Speed of light

[martin_h]

https://youtu.be/7Ys_yKGNFRQ?si=SRtRX6QEwD4QJVsW

[saac]

The chaotic cavity is mesmerising. Cool video, raises more questions in my feeble mind than it answers lol  

Jim 

 

Edited December 17, 2023 by saac

[Ouroboros]

Fun video. Impressive camera. Bit of a missed opportunity that they didn’t spend a little time explaining how it works.  

[vlaiv]

6 hours ago, Ouroboros said:

Bit of a missed opportunity that they didn’t spend a little time explaining how it works.  

My first impression was that it is fake.

I did not do the exact math, and to be honest, I just skipped thru the video to see what it is all about, but when they started talking about incredible amount of FPS - my gut feeling was to think along the lines:

- in order to record smooth image at certain FPS - you need to have enough photons

- if you capture that much photons in such short amount of time - it means that total flux must be extraordinary.

- That much photons with photon energy at certain level - equals enormous amount of energy.

Gut feeling "calculation" pointed out that it's just too much energy in light to be feasible / real.

[vlaiv]

Ok, here it is again

I just again opened link and title is:

10 Trillion FPS

and two guys stand there and talk about actually recording at 10 trillion FPS.

How many pixels, at which bit depth? Someone care to calculate data throughput of that needed?

Someone care to calculate speed of readout in electronics and how high clock rates do you need to have in order to do that? Then proceed to add the fact that electronics also works on EM field like light and that it takes time for electronic impulses to travel certain distance. With very clock speeds (like in Gigahertz range already) - you have issue with length of your wires. Want to go to terahertz speeds - how are you going to sync left and right side of your microchip when there is not enough time for signal to travel that distance?

[andrew s]

A quick search found this https://www.caltech.edu/about/news/ultrafast-camera-takes-1-trillion-frames-second-transparent-objects-and-phenomena and https://www.popularmechanics.com/technology/design/a32434104/worlds-fastest-camera/

Regards Andrew 

Edited December 18, 2023 by andrew s

[vlaiv]

13 minutes ago, andrew s said:

A quick search found this https://www.caltech.edu/about/news/ultrafast-camera-takes-1-trillion-frames-second-transparent-objects-and-phenomena

Regards Andrew 

On the page you linked:

Quote

The fast-imaging portion of the system consists of something [removed word] calls lossless encoding compressed ultrafast technology (LLE-CUP). Unlike most other ultrafast video-imaging technologies that take a series of images in succession while repeating the events, the LLE-CUP system takes a single shot, capturing all the motion that occurs during the time that shot takes to complete.

So it's a bit like instead of capturing lightning as it progresses you capture this:

and the then do what exactly to get the actual motion? You make things up ...

Not science in my book ....

Edited December 18, 2023 by vlaiv
missing letter

[andrew s]

7 minutes ago, vlaiv said:

Not science in my book ....

Looks like legitimate technology to me. See the second link I cross posted with your reply. Also this on sreak cameras .

Regards Andrew 

[vlaiv]

1 minute ago, andrew s said:

Looks like legitimate technology to me. See the second link I cross posted with your reply. Also this on sreak cameras .

Regards Andrew 

Let me just contrast that with following argument (to be honest, I've look at streak camera and saw this:

and did not bother to read the rest).

So you want to capture light scatter from light propagating in some medium and you capture frames in succession and two frames next to each other are say 1 millimeter apart in sample that scatters).

This simply means that whatever travels at speed of light can cross 1 mm between two frames.

We have scientific sensor. It does not need to be large - let's make it rather small - say 3x2mm in size with handful of pixels.

First frame is read out of those pixels and signal is sent out. That signal can travel at most 1mm before next signal - next frame needs to travel down the same line. We can't have any sort of multiplexing for readout in that case as we would need to put signals from many pixels at the same time in one wire - but we can't have that as whole mm is taken up by one pixel. This means that we would need to have wire for each pixel.

Even if you do that - no recording device can record this information unless there is whole thing per pixel as at some point you need to do some multiplexing. I don't want to even start on issues like interference with such high clock speeds and multiple wires coming from pixels and ultimately - some part of system needs to move slower than speed of light.

We have electrons in potential well of pixels and we have electrons in electronic device that records the signal. Those don't travel at the speed of light.

In that context - streak camera with "sweep circuit" would need sweep to move faster than the speed of light in order to sweep necessary surface in needed amount of time (remember - we are dealing with changes that occur when light moves 1mm or less - how is sweep circuit going to move more than 1mm in time that it takes speed to move 1mm?).

[Ouroboros]

9 minutes ago, vlaiv said:

In that context - streak camera with "sweep circuit" would need sweep to move faster than the speed of light in order to sweep necessary surface in needed amount of time (remember - we are dealing with changes that occur when light moves 1mm or less - how is sweep circuit going to move more than 1mm in time that it takes speed to move 1mm?).

…. And yet they work. When I was using streak cameras 30+ years ago to measure fluorescent decay in a laser-pulse excited samples we had time resolutions of hundreds of picoseconds. Time resolutions are now sub-picosecond from commercial cameras. I can see how this would work one dimensionally by imaging the scattered light from the sample along the slit of the streak camera.  You could get a two dimensional image by using multiple pulses and scanning the laser beam (or imaging system) in the direction perpendicular to the direction of laser propagation.  But I understand there are clever ways to do this single shot too. I don’t think any of this is particularly novel. I’ve been out of the laser game for 15 years and I can recall this sort of thing being done back then. 

[vlaiv]

6 minutes ago, Ouroboros said:

Time resolutions are now sub-picosecond from commercial cameras.

When you say this, does it mean:

- you can record event that lasted sub-pico second with this type of camera

or you are saying:

- you can record event that lasts sub-pico second, every sub-pico second with this type of camera?

[vlaiv]

@Ouroboros

Just to be clear on above post:

First is shutter speed of sub-pico second and second is having FPS that matches sub-pico second.

I have no issues against the first. I just say that second is impossible.

[Ouroboros]

11 minutes ago, vlaiv said:

When you say this, does it mean:

- you can record event that lasted sub-pico second with this type of camera

or you are saying:

- you can record event that lasts sub-pico second, every sub-pico second with this type of camera?

I would say yes to both of those. Here’s a link to the brochure for a Hamatsu streak camera. They show there a picture of a series of sub-ps pulses repeating every 10ps or so. I guess they’ve used an etalon or similar to produce a comb of pulses like that.  So if for example you used a single light pulse of 50 femtoseconds duration to excite some emission in a sample lasting a few picoseconds (say) you would easily resolve that event with such a camera. 

[vlaiv]

12 minutes ago, Ouroboros said:

I would say yes to both of those.

You are sort of right with that statement.

There is only one exposure as far as readout goes, but clever mechanism allows for single row (1D spatial) signal to be shifted in Y direction by change of electric field to get multiple images spread in time.

Not sure how sharp stair stepping can be achieved as it would require very sharp changes in electric field to create fast transition in deflection angle, but I guess it's possible?

It does not explain this however:

How can such device be used to record 3 dimensions? Two spatial and one temporal?

[Ouroboros]

12 minutes ago, vlaiv said:

How can such device be used to record 3 dimensions? Two spatial and one temporal?

Two dimensions (1 time, 1 spatial ) is easy by focusing an image along the slit of the streak camera. The slit is perpendicular to the scanning electrodes in the streak camera. At the back of the streak camera is a phosphor screen attached to an image intensifier. The output of that is coupled to (these days) a cmos camera which provides a two-dimensional image - time in one direction, spatial in the other.

How do you get the other spatial dimension is the question. I can see this can work easily if multiple laser pulses are used. Basically you could scan the subject or laser or imaging system up and down.  However, I believe some of these techniques work with a single laser pulse and I’m not quite sure how that’s done. 

Incidentally, as explained in the video, the image of the figures is provided by a conventional camera. 

[saac]

Tbh this high speed camera technology has been demonstrated before by MIT about 5 years ago. The previous video that I recall showed a high frame rate capture (trillion frames per second) of a single pulse of laser light propagating through a cola bottle and washing over an apple. It had been linked to a camera technology that was able to see around corners being able to capture then render minute amounts of reflected light. 

Jim 

 

Edited December 19, 2023 by saac

[vlaiv]

10 minutes ago, Ouroboros said:

Incidentally, as explained in the video, the image of the figures is provided by a conventional camera. 

Yep, that part is evident.

There is one more detail, that I might have gotten wrong - it took 8 hours to "analyze" the recording?

[Ouroboros]

51 minutes ago, vlaiv said:

Yep, that part is evident.

There is one more detail, that I might have gotten wrong - it took 8 hours to "analyze" the recording?

Yes. I’ve skim read some review papers in this field over the last half hour or so. It’s not immediately obvious how this technique works.  But I think I can sort of see that they use imaged chirped laser pulses to encode wavelength information with temporal information. They then use two cameras - the streak camera and a cmos camera - plus a whole load of processing to get back temporal and 2D spatial information.  The frames per second thing is slightly misleading. It’s more like an equivalent fps. 

Edited December 18, 2023 by Ouroboros

[vlaiv]

18 minutes ago, Ouroboros said:

Yes. I’ve skim read some review papers in this field over the last half hour or so. It’s not immediately obvious how this technique works.  But I think I can sort of see that they use imaged chirped laser pulses to encode wavelength information with temporal information. They then use two cameras - the streak camera and a cmos camera - plus a whole load of processing to get back temporal and 2D spatial information.  The frames per second thing is slightly misleading. It’s more like an equivalent fps. 

They actually use two sets of 1D + temportal information to construct 2D+time, right?

Does scattered light "freeze" frequency at the moment of scatter? I'm trying to figure out how chirp plays into all of that - it is probably combined with diffraction to separate different moments, but I'm failing to see how that would be possible if scattered light continued to change in frequency with time.

[Ouroboros]

1 hour ago, vlaiv said:

They actually use two sets of 1D + temportal information to construct 2D+time, right?

Does scattered light "freeze" frequency at the moment of scatter? I'm trying to figure out how chirp plays into all of that - it is probably combined with diffraction to separate different moments, but I'm failing to see how that would be possible if scattered light continued to change in frequency with time.

There you are - try this paper. Single-shot compressed ultrafast photography: a review.  Bits of it are quite readable. 😀.  I admit that I don’t grasp how the digital micromirror device they mention allows them to reconstruct the spatial and temporal information.  It’s clever stuff though. 

[johannes1]

Imaging with a time resolution of few 10's picoseconds and sometimes even faster is principally possible. There are different techniques. I do not see a direct reference in the video description, but it looks very much like a single-photon avalanche diode (SPAD) camera, similar to this paper:

https://www.nature.com/articles/ncomms7021

SPAD cameras are intriguing, because they can be made in standard CMOS processes and could potentially in a few years be available in a format and cost that might be in reach for amateur astronomy. However, there are some catches. If taking images at say 50 picoseconds/frame, you have a couple of challenges:

1. You need to have some form of shutter that goes at picosecond rate - this is solved by SPADs
2. You have the photon flux problem pointed out above: there will on average be very few photons in the integration interval and thus gigantic shot noise
3. You somehow need to get your image data off the chip into some form of memory before the next frame

In the light-in-flight demo, problems 2&3 are solved by using a periodic process (pulsed laser), and accumulating data over many many pulses. I.e. you synchronise your SPAD camera with the laser, for each individual pulse you only get a few photon detections, and hence little data to deal with, and under the assumption that each pulse has the same effect, you fill your frames bit by bit - timed in reference to your trigger pulse. Not exactly the same as actual live trillion fps, but rather trillion fps for very specific processes.

Whether that translates to astronomy in any way that is not accessible to more traditional cameras, not sure about that but perhaps someone comes up with a clever idea some day. (the exploitation of periodicity makes me think about pulsars, but since they are point sources I am not clear why you would need imaging capability for those)

[robin_astro]

4 hours ago, johannes1 said:

(the exploitation of periodicity makes me think about pulsars, but since they are point sources I am not clear why you would need imaging capability for those)

Imaging the crab nebula and pulsar through a synchronised shutter was good fun  though

http://www.threehillsobservatory.co.uk/astro/astro_image_33.htm