I just happened to see it on that blog, and wanted to quote its phrasing (which doesn't appear on the video's own page as far as I can see). Sometimes a sentence adds a lot.
This is a clear violation of HN guideline. HN guideline is: "Please submit the original source. If a blog post reports on something they found on another site, submit the latter."
He's obviously new here, so why not cut him a break.
luckylemon, it's true that there's a guideline regarding blog posts that link to other material. You can see of the guidelines by following the link at the bottom of the page [1]. Thanks for contributing something you feel is interesting.
I don't really buy this idea that it's not "really" a camera.
Think of it like this: A still camera has a number of small receptors. When you take a photograph, you wind up with a bunch of numbers that average the amount of light that hit each one. It's not a perfect record of the light that was really there-- does that count as a "real" picture? A video camera takes a series of slices at particular moments in time. Information about events that occur between those slices is lost, either blurred out of existence or ignored entirely. Does that count as "real" video?
So, this camera can record a single line once at a precise instant in time. If we want to turn that into a moving picture, we can either parallelize the camera, or serialize the event. So let's say we put together a hundred million of these cameras using mirrors and amplifiers and whatever else we need to record video of a single pulse of light happening once. Let's say we also use a single camera to record video of a hundred million pulses of light which we know will have a variance of less than the resolution of the camera. Let's say we crunch the data from both of these "cameras", and they come out as identical.
Which one is the "real" video?
Anyway, there's a much more salient reason why this wouldn't work for a moving object: As the man pointed out, at this rate, even a speeding bullet would not move perceptibly.
I could be wrong, but as long as the object is moving at "normal" (e.g. significantly less than the speed of light) speeds, this approach would work for moving objects too.
The "video" of the pulse of light going through the bottle is actually the result of sending trillions of light pulses through the bottle, taking pictures at very precise (recorded) times and syncing them all together.
They are not sending trillions of light pulses, otherwise they'd be making a very long movie. If they want to do a 10 second video, they just need to send 240 pulses.
They send a pulse and take some kind of picture at some particular point in time. If you watched the video you'd know that these pictures are taken so quickly that the exposure is incredible low - this means the image is incredibly dark.
They repeat this process of taking some kind of picture at some point in time trillions of times - this produces the terabytes of data that was mentioned in the video.
From these terabytes of data, which are essentially incredibly under exposed (dark) pictures of a bottle they use computers to add them all together based on their time and produce a video (which is actually trillions of different pictures added together to compensate the massive amount of under exposure).
Note, the "time" of the picture is related to when the pulse of light was "released".
right but doesn't this depend on how fast you can send and receive those trillion light pulses? if that can be done in sufficiently short time, the object in motion will have only moved by a small amount, and it will look like it has been elongated in the direction of travel, like a motion blur.
The above is my uninformed and optimistic speculation.
Nope, because a single frame of the video is the result of many trillions of light pulses.
Each time a picture is taken the exposure is so low that the image is incredibly dark. These means they must take lots more and "add them together" to produce a bright image.
If the object is moving it would be impossible since every new image would have the object in a different place, and when you added them together to counteract the fact each image is so under exposed (dark) the object would be blurred.
Right. But the point of high speed photography, like the apple bullet that is disingenuously compared in the video, is that you avoid motion blur. My 200 fps camera can take a trillionth of a second snapshot, with some, um, motion blur
Why does the researche call it "trillion fps", when it is really 1/trillionth sec frame? That sort of misrepresentation is supposed to be saved for the news office or the popular press.
The achievement is interesting enogh without the deceptive exaggerations.
This is a clear violation of HN guideline.
HN guideline is:
"Please submit the original source. If a blog post reports on something they found on another site, submit the latter."
So instead of talking about the content, we are all discussing the way that he posted his content.
Who cares? One click, two clicks. It really doesn't effect me that much.
I think all of us are talking way too much much about how much HN needs to be fixed. Almost every large thread has a reference to how HN is broken. HN may or may not be "going downhill" but if that's all we talk about... It certainly will.
Again, it's NOT trillion frames per second. Stop spreading this bullshit.
That camera takes extremely short exposures (1 trillionth of a sec), but it cannot take 1 trillion of such exposures in one second. Not even close. Not one billion, not one million (and probably not even one thousand, as there are only a few sensors that can do that).
They capture data at about half a trillion "fps". Then they repeat this over and over and combine the data - as the light from just one exposure is too weak.
This just shows you don't understand how their technology works. They don't capture the data at half a trillion FPS. They take very short exposures and shift the phase very precisely, which creates an illusion of 1 trillion fps.
