able to see an advancing light front off of nuclear test footage, at least in a 2D film instead of something like a 1D stream camera. By the 60s, I think the highest speed cameras were on the order of 10,000 fps, which means light would go about 30 km in one frame. Even with current technology, it would be difficult to have a full movie at such speeds of a single event, as the highest speed cameras are usually limited to some set number of frames before requiring some time to reset (something like 1-8 frames are common, although larger ones exist).

Re:What

[The Raven]

This was not a single camera; what you're seeing is the images from 10 closely spaced cameras that were triggered at slightly different times. We could take very quick images in the 50's... we couldn't take a video. Video capture is new.

Re:What

[jdastrup]

TFA says they are using 500 cameras (CCD's) to accomplish this task. And isn't video simply many still frames put together? So what's the difference?

Re:What

[timeOday]

Not a new technology, just a variation on existing tech.

I have never seen a new technology appear that was not met with that reaction on slashdot. I think people have some serious misconceptions about what science looks like when you follow it on a daily or weekly basis. Step functions exist only in theory.

Re:What

[vlm]

Then once you see the person aiming their "LIDAR" at you, you swivel the tank's gun and send some high-velocity non-photonic matter their way... Probably the most effective countermeasure.

I detect an economic problem when the cost of a dumb "transmit only" laser diode and microcontroller to pulse it costs less than cannon round. A tactical problem if you build a "tank detector" fortification using a dazzler in a safe area (nothing important downrange of it) and an anti-tank team preposition along a flank of the dazzler's LOS, the big boom from the tank wakes up the sleeping opfor anti-tank team who promptly make an even bigger boom outta the tank... Also a tactical problem if your tank only carries 40 odd rounds and the opfor issues a clip on dazzler decoy device to a couple hundred infantry escorting like two real live opfor tanks, hmm which do I pop odds are only 1 in 100. I suppose you pop the two big thermal plumes, but still the freak out factor must count for something, maybe combined with other surprises...

20 yrs ago I worked on the logistical computer systems for ammo in the USAR so I know those shells are quite expensive, and I do stuff with cheap microcontrollers and some laser diode stuff now, so I know I can build a decoy for maybe 1/10 the cost of a shell.

Re:What

[dna_(c)(tm)(r)]

Well they can, just not individual photons or individual photon events. It's exactly the same as an oscilloscope -- you also don't see the shape of an individual pulse. You under-sample, and then add the samples together assuming it was always the same pulse.

That would make an interesting /. poll, 1) I use an analog scope 2) I use a digital scope 3) Cowboy Neal is a my scope 4) Whats an oscilloscope?

3) CowboyNeal is out of scope

Re:What

[Surt]

Bah, and today is the first time in seemingly months that I have no mod points for such a worthy post.

Re:What

[freeze128]

What's a Cowboy Neal?

Re:What

[Surt]

His point is that what you are capturing is different photons involved in a wave. As soon as you capture any given photon, it's no longer a participant in the wave, so you can't capture the movement of even one photon, much less multiple photons. What you do capture is some photons characteristic of those moving across a scene.

But really, he's just being a pedant troll.

Re:What

[Hatta]

What you are seeing is probably a sphere of expanding plasma. It can't possibly be a "sphere of light" because light travels at the speed of light. By the time you could see the sphere, you would be inside it.

Re:What

[kimvette]

Gated cameras have been out for a while - initially they were classified as munitions (they still may be, I'm not sure). ExtremeCCTV (Now part of Bosch) had a full line of such cameras, and the huge advantage of this tech is it enabled surveillance through even dense fog and blizzard/whiteout conditions. I can't find the tech on Bosch's site now but I found these:

http://www.usborderpatrol.com/Border_Patrol1604_8.htm (just an article about the practical application)
http://www.laseroptronix.se/gated/aqly.html
http://www.obzerv.com/night-vision-cameras/range-gating-technology/core-expertise/

Oh, and the range of this tech is much greater than those pages indicate; in 2003 the units were good for 25km+ (that was the publicized spec), and the tech (both lasers and sensors) has improved since then.

Re:What

[geekoid]

no, you can't. The poster is talking about packets of photons.

What you are describing is different. And it an't a trillion frames a second.
.

Re:What

[rev0lt]

In many modern processors, you don't really need to "hand count" instructions, you can probe the performance counters. In many microcontrollers, you can do the same with one of the timers. By your description, I'm assuming you are using the oscilloscope on a somewhat standard serial connection without buffering. If you'd try it over a buffered connection (such as a decent network card), you'd see (and given that you can actually use an oscilloscope, you probably already know that) why there is no direct relation between bitrate and interrupt rate. I'd argue that a device driver/os developer that would be impressed by that probably is lacking in many areas.

Re:What

[Silfax]

1) I use an analog scope
2) I use a digital scope
3) Cowboy Neal is a my scope
4) Whats an oscilloscope?

analog & digital - but mostly analog

Re:What

[falzer]

1) Analog, used it just yesterday.

Re:What

[atisss]

Except that tank won't probably waste their shells on unconfirmed target. At best you can hope on couple of .50 machinegun shells in direction of microcontroller.

Re:What

[atisss]

So in order to capture nuclear explosion in 720p you would have to make 720 perfectly synchronized detonations at the same spot.

Re:What

[wagnerrp]

Active gated cameras aren't a new technology, they're just a variation on Fizeau's original apparatus for measuring the speed of light. This is nothing more than 1850's tech. Friggen steampunk imagers...

Re:What

[EdIII]

To put it another way, it's like saying the BBQ is just a variation on a campfire and not really all that impressive. Well... that's a pretty interesting variation with many practical applications.

Variations, if they allow new functions and capabilities, does not invalidate it's worth simply because it re-purposes existing tech.

Re:What

[wagnerrp]

It's one of the only technologies that can track supercavitating weapons underwater.

The other being bog standard passive sonar? Cavitation makes noise. The bubbles collapse violently with a loud pop, making a ton of noise, which is why submarines try so hard not to do so. Supercavitation makes a fuckton of noise. The idea is that you're just going so damn fast, no one can do anything about it. Most supercavitating concepts run 50-150m/s, while the speed of sound in water is 1500m/s. That's plenty far out in advance to track it acoustically just fine.

Re:What

[Dogtanian]

I'm pretty sure non-techies don't have a clue what RS232 is

That's what you think. Of course we've heard of RS232, he was C3PO's boyfriend in the Star Trek films!

Re:What

[Dogtanian]

What's a Cowboy Neal?

Is this a joke? Your user ID is (marginally) lower than mine, and I've heard of him and remember the days when almost every poll had a Cowboy Neal option.

Re:What

[Agripa]

Not true either -- some digital scopes (the super-high-bandwidth ones) work that way, so all you need is a wicked fast sampling circuit, instead of making the whole thing wicked fast, but most "ordinary" digital scopes are perfectly capable of one-shot captures.

Most digital storage oscilloscopes now have single shot sample rates high enough to support their full bandwidth. Analog to digital converters from 500 MS/s to 2 GS/s and faster have been readily available for years. Only the toy ones are slower. Internally the digital interface to the converter is more difficult now than the analog to digital conversion itself but you can do it with an FPGA.

My old (old enough to drink) slow 20 MS/s and and 100 MS/s DSOs are what I use the most. Of course in equivalent time sampling mode, even they support 2 GS/s.

One problem with most digital storage oscilloscopes is that the dead time between record captures can make elusive behavior difficult to find without special triggering. The digital storage oscilloscopes with the fastest capture rates (and highest prices) are still slower than the fastest obsolete analog oscilloscopes.

And OTOH, most analog scopes I've used, when you use the fastest available horizontal sweep, the beam is dim enough I can't see a single event -- so unless you're using a storage scope or one of those polaroid capture widgets, you're forced to visually average across multiple sweeps here as well.

Some analog oscilloscopes are better than others for that. I have a Tektronix 7834 400 MHz storage scope for that very reason but in direct view mode, a 7904 gives a significantly brighter image. At least two oscilloscopes Tektronix made use a microchannel plate to intensify the CRT image so you actually can watch low repetition rate high sweep speed phenomena without a dim trace.

In the case of the 7104, the combined horizontal and vertical bandwidth support a writing rate faster than the speed of light and you can actually see it even in single shot mode.

Re:What

[The Raven]

You are misunderstanding what we are discussing. The nuke test images are from the 50s, very old images. There were several insanely high speed stills of a nuke's first moments. The difference between a photograph and a video in this case is that the position of the lens is different for each still.

Wish to see?

[unixcrab]

1. The electron beam scanning in a CRT. 2. Inside a cylinder of an internal combustion engine. 3. A lightning strike (too difficult maybe)

Re:Wish to see?

[Fahrvergnuugen]

Pretty cool even for "normal" high speed photography

Re:Wish to see?

[Jaqenn]

Tim Samaras is a storm researcher who has captured lightning strikes at 10,000 frames per second:

http://www.youtube.com/watch?v=EyUsjsJ-E0c

It's not 1,000,000,000,000 FPS, but it's still pretty cool.

Re:Wish to see?

[RivenAleem]

10,000 FPS should be enough for anyone.

Re:Wish to see?

[Fri13]

24 is....

Re:Wish to see?

[Surt]

I know you're being funny, but just because I hope to correct this misconception in the long run:
24 is not enough for everyone. The actual number required is closer to 240. That is the point at which not even the 99th percentile of eye responsiveness can detect the frames, and perceive instead smooth motion.

Re:Wish to see?

[Time_Ngler]

None of these things would work with their idea. The light has to come from a laser they control, because they are actually viewing hundreds of separate repeatable events and reconstructing them into a single movie.

Re:Wish to see?

[UnknownSoldier]

No offense, but Fuck You and your archaic stuttering video.

If you can't tell the difference between 24 Hz and 60+ Hz, you're blind.

Re:Wish to see?

[Anonymous Coward]

No offense, but Fuck You...

You don't really understand what "no offense means, do you? Also, fuck you for being disagreeably disagreeable.

Re:Wish to see?

[Hentes]

There are lightnings (especially the ones that turn back) slow enough to see with the naked eye. Some take 1/4s or even 1/2s to strike.

Re:Wish to see?

[inode_buddha]

1. The electron beam scanning in a CRT. 2. Inside a cylinder of an internal combustion engine. 3. A lightning strike (too difficult maybe)

A cumshot.

Re:Wish to see?

[avandesande]

Are these European or African photons?

They can see a photon???

[Joce640k]

I really, really doubt they can see a photon...

Re:They can see a photon???

[jellomizer]

At least now without altering the results.

Re:They can see a photon???

[vikingpower]

You can "see" a photon, although the act of "seeing" it destroys it. You can, however, definitely NOT "observe" a photon. Ever.

Re:They can see a photon???

[GrpA]

Nope, they can actually visualise a single photon if the gain is sufficient, eg, Super Inverter Image Intensifier ( also known as Gen3+1 ) - Typical photonic gain levels of around 300,000x -

Neat huh?

Of course, that's assuming the photon is converted into a photoelectron by the photocathode, which depends on the QE ( Quantum Efficiency ) of the photocathode material.. And assuming the photon isn't lost in any AlO films inside the device... Then yes, they can actually see individual photons.

They can count photons too with photomultipliers, but image intensifiers will let you put them into an image... :)

GrpA

Re:They can see a photon???

[vlm]

I really, really doubt they can see a photon...

If not photons, what are you using to see? The "photons moving across the scene" doesn't mean you see one moving from A to B, it means you see one presumably came from A because it was a pitch black room, hits B, bounces off and hits yer eye or camera. There is some geometrical / trig foolishness to correct the actual speed, but to the simplest approximation if you suddenly turn on a light at A, the position of B moves away around the speed of light. (Bonus points for calculating how fast imaginary construct point B moves when a beam of light hits a nearly perpendicular mirror "warp speed" here we come)

Re:They can see a photon???

[skids]

Actually there's an equivalent of the uncertainty principle for photons. Since we know their speed is c, the quantities we can know one or the other (or a probabilistic mix thereof) are not speed and position, but rather:

"The analogous uncertainty principle for photons forbids the simultaneous measurement of the number n of photons (see Fock state and the Second quantization section below) in an electromagnetic wave and the phase of that wave" ...so it is arguable that destroying one photon worth of coherent light counts as an "observation", since there's no distinguishing of one photon from another within the coherent packet.

Re:They can see a photon???

[X0563511]

Of course, that's assuming the photon is converted into a photoelectron by the photocathode

... which in turn alters the object/system under observation?

Re:They can see a photon???

[catmistake]

I really, really doubt they can see a photon from the side

FTFY (because all anything that sees can see is photons)

Re:They can see a photon???

[RogerWilco]

Single photon detectors have existed for more than 10 years I think. A guy I knew in the second half of the nineties was building them, I remember him getting an article published about it in Nature so it must have been pretty cool back then.

i'd be dead before that water balloon pops

[Anonymous Coward]

played back at 24fps, it would take over 1,000 years to watch 1 second of video captured at 1,000,000,000,000fps.

Re:i'd be dead before that water balloon pops

[SJHillman]

If my math is right, light travels a little over 1/100th of an inch (or a little under 0.3 mm) per frame.

Re:i'd be dead before that water balloon pops

[Anonymous Coward]

The problem is that photons ARE light, and the camera CONSUMES them by seeing them.

You can never see the same photon twice. The act of seeing absorbs it.

Re:i'd be dead before that water balloon pops

[Anonymous Coward]

My blind = mown.

Re:i'd be dead before that water balloon pops

[jellomizer]

Granted however if you run algorithms that will correctly motion blur each image. you will be able to view things at mostly any speed you want and not have it seem choppy.

I am happy to see technology getting to a point where improvements in some areas have become impractical as they are surpassing our ability to observe them.
Just like how the Now High resolution displays are getting to a point we no longer need to get any higher (We need to get a bit higher so we cannot see a pixel attached to a display on a contact lens) . Once we get computing to exceed our physical limitations improvements in system will then shift to focus on other areas that needs improvements. Part of the reason why PC don't seem to go any faster is the fact you were happy with windows at 600x480 display back in the day. Now anything less then 1024x768 is insane. And you should be higher then that. So that is more pixels on your screen that needs memory and processing power (now offloaded to the video card mostly) but still a lot of processing in terms of percentages goes to your interface. As we exceed our physical limitations processing will go to faster applications and processing of data and less to keeping up with the user experience.

Re:i'd be dead before that water balloon pops

[GrpA]

They are talking about shutter rates, not image capture rates... Big difference.

It probably has quite a slow frame rate.

GrpA

Re:i'd be dead before that water balloon pops

[DriedClexler]

Right, dude. Just how when you smell something, you consume molecules from it through your nose.

Therefore, you never smell the same molecule twice.

Therefore, you destroy every rose you smell.

Er ... yeah, maybe we need to not get ahead of ourselves.

Re:i'd be dead before that water balloon pops

[Surt]

But only a lifetime to watch at 240, which would be better for a smooth viewing experience.

Re:i'd be dead before that water balloon pops

[Eponymous Hero]

molecules != particles

Re:i'd be dead before that water balloon pops

[mwfischer]

Just speed it up. It'll only take about 1 second.

Re:i'd be dead before that water balloon pops

[catmistake]

analogous to the strobe light, dripping water thingy... where fast drops of water appear to be moving in slow motion because the strobe light only lets you see a drop at a certain point, and then another drop at a point a little further than the last... what looks like a slow motion drop of water is actually many drops of water, strobed.

I could use this at work...

[SJHillman]

I love the whooshing sound deadlines make as they fly by, maybe this will slow them down enough to see what they look like too!

Hmmm

[DWMorse]

What would you like to see slowed down to such a degree?

Hint: It involves a trampoline, or maybe a wet tshirt...

Re:Hmmm

[mbone]

You realize that at a picosecond frame rate it would take about a year of watching at 30 frames per second to actually see any motion of a person on a trampoline at all, and maybe a century to observe an single bounce ?

Re:Hmmm

[mbone]

Relying on Slashdot for sexual gratification is worse than relying on Slashdot for legal advice, and that's saying something.

Re:Hmmm

[Surt]

But if you want a smooth playback you should watch at 240 hz. So that would only take you most of your adult life to observe a single bounce.

Re:Hmmm

[SleazyRidr]

But what a way to spend a century!

Bullshit detector goes beep

[tpotus]

I watched the video and can only conclude that it doesn't make any sense at all. Slow motion video of a moving photon? Give me a break.

Re:Bullshit detector goes beep

[Anonymous Coward]

You don't get it. What you see are the photons deflected by the air off of their original path. When the light is in the middle of the bottle it has actually already exited it. The cameras capture the deflected photons. In a perfect vacuum you would not be able to see anything.

Re:Bullshit detector goes beep

[Aqualung812]

Exactly. How do you see a photon?

Re:Bullshit detector goes beep

[swalve]

It sounds like they aren't actually capturing 1T fps in real time. They are simulating it by capturing identical scenes at very slightly different intervals. Sort of a wagon wheel effect, or that effect that made the rounds a couple of months ago where they "captured" the movement of guitar strings. Take a machine gun that fires bullets once per second. Take a camera that takes photographs every 1.000000001 seconds. Fire a trillion bullets and take a trillion photographs. Each photograph will show a different bullet, one trillionth of a second further along the path. If you play them back, it looks like a single bullet going really slow.

Re:Bullshit detector goes beep

[root_42]

The FAQ on Rashkar's website (http://web.media.mit.edu/~raskar/trillionfps/) gives a good explanation. The whole thing is a repeated measurement process, with accurately timed recordings of the detector. The explanations and PR are (sadly) typical for SIGGRAPH papers nowadays. It's a bit of bullshit, you never "see" a photon flying through the scene. For this to happen, another photon would have to interact (reflect) off of this moving photon and be recorded by your camera.

Still, this method is quite interesting to visualize the propagation of spherical light wave fronts through a static scene. However, what I was missing was the imaging of reflected light, as it bounces through the scene. That would truly be educational. I could only make out the sampled pulse moving and intersecting the different objects in the scene.

Re:Bullshit detector goes beep

[GrpA]

Well I could just say "With your eyes" but I figure the question is "How do you see a single photon?"

You amplify it by converting it to a photoelectron with a very sensitive photocathode, then you add more electrons through either linear acceleration and multiple electron/photon stages or with a MicroChannel Plate ( MCP ) which causes secondary electrons to multiply the number of electrons, then you accelerate it over a short distance to around 5,000 to 10,000 eV and then smack it into a aluminized phosphor screen, which converts the electrons back to photons, but a HEAP of them so they are visible.

They can also focus and steer the electrons inside the tube. That's why it's called a "streak tube"... :)

I have seen photons many times. Kind of cool seeing a picture made from just a few photons, but it has to be REALLY dark to do this and you have to get your own eyes accustomed to the dark as well. The pictures sometimes just look like static until you collect a whole heap of them in a timed exposure.

When you amplify light about 100,000 times and then take a 15 second exposure and it *still* looks dark, you know the original image was exceptionally dark.

GrpA

Re:Bullshit detector goes beep

[vlm]

Exactly. How do you see a photon?

If its high enough energy / high enough freq / low enough wavelength we call it a gamma particle and you watch ionized air particles it left behind as it passes thru.

Thats kinda abstract because we're not looking at "it" but more what it did to the air as it wooshed by.

Maybe a closer example would be cerenkov radiation, essentially a visual sonic boom as one goes thru something with a lower speed of light (some plastics have a really low speed of light, which makes them pretty good lens material). I guess if you make a plastic scintilator type of thingy with a continuously changing refractive index so the particle continuously emits cerenkov radiation then you could kinda argue you're seeing "it", although we're destroying it in the process of seeing it so...

Cerenkov radiation is what leads to 2nd order BS WRT the color of radiation. Comic books trained kids radioactive stuff glows green, thats the 1st order BS. Actually visiting a test or experimental reactor quickly trains them to the 2nd order BS that radiation is blue (inspiration for star trek tng engines, I guess). Trust me, neutrons, by themselves, are not actually green or blue.

Re:Bullshit detector goes beep

[mbone]

I think they mean that in a perfect vacuum, there would be nothing in the bottle and thus nothing to scatter.

Re:Bullshit detector goes beep

[dmatos]

In order to resolve spatial location of the light pulse, at the very least they need to have a sensor with a response time on the order of 1ps. Being able to control the exposure time accurately with ps resolution is also a pretty incredible feat.

They have definitely admitted that they're not capturing the entire 2-d image at 1T fps. They might be capturing a 500-pix line at 1THz line rate, but that's unclear in the article and the video.

Re:Bullshit detector goes beep

[dmatos]

Ah. Further reading at the MIT site indicates that they are reading at "1THz line rate". They use a varying electric field inside the camera slit to deflect the photons by different amounts onto a 2-D image sensor. Thus, on the sensor, the x-direction contains spatial information, and the y-direction contains temporal information.

They can do this by sweeping the strength of the electric field inside the streak camera's slit quickly. Photons arriving at different times are deflected by different amounts, and thus hit different pixels in the 2-D sensor behind the slit.

Re:Bullshit detector goes beep

[cmd]

swalve is correct, the camera takes about an hour to capture an entire scene. To quote the video (starting at 1:32):

"So what's happening is, the camera keeps taking images and we very slowly rotate this mirror to scan our field of view across the entire scene. And because all of our pulses look the same we can in the end go and combine all these images we took to get one complete movie of the scene."

That movie of the light pulse had a different, but identical, pulse in each frame.

So what they have invented is a very accurate shutter trigger and a sensor with a very fast shutter speed.

The image sweep is accomplished by mechanically moving a mirror very accurately, but not quickly.

This camera rig is not really a slow motion camera, it does not capture a single event at 10^9 frames per second.

Cue the $6M man theme sound

[Rosco P. Coltrane]

Each movie that camera makes is dubbed with the sound of Steve Austin running for dramatic effect.

How soon until it's in my Droid?

[schwit1]

I'm sure somebody is going to shrink it so it can fit in my phone.

Slashdot Story Misses the More Exciting Point

[MyLongNickName]

This grew out of a system to see around corners. The professor wanted to build a camera that could analyze the path of reflected light to get pictures around ninety degree angles. This is a really amazing concept, moreso than simply getting a camera to take ever increasingly fast pictures.

if you are interested in learning more and have a lecture's worth of time on your hand, please check one out here: http://www.youtube.com/watch?v=aKu20y1f_RU

Re:Slashdot Story Misses the More Exciting Point

[swalve]

What IS really cool about this is that they have visualized light echos.

Re:Slashdot Story Misses the More Exciting Point

[RandCraw]

More info here:
        http://web.media.mit.edu/~raskar//trillionfps/

The work is based on 'streak camera' technology, which measures the change in the flow of a stream of photons. Raskar is extending the work from 1D to 2D and adding multi-point reflectance data to infer the 3D shape of objects that are otherwise occluded from view.

Clearly the image's spatial resolution will be limited by the smoothness of the reflected surface, the rate of motion of the target, additional noise sources (e.g. ambient light), perhaps even variations in temperature in the air (refractive noise), etc

AFAIK, all streak cameras are currently used only in very structured environments (particle accelerators, flow mass spectrometers, etc). The prospect of using a streak camera to render 2D objects is ambitious. Doing so in an unstructured setting like the battlefield (where do you think his funding is coming from?) is implausible. The prospect of combining such technology with computational photography techniques to practically 'look around corners' degenerates to hopeless hyperbole.

IMHO, the MIT Media Lab's reality distortion field is alive and well.

Streak cameras

[Anonymous Coward]

Streak cameras have been around for decades. They take a one dimensional source of light, and sweep it across a 2D detector very quickly so that the second dimension gives you the time resolution much shorter than the exposure time used by the sensor. Streak cameras with time resolution in picoseconds is pretty common, and many have sub-picosecond resolution. The problem is that once the a light source is swept across the camera, you are limited by the time it takes to read and reset the sensor before you can repeat the process, giving you the same repetition rate as high speed 2D cameras. So you might have 100 fs time resolution, but it would be one dimensional, and only last for 100 ps, before having to wait a few microseconds to milliseconds to take another image (there are some tricks to get two images given one sensor before reading it, and some high end cameras will just have multiple sensors in parallel to get faster successive images).

The novelty here seems not to be the camera, but the use of a laser for illumination and the stitching of many 1D images taken over an hour or so together into one 2D image.

Re:Streak cameras

[bws111]

Isn't that basically what photo-finish cameras for horse races and athletic events do? Those have indeed been around a long time.

Re:Streak cameras

[bws111]

Thanks for the info

Baywatch

[gaelfx]

Any time there is any advancement in the field of slow motion video capture, the only answer is Baywatch.

Light is about it

[mbone]

The current camera only does one line of a frame at a time and uses repeated laser pulses to synthesize a movie, but suppose that they upgrade this camera so that it can take a full frame in a picosecond (it's only engineering !). There is not much besides light (or a beam of high energy particles) that actually changes much in a picosecond - or even in a nanosecond (~ 1000 frames, or 30 seconds at 30 fps).

In 1 picosecond the ISS moves about 80 angstroms, or ~8 micro meters in 1 nanosecond. A bullet is considerably slower, as is a chemical explosion. The only thing else that I know of that changes much in a nanosecond is a nuclear explosion, and I bet that has already been imaged at the picosecond level, not that we are likely to see the films.

Oblig. Futurama (paraphrased)

[Dachannien]

Morbo: Photons do not work that way! Good night!

Seriously. You can't detect a photon unless it actually collides with the detector. So how do you detect movement of photons across a scene?

Re:Oblig. Futurama (paraphrased)

[DriedClexler]

So how do you detect movement of photons across a scene?

Well, you assume the photons expand out radially from the source, and therefore, watching the light propagate from left to right across your field of vision, you assume that the photos that hit your detector (such as your eyes), are roughly coplanar with the (majority of) photons that aren't hitting your detector.

You know, like what happens when you look at the beam coming out of a flashlight from the side.

(Remember: to make inferences, you must make assumptions. Your visual system makes inferences.)

It's just that, with fine-grained enough snapshots, you can see what the beam looks like before it has propagated across your entire field of vision.

Yes, quantum behavior of photons is complicated, but sometimes we overcomplicate it.

Re:Oblig. Futurama (paraphrased)

[X0563511]

So? Photons start striking an object. There's a "front" moving across that object as the photons hit it, assuming it wasn't perpendicular to the emitter.

There's no reason you shouldn't be able to see this... as you would see those photons moving across that front that reflected back towards the sensing device (camera)

Re:Oblig. Futurama (paraphrased)

[Surt]

The plural does work that way, though, and the plural form was used, so technically it was correct.

Re:Oblig. Futurama (paraphrased)

[Anonymous Coward]

You can't watch light propagate across your field of vision. What you see from a flashlight is _scattered_ light coming from a beam propagating across your field of vision. Any thing that you sense by vision, or that in instrument detects through photon interactions, is the result of photons hitting your eye or the instrument respectively; that implies that those photons are _not_ traveling across the field of vision.

Related is the quantum measurement problem--anything you do to that photon is likely to perturb its momentum dramatically, so there's not much chance of "watching" a photon propagate at all. Detect where it is one instant, and the next instant it will be on a trajectory unrelated to its previous one.

Not sure who "we" are, but I think you might be over-representing your expertise in this matter. Not being able to see photons propagating isn't a quantum vs. classical issue, it's a problem with how light propagates and what it means to "see" something.

Re:Oblig. Futurama (paraphrased)

[Hentes]

Yeah, light would have been a better word.

No idea what to use it for

[Yvan256]

But you can bet the Mythbusters are going to want one. You know, for science.

Twilight Zone

[Jeremiah Cornelius]

I want to use this thing to finally catch that blowhard McNulty in action!!

Time to watch

[mbone]

Times, at 30 fps, to watch

- a lightning strike move 1 meter : ~ 1 week
- one bullet streak by Neo's head : ~ 100 days
- one boob bounce on Baywatch : ~ 1 century

Better bring lots of popcorn.

Re:Time to watch

[Nationless]

Or put it on a thin wall display and call it art. People can see it change tiny amounts every time they visit and I can guarantee it would be a very cool topic of conversation.

Summary misses the camera's use-case: Chemistry

[gentryx]

The real use-case for the camera is not to watch at coke bottles at super slo-mo, but to investigate how molecules absorb light of different wave-lengths. There is a real scientific need for this camera. And of course, as mentioned earlier, it can't trace individual photons.

ps: needless to say that I did like my own summary much better (for being informative), but that may just be me.

What would you like to see slowed down?

[Keyslapper]

"What would you like to see slowed down to such a degree?"

My 5 year old. I might be able to keep up with him that way ...

Things I'd like to see...

[PortHaven]

- Match light at uber slow speed.
- Bullet impact
- The internals of a cell working
- A plant photosynthesizing

Re:Things I'd like to see...

[PPH]

- A dozen Slashdotters vying for First Post.

Clear Explaination

[trout007]

Here is my take of what is going on.

They want to take a movie on how a light travels through a scene. Just like watching waves in a wave tank.

Here is how they do it.
The fire a laser pulse into a scene.
Then after a certain amount of time in the picoseconds they record the image with a line sensor.
This only records one horizontal line of information
Repeat with a slightly longer delay.
Continue until the light is no longer in the scene.
Rotate a mirror in order to record the next horizontal line in the scene and repeat the above.

Finally combine all of the recorded lines with the same delay into a picture and combine the pictures at different delays into a movie.

So this isn't a high speed camera in the sense that could slow down a physical event like a bullet firing or explosion. It is more like doing stop motion photography.

Re:Clear Explaination

[The Gaytriot]

Most clear explanation I've heard so far, wish I had mod points today.

I wanna see...

[g0bshiTe]

What would you like to see slowed down to such a degree?

Bay Watch running sequences.

Re:I wanna see...

[AaronLS]

I guess it'd be a little like watching grass grow.

What would you like to see slowed down ?

[maroberts]

The exit of money from my bank account...

How do you watch a photon?

[AaronLS]

If a photon hasn't struck your "camera" how do you watch it move across your field of view? I didn't see anything in the article that indicated that had a way to watch photons that are passing through your field of view but do not impact the sensor array.

"The laser pulses, with very complex timing circuitry, are then picked up by an array of 500 sensors in the camera"

This means the camera is acting just as any other camera would. It has a surface that reacts to photons impacting on it as the result of reflected laser pulses. This how our eyes work and any other camera. So there is nothing here that indicates you can somehow watch a photon passing through your field of view.

Re:How do you watch a photon?

[anonymov]

It's just a bad description.

a) Laser shots a pulse
b) At moment T, camera captures reflected/refracted portions of the pulse
c) Repeat for moments T+n

After combining the pictures, you can infer the path and shape of the wavefront.

Re:How do you watch a photon?

[AaronLS]

Exactly. So if the pulse hits something like a surface or fluid and some of the photons are reflected towards the camera, then it will pick those up. It can't actually see the photons that stay on their path and aren't reflected towards the camera.

I recommend one look, and then one more...

[vikingpower]

...at the YouTube video. At the moment you realize what actually is going on here ( you *SEE* a wave front travelling over or through an object ), you gawk and think "how godawfully beautiful". I mean: wow.

Super slow-mo, eh?

[kno3]

What would you like to see slowed down to such a degree?

My maths lectures.

Re:Super slow-mo, eh?

[mbone]

XKCD has a cartoon for this.

"If you think
the minutes in
your morning lecture
are taking a long time
to pass for YOU..."

Warp core explosion

[bjb_admin]

Would that give me time to draw a happy face on a warp core explosion?

Re:Warp core explosion

[Surt]

Probably not. A warp core explosion can actually happen faster than the speed of light, so potentially this is just not fast enough.

What would you like to see slowed down to such a d

[Anomalyst]

Congress and the Oh, bummer administration.

Re:So

[Phrogman]

So lets hear it for the GIIPer !
Couldn't resist :P

CGI

[anonymov]

Computer graphics researchers would surely be excited to see some shots of Cornell box with this thing.

BOOBS

[MBAslug]

Watch a jiggle slowly spread from one to the other

What would I like to see?

[roc97007]

> What would you like to see slowed down to such a degree?

Sofía Vergara, jumping rope, in the nude.

Saw this in SIGGRAPH in August

[fleeped]

This was presented in a talk in SIGGRAPH this year, in August. And was brilliant and inspirational :) Slow Art With a Trillion Frames Per Second Camera

Participating media and more

[fleeped]

smoke, fog, dust, translucent objects, water, and the list goes on. It still collides with stuff, but hey, it keeps going (sometimes almost straight)

What would I want to see slowed down that much?

[bwall]

A money shot of course.

NOT trillion frames a second?

[bradgoodman]

If I understood the video, they are *NOT* taking video at a trillion frames a second. They are continuously pulsing the laser, then snapping a picture at progressively longer periods of time - the net result being an series of images from SEPARATE EVENTS, that would MIMIC high-frame rate video if a single event, ONLY if the separate events are identical (e.g. a laser flashing, each flash illuminating the same object over and over again).

Re:NOT trillion frames a second?

[thisisauniqueid]

See my other comment about this -- http://hardware.slashdot.org/comments.pl?sid=2572024&cid=38366336

What about strange quantum effects?

[mpfife]

What about this: Since they're taking photos from multiple 'flashes' of the illuminating laser over time - conceptually - shouldn't the quantum properties of light bending/scattering be visible?

What if we used this to shoot the standard diffraction grating quantum experiment or other examples of strange quantum properties. Would we see frame-to-frame quantum discontinuities based on when the sampling occurred?

Neutrinos!

[VanessaE]

Couldn't this technology be used to augment the work being done to study those faster-than-light neutrinos? I realize they're hard to detect because of their weakly-interacting nature, of course, but couldn't one construct a similar experiment with both light and neutrinos and watch the two propagate?

"It's not a trillion frames per second"

[thisisauniqueid]

Well assuming the laser pulses are completely uniform (which they are very close to being), then each wavefront of light is mathematically indistinguishable from the one before and after it, so in a very metaphysical sense, you could say that stitching together a video with frames of data taken from successive pulses is absolutely no different in the end than if they had collected all the data in a single shot. However they actually need to aggregate data from millions of frames into a single shot, because they get on average something like half a photon expected per frame without aggregation to increase the SNR -- see the longer presentation here (it's quite a remarkable presentation, I have seen about a third of it so far -- it talks a lot initially about their work to see around corners -- this is not a new thing to try to do, it has been done before with standard camera equipment and a projected pattern with some success, but doing it with laser pulses is new).

The cool thing about what they have done is that you can watch the actual wavefront move like an expanding contact lens through the scene. To my knowledge, nobody has ever seen this before. Sure, it's data from billions of expanding contact lenses, but it shows you in a very visual way that the universe works the way the mathematics say it does.

No, the camera is not a trillion frames per second, but it shows you events that happen over trillionth-of-a-second timescales, if it were possible to capture data that quickly (which it is not) and if it were possible to solve the SNR issues over that timescale (which it is not).

Wish to see!

[KlaymenDK]

What would you like to see slowed down to such a degree?

The Big Bang.

Whoa!

[jman.org]

What would I like to see slowed down to such a degree?

Anything from the U.S. Senate. Those guys move waaaaaay to fast.