I built a 1,000,000,000 fps video camera to watch light move

So apparently buying a high speed camera wasn't enough, because after two videos with it I decided to build my own, but 5 orders of magnitude faster…

In this video I'm filming the motion of light as it flies across my garage at… well, the speed of light! It's fast. So fast that even with my best setup so far, I get 18 frames of video from one end of the room to the other, and those frames have a lot of temporal blur so realistically each "frame" is actually kind of an average of the information that by right should belong to 5-10 frames. It's a mess, but it works.

I'm using the technique from the electricity waves video where I used repeated oscilloscope measurements synced after the fact to produce "videos" of electricity moving down a wire. The only difference is that instead of measuring electricity waves, I'm measuring light emitted by a laser, bouncing off the wall, traveling to my camera, and landing in the window of a photomultiplier tube. UNLIKE the electricity waves video, this setup (thankfully) is automated, and an optics assembly slews across angle space, building up a 3d dataset of video, collecting all the time information from each pixel sequentially.

It's a really fun project that I've wanted to do for a long time, and just recently got pulled together.

Hope you enjoy!

Relevant AlphaPhoenix videos:

Massive thanks to my top Patreon supporters!
birdiesnbritts
John Sosa Trustham
Vladimir Shklovsky
Aloysius Sparglepartz
Jason Whatley
Lohann Paterno Coutinho Ferreira
Jeffrey Mckishen
nothings
Eugene Pakhomov
Glenn Willen
R520
Nick F
Mirko Rener
Chris Connett
Tyler Filla
Miles Freeman
Benjamin Manns
MPG
Seth Reuter
Danny Thomas
Toby T
Lucy Fur
tiaz
David Antoš
Chris Duvarney
Nick Wage
John T
Jack Serrino
Ethan Sifferman
Dit55

Chapters:
0:00 Intro - What's a lightspeed camera?
2:30 Watch light move
3:43 Forming an image from many observations
6:20 Time-of-flight delay
9:52 The light has to turn on fast
14:14 The sensor has to turn on fast
18:23 Photomultiplier tubes
20:11 Building and testing the camera
23:56 Breaking the camera
25:00 Hero shots

Media Credits:

Run by Ethan Meixsell - YouTube Audio Library License

Switched on Carcassi by Brian Bolger - YouTube Audio Library License

Disco Knights by Quincas Moreira - Youtube Audio Library License

Visualizing video at the speed of light — one trillion frames per second
Massachusetts Institute of Technology (MIT)

The Waymo Driver navigating freeways
Waymo

VFX Artists DEBUNK FLYING ORB UFO Videos
Corridor Crew

"In a cave with a box of scraps" meme from Iron Man

[AlphaPhoenixChannel]

Corrections and FAQ in this comment!

0) Stuff I want to shoot with this camera (adding suggestions to list):
Race different transmission lines vs fiber optic
race light in clear tube of air vs clear tube of water
refraction, bending and slowdown of wavefront
patterns with lots of mirrors somehow?; double-slit interference
interference with one beam delayed (really excited about this one); infinity mirror with an object in the middle that would get cloned slowly looking off into mirror-world
short pulse to get "star wars blaster" effect (this might require buying a different driver!)
shine the laser on a fluorescent material
large lens (I have an old projection TV fresnel in my parents' attic I could pull out for this!)
Electric arcs - this would have to be a macro shot, so observing actual light would be impossible at this framerate, but if I could get an arc to form extremely consistently, i bet it doesn't actually FORM at lightspeed - you have to chug some ions around through the air gap. this could be really awesome!
Light solving a maze or shining into an penrose room. flash bulb in fog - like that one nebula light echo timelapse
long stretch of fiber optic
do a time-delay process with the camera and light in the same place to get a TOF depthmap
two more different color lasers to do full-color imaging and/or refraction
One of the standard computer rendering things like cubes in a mirrored room, or that one teapot or something to do “ray tracing” irl

1) This isn't a "stop motion" technique - it's more like making a 1Bfps camera that's 1x1 pixel, then tiling lots of videos next to each other. I did consider using a Pockels Cell in front of a regular camera to try to do stroboscopic stop-motion-style data capture, but I think it would have been more difficult/expensive than what I ended up with.

2) I've gotten lots of great control recommendations - a few really high-precision servo brands, and things with spinning mirrors/lead screws/stuff like that. My first concept for the camera was actually to mount my one pixel behind a static lens on a 3d printer gantry and raster it back and forth to create a "virtual" sensor. I didn't do this because In the garage (a relatively tiny space) i needed a really wide field of view and from the cyanocamera project, I knew maintaining focus at the edges would be impossible. The spinning mirror is really appealing, but "most" of the time is spent with the mirror looking the wrong way, so I'd be multiplying my scan time by 3-4x. it WOULD give a more consistent image though! I'm also about to look up a galvanometer.

3) most of the scans i scaled to take an hour or two. the more averages you take on each pixel, the better you can resolve the "darks" and get illumination in the room that isn't the primary beam. If i only wanted the primary beam and a noisy signal everywhere else, I could easily get away with one laser flash per pixel! (this gets better if I overcrank the diode, but we all saw how that ended)

4) There are a lot of comments asking about the "early" light from the laser, the fact that the beam is traveling at an angle to the camera, etc. if you look, the first half of the beam moves really fast and then it slows down as it starts moving "away" from the camera. this is a sort of doppler effect, and it's an unfortunate limitation of the size of my working area. it's a really cool effect though, and one I want to discuss more in a future video!

5) I DID use fog for the disco ball! in fact my fogger caught on fire so I got bonus fog... That's the only reason the "streaks" were visible.

6) The part about adding an aperture being similar to a "real" camera lens' aperture may be a bit muddled. in a regular camera, the resolution is set by how far apart the light sensitive pixels are on the chip, but in this case, my light sensitive area is probably more than 1 cm^2, so the aperture restricts defocused light but ALSO light that comes in off-angle.

7) How many subscribers do I need to drill a hole in my garage door? let's say 10M - that may or may not happen one day in the future xD

8) a few commenters have very accurately pointed out that the "start recording now" signal should arrive at the oscilloscope AFTER the laser light does - this is correct! thankfully the oscilloscope can back up slightly from its trigger point, and that's how I'm doing it. it really doesn't matter when i see that signal, as long as it's the same for every single flash I can use it to synchronize the data.

9) I don't think there's any simple way to make this camera faster. the scope is 1GHz sample rate, but actually 100MHz for signals, which means ~3ns rise time is all I could hope for. The laser driver ALSO is in the range of units of nanoseconds, and the PMT is also in units of nanoseconds. in order to speed this camera up, I would need a new light, a new detector, and a new data collector - basically a new everything. It'd be better to switch to a legit streak camera at that point. maybe someday.

10) a lot of people are suggesting that to see light move, the camera would need to go “faster than light” but that’s not the case - you can take a video of a supersonic jet with your phone without any part of your phone breaking the sound barrier

11) this cannot be done reasonably with mechanical shutters. If you have the laser going through a 1mm pinhole and you want to slide a shutter out of the way to turn on the light in a few nanoseconds, that shutter itself would need to be going 1mm/5ns = 200km/second = 124 mi/sec

[dl]

Suggestion: I always wanted to see a laser bouncing between two mirrors back and forth zigzagging at a slight angle

[ElectroBOOM]

Yo that was a beautifully complex project!! Well done!

[jackfloof]

23:16 "Why is onenote using 22% of my CPU?" Most real thing ever said in any video ever

[khoury]

The creation of the interference pattern from the double slit experiment would be really interesting to see. Thanks for creating your fascinating channel

[thethoughtemporium]

Absolutely breathtaking! I wonder if you had a coil of bare fiber optic that was just glass with no insulation, if you could see the light looping through the coil

[hamcha]

First StuffMadeHere doing "fake custom lenses", now this! Love seeing all this optical physics content!

[swivel_z]

It's crazy to think, at these speeds, we're seeing the events of like 10 nanoseconds ago, which means there's 10 frames where we don't yet know that the laser is on, but the laser does know

[BPSspace]

The hole in the board at around 4:20 is breaking my brain. Also 19:15 blue tape is ALWAYS a great idea. This is so sick, great work!!

[TastySlowCooker]

4:29 oi, don’t make me watch this 25 times. How dare you make such a smooth edit dammit

[kimlundgren]

Digging the hole out of the box was so good. Didn't give the joke away at all, just played it off like it was a real thing, and the effect was really well executed. Loved it. Cool little nudge to Corridor later on too.

[aubadoir.]

I'm actually crying over the "ACME Wall Hole" bit, I got confused and had to rewind

[incription]

slo mo guys have been real quiet since this one

[Sonic_Shroom]

You fully got me. I was like OK you just actually drilled a whole in your garage door.

[hanklestank]

Man, Alpha Phoenix videos are so goddamn cool.

Sometimes I can’t believe we have someone just hanging out, on YouTube, literally building cameras fast enough to capture the movement of light, or sensors to track the speed of atomic movement in a piece of rebar, systems for displaying electric wave propagation models, and horrifyingly powerful gerrymandering algorithms, and doing it just hanging out in his cave, with a bunch of scraps.

To have someone this smart, exuberant, and good at presenting such wildly disparate topics in such understandable ways (without dumbing it completely down), with some of the coolest homemade contraptions I’ve ever seen is awesome.

So thanks for putting out the best videos on this site!

[Pedro-Pereira]

Frankly one of the best YouTube videos I've seen in a long time! The enthusiasm, the explanations, the outcome, freaking amazing. I just feel like going to my office and start tinkering with stuff right now!!!

[DrArtiePoole]

I just want to thank you for your videos.

I have a PhD in physics and worked on a combination of electrical measurements and ultrafast lasers (<100fs pulse widths) for magnetooptic spectroscopy. Although I usually know the pricniples quite well, I still learn from your videos and find your explanations to be the most clear and concise for many core principles that are missed in the syllabi. I taught first year university students for a short while and they would ask questions that are hard to answer. I would often draw from your videos and point the students to this chanel for a more rigorous explanation. Well done.

p.s. your extra effort in editing in this video was appreciated :)

[markedis]

As a lighting control designer, I found watching the second and third order reflections fascinating

[Mixmum]

That might be the best pitch for Brilliant I've heard. Take note, Brilliant, pay this man more.

[peterhoulihan]

Great video! I figure it deserves a beer.

Two ideas for your camera:

1. Set up a bunch of mirrors so your laser bounces it's way through a maze. Or maybe along a racetrack.

2. See if you can get your hands on a laser illuminated hologram (as opposed to the kind that use regular light) and film it lighting up.