4 ms·

New camera offers ultrafast imaging at a fraction of the normal cost

124 points by wglb 3y ago

billfruit 3y ago
Does it use compressed sensing?
- sp332 3y ago
  No. It's a really fancy mirror.
  - hinkley 3y ago
    pointing at several different sensors, or relying on the speed of light to time shift arrival at the sensor? I still can't tell from the article, but any of those would be a neat trick.
    - sp332 3y ago
      It rapidly points to different places on the sensor, allowing one sensor readout to contain multiple images from different times.
      - ezconnect 3y ago
        So it's like the drum exposure technique but using mirrors on a single CCD or CMOS sensor right?
- hgomersall 3y ago
  Without having read any more than can be gleaned from the article, it sounds like it's using similar principles. That is, the image is transformed into a domain (frequency) that is more amenable to rapidly capturing the necessary information quickly without time gating directly. It's not obvious to me what the setup is, but this type of approach is neat.
numpad0 3y ago
> He realized that rapidly changing the tilt angle of periodic facets on a diffraction grating, which can generate several replicas of the incident light traveling in different directions, could present a way to sweep through different spatial positions to gate out frames at different time points. Is this saying that light through a lenticular lens sheet comes out slightly early/late depending on angles, and the sheet could be used as an array of delay lines with different time constants?
- pyinstallwoes 3y ago
  There seems to be something to that in regards to fresnels lenses and meta materials.
- foota 3y ago
  I don't think so, I think it's more like sweeping an image across multiple sensors over time using the grating (probably via one of the piezo optical effects?) But I'm not familiar with your terminology, so maybe that's what you said :) Edit: Ah, they're using a DMD. The paper is published here: https://opg.optica.org/optica/fulltext.cfm?uri=optica-10-9-1223&id=538060 https://opg.optica.org/optica/fulltext.cfm?uri=optica-10-9-1.... Iiuc when the DMD mirrors are switching between on and off you can get different diffraction patterns while they're shifting, and these different diffraction orders will move the image on the camera.
- dylan604 3y ago
  how could light ever be early?
  - tzs 3y ago
    Suppose you have two points, X and Y, each distance d from some origin O. At time 0, some light is emitted from O towards both X and Y. The earliest it can arrive at either of them is time d/c, where c is the speed of light in a vacuum. If you want the light to arrive at different times you can either do something to make the paths different lengths, or to make the speed of light different on the two paths, or both. You can't make the speed faster than c on either of the paths but you can make it slower by making the light go through something with an index of refraction greater than 1. Now suppose you've got some thing that you have built to process light, and you care about the differences in arrival time of the light to different parts of the thing. There's probably going to be some part of your thing that it makes the most sense to use as a reference, and then describe light arrival times elsewhere relative to that reference point. There's usually no particular reason to pick as the reference the point on your thing that light first reaches. If you do indeed pick a reference that is not the earliest place of contact with the light, then you can have light arrive earlier at other parts of your thing.
beckerdo 3y ago
Photos or it didn't happen!
- araes 3y ago
  The full paper can be found here: https://opg.optica.org/optica/fulltext.cfm?uri=optica-10-9-1223&id=538060 https://opg.optica.org/optica/fulltext.cfm?uri=optica-10-9-1... It has pictures of: Fig. 3. Side-view observation of laser-induced breakdown in distilled water https://i.imgur.com/s4NeSmP.jpg https://i.imgur.com/s4NeSmP.jpg Fig. 4. Front-view imaging of cavitation dynamics https://i.imgur.com/GCyW6xY.jpg https://i.imgur.com/GCyW6xY.jpg Fig. 5. Imaging the laser ablation of single-layer onion cells https://i.imgur.com/K7H0uso.jpg https://i.imgur.com/K7H0uso.jpg As well as setup and configuration
  - dekhn 3y ago
    "cuvette with carbonated drink"
- hinkley 3y ago
  The claim to have photos. They did not seem to have shared them :/
xeonmc 3y ago
So basically it's like a CRT sweeping electron beams across the screen
h1fra 3y ago
4.8 million frames per second is wild. Currently the phantom can achieve I think 1.5M fps in black and white with a "postal stamp" resolution. I wonder what resolution this camera could output but I guess it's probably smaller than < 120px.
- xattt 3y ago
  From Wikipedia: > 1,750,000 frames per second at a resolution of 1280 x 32 It’s just a slit! This is wild it’s even possible.
- credit_guy 3y ago
  It's wild indeed. Maybe even wilder is that a camera capable of taking on million frames per second was developed in 1940. It was called the Repatronic camera [1],[2]. I can't provide a link, but I read in the book "Dark Sun" by Richard Rhodes that they had a 3.5 million frames per second camera when they detonated the first thermonuclear bomb, Ivy Mike, in 1952. I found in a different place that the guy who built this camera was a certain Berlyn Brixner. The wikipidia page for Berlyn Brixner [3] does not mention this detail, unfortunately. [1] https://www.nnss.gov/docs/fact_sheets/DOENV_1136.pdf https://www.nnss.gov/docs/fact_sheets/DOENV_1136.pdf [2] https://en.wikipedia.org/wiki/Rapatronic_camera https://en.wikipedia.org/wiki/Rapatronic_camera [3] https://en.wikipedia.org/wiki/Berlyn_Brixner https://en.wikipedia.org/wiki/Berlyn_Brixner
- davidhyde 3y ago
  They should probably have indicated how many frames they can capture too. > The team created a DRUM camera with a sequence depth of seven frames, meaning that it captures seven frames in each short movie. Seems like their movies could possibly only be 7 frames in length so not really comparable to Phantom type cameras which can capture a lot more content. Still, it’s great to see the mirror tech used in DLP projectors having other uses. Some 3D resin printers use it too but that’s still projection I guess. I think those dlp mirror chips are amazing and probably fascinating to hack around with.
  - h0l0cube 3y ago

124 points by wglb 3y ago

billfruit 3y ago

Does it use compressed sensing?

sp332 3y ago
No. It's a really fancy mirror.
- hinkley 3y ago
  pointing at several different sensors, or relying on the speed of light to time shift arrival at the sensor? I still can't tell from the article, but any of those would be a neat trick.
  - sp332 3y ago
    It rapidly points to different places on the sensor, allowing one sensor readout to contain multiple images from different times.
    - ezconnect 3y ago
      So it's like the drum exposure technique but using mirrors on a single CCD or CMOS sensor right?
hgomersall 3y ago
Without having read any more than can be gleaned from the article, it sounds like it's using similar principles. That is, the image is transformed into a domain (frequency) that is more amenable to rapidly capturing the necessary information quickly without time gating directly. It's not obvious to me what the setup is, but this type of approach is neat.

numpad0 3y ago

> He realized that rapidly changing the tilt angle of periodic facets on a diffraction grating, which can generate several replicas of the incident light traveling in different directions, could present a way to sweep through different spatial positions to gate out frames at different time points. Is this saying that light through a lenticular lens sheet comes out slightly early/late depending on angles, and the sheet could be used as an array of delay lines with different time constants?

pyinstallwoes 3y ago
There seems to be something to that in regards to fresnels lenses and meta materials.
foota 3y ago
I don't think so, I think it's more like sweeping an image across multiple sensors over time using the grating (probably via one of the piezo optical effects?) But I'm not familiar with your terminology, so maybe that's what you said :) Edit: Ah, they're using a DMD. The paper is published here: https://opg.optica.org/optica/fulltext.cfm?uri=optica-10-9-1223&id=538060 https://opg.optica.org/optica/fulltext.cfm?uri=optica-10-9-1.... Iiuc when the DMD mirrors are switching between on and off you can get different diffraction patterns while they're shifting, and these different diffraction orders will move the image on the camera.
dylan604 3y ago
how could light ever be early?
- tzs 3y ago
  Suppose you have two points, X and Y, each distance d from some origin O. At time 0, some light is emitted from O towards both X and Y. The earliest it can arrive at either of them is time d/c, where c is the speed of light in a vacuum. If you want the light to arrive at different times you can either do something to make the paths different lengths, or to make the speed of light different on the two paths, or both. You can't make the speed faster than c on either of the paths but you can make it slower by making the light go through something with an index of refraction greater than 1. Now suppose you've got some thing that you have built to process light, and you care about the differences in arrival time of the light to different parts of the thing. There's probably going to be some part of your thing that it makes the most sense to use as a reference, and then describe light arrival times elsewhere relative to that reference point. There's usually no particular reason to pick as the reference the point on your thing that light first reaches. If you do indeed pick a reference that is not the earliest place of contact with the light, then you can have light arrive earlier at other parts of your thing.

beckerdo 3y ago

Photos or it didn't happen!

araes 3y ago
The full paper can be found here: https://opg.optica.org/optica/fulltext.cfm?uri=optica-10-9-1223&id=538060 https://opg.optica.org/optica/fulltext.cfm?uri=optica-10-9-1... It has pictures of: Fig. 3. Side-view observation of laser-induced breakdown in distilled water https://i.imgur.com/s4NeSmP.jpg https://i.imgur.com/s4NeSmP.jpg Fig. 4. Front-view imaging of cavitation dynamics https://i.imgur.com/GCyW6xY.jpg https://i.imgur.com/GCyW6xY.jpg Fig. 5. Imaging the laser ablation of single-layer onion cells https://i.imgur.com/K7H0uso.jpg https://i.imgur.com/K7H0uso.jpg As well as setup and configuration
- dekhn 3y ago
  "cuvette with carbonated drink"
hinkley 3y ago
The claim to have photos. They did not seem to have shared them :/

xeonmc 3y ago

So basically it's like a CRT sweeping electron beams across the screen

h1fra 3y ago

4.8 million frames per second is wild. Currently the phantom can achieve I think 1.5M fps in black and white with a "postal stamp" resolution. I wonder what resolution this camera could output but I guess it's probably smaller than < 120px.

xattt 3y ago
From Wikipedia: > 1,750,000 frames per second at a resolution of 1280 x 32 It’s just a slit! This is wild it’s even possible.
credit_guy 3y ago
It's wild indeed. Maybe even wilder is that a camera capable of taking on million frames per second was developed in 1940. It was called the Repatronic camera [1],[2]. I can't provide a link, but I read in the book "Dark Sun" by Richard Rhodes that they had a 3.5 million frames per second camera when they detonated the first thermonuclear bomb, Ivy Mike, in 1952. I found in a different place that the guy who built this camera was a certain Berlyn Brixner. The wikipidia page for Berlyn Brixner [3] does not mention this detail, unfortunately. [1] https://www.nnss.gov/docs/fact_sheets/DOENV_1136.pdf https://www.nnss.gov/docs/fact_sheets/DOENV_1136.pdf [2] https://en.wikipedia.org/wiki/Rapatronic_camera https://en.wikipedia.org/wiki/Rapatronic_camera [3] https://en.wikipedia.org/wiki/Berlyn_Brixner https://en.wikipedia.org/wiki/Berlyn_Brixner
davidhyde 3y ago
They should probably have indicated how many frames they can capture too. > The team created a DRUM camera with a sequence depth of seven frames, meaning that it captures seven frames in each short movie. Seems like their movies could possibly only be 7 frames in length so not really comparable to Phantom type cameras which can capture a lot more content. Still, it’s great to see the mirror tech used in DLP projectors having other uses. Some 3D resin printers use it too but that’s still projection I guess. I think those dlp mirror chips are amazing and probably fascinating to hack around with.
- h0l0cube 3y ago