[Brian Haidet] published on his AlphaPhoenix channel a laser beam recorded at 2 billion frames per second. Well, sort of. The catch? It’s only a one pixel by one pixel video, but he repeats it over and over to build up the full rendering. It’s a fascinating experiment and a delightful result.
For this project [Brian] went back to the drawing board and rebuilt his entire apparatus from scratch. You see in December last year he had already made a video camera that ran at 1,000,000,000 fps. This time around, in order to hit 2,000,000,000 fps at significantly improved resolution, [Brian] updated the motors, the hardware, the oscilloscope, the signalling, the recording software, and the processing software. Basically, everything.
One of the coolest effects to come out of this new setup is how light appears to travel noticeably faster when coming towards the camera than when moving away from it. It’s an artifact of the setup: laser beams that reflect off of fog particles closer to the camera arrive sooner than ones that bounce back from further away. Or, put another way, it’s special relativity visualized in an experiment in [Brian]’s garage. Pretty cool.
If you found all this intriguing and would like to know more, there’s some bonus material that goes into much more depth.
One of the best videos I saw in a long time.
Also: He should apply for the abuse-challenge.
Clever use of an oscilloscope for data capture and probably more accurate that he used the same channel for the laser pulse “trigger” rather than a separate channel as he originally intended. I suspect that if I had tried such a poject that I would have likely wasted time and effort attempting to concoct some sort of counter and a sample-and-hold ADC to accomplish the same thing (and likely given up before completion). He’s got some additional videos on his other channel about some other aspects related to the project. It’s a great accomplishment and video!
Whoops. That was meant to be a standalone comment
Interesting procedure he uses.
Take super fast video clip 1px by 1px at a few hundred thousand positions.
Pulse the laser on for each new position. – Start video recording at the same time of pulse for each clip
Stitch videos together and here is the result.
Results look awesome – I assume the speed difference at the end may be related to doppler effect but could also be a focusing situation.
I follow his channel, he has some other really great videos, I am not putting my garage to good use
Wow! Not only it the high-speed of the electronics impressive, and the synchronization of signals (including the actual electron flight-time in the p.m. tube), but getting the physical stability to build up a recognizable image through hundreds of thousands of pixel snapshots is darn near inconceivable. If only Michelson and Morley were alive to see this…
That’s pretty good work. I’ve seen Masters degree projects with less.
I didn’t catch how long it actually takes to grab that image. At the 3000 Hz rep rate of the laser he mentioned you’d think at least 300 seconds, but I don’t know how much averaging he does, nor how long it takes to retrace the mirror, or dump the data from the scope..
But how fast could it go? If you assume (say) 1000 ns as the maximum reasonable range of detection (1000 feet), any photons launched from one pulse would have been absorbed after a few bounces off walls, or long left the vicinity. That’s a million shots per second. The average power of the laser goes up by a factor of 300, but even that’s not too crazy (it you’re not standing in front of it, anyway).
So, it’s conceivable that you could get a 1 frame per second out of this, with a data rate of ‘only’ around 100 megabytes per second. Almost easy nowadays.
It takes about an hour to record one of these frames, and each pixel is made from 1-3 laser flashes (binned by encoder data after recording). Each row is collected with the scope in sequence mode and the data is moved to the computer while the mirror slews back to the start. That data transfer is actually bottlenecking the entire thing. Originally I was blinking at 30 kHz for much cleaner images, but it took like 20-30 seconds to pull each line’s data to the computer. There’s some processing on the scope that happens per captured waveform that grinds the usb transfer rate to a crawl – wish I could just ask for the whole memory buffer raw cause the code already “retriggers” each scan. You can see the transfer rates if you zoom in on the bit where the screen recording of Python is shown
Cool. Thanks for the followup and additional info.
Not cheap, but I guess it would be a pretty good project to showcase a ThunderScope (Thunderbolt scope, supposedly gets to over 1GBps, as it’s just encapsulated PCIe). Who wants a rev3? (?
The University of Illinois Champagne-Urbana could use a person like you. I think the electric engineering or the physics department should send you an offer.
Awesome video!
However, as you know the path the photons took at which pixel at a certain point in time, could you take this into account and compensate for the different path lengths?
Kudos!
Special relativity? What?
It’s basic arithmetic.
1m travel ray + 2m reflection ray = 3m total travel
3m travel ray + 4m reflection ray = 7m total travel
The second ray is 4m longer, thus it appears slower than it should be because the travel distance of 2m is what we are trying to measure for.
If you want to be fancy and talk about it as a system you can go as far as basic trigonometry.
We aren’t talking about reference frames in context.
Only LITERAL “frames” as a reference.
But that’s actually all there is to special relativity! Just the realization that light has a finite maximum velocity, and Pythagoras. The rest falls right out. But as the video notes, it’s hella strange to think of events being non-simultaneous just because they happen at different distances.
I looked around for a good explainer, found this: https://newt.phys.unsw.edu.au/einsteinlight/jw/module4_time_dilation.htm Maybe other people have better ones?
Imagine that he were filming a slow-moving toy car with head and tail lights instead of a “moving” cloud of smoke illuminated by laser. Would you expect the observed forward and backward velocities to be significantly different? What happens when you speed the car up?
Eh, there’s a lot of special relativity that you can think about while looking at this demo but like Ian said, this isn’t a demo of special relativity because all the reference frames are basically stationary with respect to eachother. The laser, the camera, and the cloud are all in the same reference frame. So the light emitted by the laser and the light emitted by the excited cloud are both relative to the same reference frame and there’s really nothing ‘special’ in it. Straightforward classical / Newtonian mechanics explains it well enough.
In the case of the car, the car is moving relative to the camera, establishing two different frames of reference. But the light emitted by its headlights is moving at the same speed in both reference frames, which is clearly impossible in classical mechanics and is the motivation for special relativity.
The key fact is that the light emitted by the cloud is travelling at the speed of light relative to that cloud, rather than relative to the reference frame of the laser light that excited the cloud. It’s possible to contemplate the reference frame of a photon but it isn’t necessary to understand the obvious result.
I need to think this through some more….
I do think that the “light cloud” is moving, and that it’s the headlight in the analogy. The headlight’s beam is the scattered light, here emitted in all directions. In this case, both the cloud and its emitted light are travelling at the speed of light in a garage. (The fog itself is stationary, naturally.) But you can imagine the “light cloud” being an LED if you want — a moving light source.
The light that he sees from the cloud trails the actual cloud’s location at the moment of exposure, because of the finite speed of light and the distance between cloud and camera. He calls this “seeing the past”. That’s the “relativity of simultaneity” logic of special relativity (https://en.wikipedia.org/wiki/Relativity_of_simultaneity) without calling it by name.
Is that necessary for the result, though?
The speedup/slowdown is because of the change in additional propagation time from cloud to camera. When it’s moving toward him, the propagation delay is getting smaller, and it “catches up” with its true position, and vice-versa. So in that sense, it’s just geometry + finite speed of light. And that’s also why it’s negligible when he points the camera perpendicular to the beams.
But at the same time, if the light cloud were an LED, moving at walking speed, there would be almost no change in its apparent velocity whichever way it’s going, because the received image of the LED won’t ever be far “in the past”, and there’s little catching up to do.
The speed of the light cloud is irrelevant for the absolute delay — that’s just the distance traveled by the light it emits. But the speed of the light cloud determines the magnitude of the change in apparent velocity. This smells like redshift / blueshift to me, and that smells like relativity. (Q.E.hand-waveD.)
So I’m still not sure. :)
There’s a whole section in the video from [Brian] which covers this in detail: https://youtu.be/o4TdHrMi6do?t=1450
Of course, when it comes to science it’s much better to measure it in terms of the length of a Tudor King’s foot (Henry VIII) than use actual scientific units; especially when the article has an international audience, 95% of which don’t use them and distance itself is defined in terms of the speed of light in metres/s and the aforementioned Tudor’s foot is defined by metres.
Of course, US citizens reply: if the other 95% of the planet doesn’t use them, they should adapt to us.
lol. In the video [Brian] says he usually uses metric but he likes “one foot” when talking about speed of light because it works out nicely to be about one nanosecond!
There are two kinds of countries…those who have been to the moon, and those who use the metric system.
AlphaPheonix is one of my favorite channels on Youtube, the amount of detail he does makes for a great learning video.
I agree. In the same class is Applied Science: https://www.youtube.com/@AppliedScience
It’s parallax. Since you are filming at an angle relative to the path of the laser, of course the laser looks slower at a distance and faster up close. This is because it covers more relative distance in the filming frame on the close end.
Well…..there has to be a balance, so the slower bottom one going away is sending a beam of Photons at light speed and the light coming to you from the bottom beam is coming off the Photon at the same speed ,but its not where the last point was , so it has to travel further so there is a slight delay. It moves so far, then, Sends the same data back to your camera ,farther away , each time. Then the one coming back, the light has to travel to you, but the light coming off of it coming at you at the speed of light. Einstein asked this question in class. If I throw a baseball at a 100 miles an hr, and then, shine a laser off the ball. Is it going at the speed of light ,plus a 100 miles an hr?
This would be ,speed of light, adding another, speed of light on top of that original speed. Can’t break the law though. So its balanced between that first part going slow ……it takes away. And then adds it back in the return beam.
Cause its not a ,bounce…and bounce….and bounce , no…..when he shot that laser beam off, the light all ready made it to the end, in that instant. This is just the physical path, the photon has to take in our space,and time. For balance. So its one beam …….one long shot. And that is why there can be a slow and fast spot. It was all ready accounted for at the end. Like putting ice cubes in a cup and filling it with water. Water is intelligence, so it all ready adjusted the mass of all that to itself. So if its gonna melt. It wont over flow.
Light does this.
Heat does this.
Sound does this.
Its the Intelligence of nature at work.