How to measure milliseconds mechanically for camera shutter calibration [video] | Hacker News Reader

How to measure milliseconds mechanically for camera shutter calibration [video]

youtube.com

126 pointsChuckNorris8925 comments

Very inventive, electronics makes things so much easier. It reminds me of the high speed mechanical cameras used in the 1950s Atomic tests that can take 200K - 10M images per second (although not for very long).

https://collection.sciencemuseumgroup.org.uk/objects/co83584...

https://gizmodo.com/filming-monster-atomic-blasts-requires-m...

> At these high framing rates the recorded event must be illuminated by an intense source of light. Just like an atomic flash during the first milliseconds of a nuclear bomb test

aside from the light requirements, how long does it take to roll out a 1000' of film at those frame rates? at 24fps, that's typically 10-12 minutes. 200,000 / 24 = 8,333x faster. 10min x 60s = 600s / 8333 = 0.07s

i'm guessing the military could afford to have a different type of film load than standard motion picture film rolls, but that's still ridiculously fast at 200k fps so maybe 10M fps qualifies for ludicrously fast. what fps would need to be achieved before the picture just goes plaid?

I love when a difficult engineering task has a built in solution to a sub-problem. You want to capture an image of a nuclear blast immediately after the explosion for only a minuscule time slice. That means you’re worried about getting a viable exposure because the film’s only under light for microseconds. Solution: nuclear bombs are really fucking bright.

The way this technique combines knowledge of mechanical and human optics is really fascinating.

Of course the calibration could also be done by recording an image to film and developing it, but it would be very slow and inconvenient. This technique is literally using the film-like qualities of the retina, which has just enough permanence to record the required pattern. Genius.

Film isn't linear the way digital is. 30% under exposure doesn't give an image that is exactly 30% brighter, but may show up more as less shadow detail, compressed mid tones, etc. Also, negative film can handle a huge amount of over exposure and still look good. You can get a better idea with positive/slide film, which is much more sensitive to exposure, but that is very expensive.

Slower shutter speeds (1/30 second and lower) ar usually the speeds that are the most incorrect with old shutters. That can be tested more or less reliably by filming it with a high speed camera (I use a Fuji XT3 at 120 fps). You can also measure the speed by recording the sound of the shutter.

Filmomat have a free app for measuring shutter speeds using audio, as well as a light sensitive doodad that you plug into a smartphone, which is accurate to up to 500th of a second. https://www.filmomat.eu/shop/photoplug

I think what they mean is exposing a frame with the stroboscope-and-shutter combination, not just shooting anything at a known light level.

So the same thing as here, but instead of shooting with the back open, doing it with film inside and back closed.

Something fairly similar, called a strobe tuner[1], is used for tuning musical instruments. (Before electronic tuners, these were more popular, but they're apparently still used.)

There's a mechanical spinning disk that turns at a precise rate. It has patterns on it, and the sound of the instrument controls the strobe light.

Your goal is to adjust the pitch of the instrument until the wheel appears to stand still.

Here's a video about how that works: https://www.youtube.com/watch?v=U8KHeyT6xB0

That video claims accuracy as good as 1/10th of a cent.

There are 12 notes in an octave, and a cent is 1/100th of the distance (in frequency) between two adjacent notes. If it can tune an A440 within 1/10th of a cent, then that means it can get the frequency within the range between 439.975 Hz and 440.025 Hz[2].

If I've got the math right, that means it's measuring way more precisely than a millisecond.

---

[1] https://en.wikipedia.org/wiki/Electronic_tuner#Strobe_tuners

[2] 440 * 2^(-1/12000) and 440 * 2^(1/12000), respectively.

I have something like this built-in to a modified cigarette pack guitar amplifier. I have two LEDs wired in a circular-series circuit (I have the positive of one attached to the negative of the other, repeat for the other end) and I have those on the speaker output in parallel with the speaker. I can watch the harmonics visually via the LEDs when I play and tune the guitar.

What I didn’t understand in the video is, it appears to an assume the rotation of the light source box is known/constant/exact.

How is the light box calibrated?

He shows that between 9:50 and 10:50 in the video

https://youtu.be/Q07CTj4fUeY?t=592

But doesn't that rely on having a strobe flashing at the specific frequency - I thought modern record players do anyway.

I'm not sure how this worked on the 1920s machine - unless it assumed 50H mains frequency, or have I missed something?

Yeah, I would assume they're using mains frequency as the reference. It was a pretty common thing back in the old days. I remember reading one electrical system engineer talking about how they had to tweak the grid hz in order to make sure that peoples' clocks stayed fairly correct. The grid would typically slow down a bit during times of peak load, so they'd run a bit faster at other times to force clocks to catch up. I think most of those clocks are gone now that there are much better sources, but they were still fairly common when I was a kid in the 90's (though I didn't learn about how they worked until much later).

No, many mains connected clocks still use the power frequency for timing. As recently as four years ago we had the same problem: https://www.nationalgeographic.com/science/article/european-...

I wonder where they are still in use? In my world, NTP is king, but there are also radio transmitters in various parts of the world that send out time signals, and satellite navigation systems that broadcast time so accurate, it can be used as a stratum 0 source for NTP. Any time I've checked my phone, it's had essentially 0 error that a human could perceive, and I assumed it was getting that from the cell network. Even my decorative/fashion quartz wristwatches only drift a few seconds a month at most. In my world, tech has certainly moved on, and I'd assumed that was the way for most people too.

I have personally only experienced it with bedside alarm clock radios, and large wall clocks e.g. in schools.

Yep, I was thinking of the bedside clocks, but my phone does that job for well over the last decade now. When I was a kid in the 90's, the school clocks were already centrally controlled - you could see them all change in unison sometimes as they corrected them. I'd imagine they might have generally used the 60hz reference and then been corrected as needed, but I wouldn't consider that quite the same thing.

It's also interesting to look at Swiss railway clock homages. The design is iconic and quite readable, but the more interesting ones have what are marketed as "stop to go" movements that mimic the actual Swiss clock system. They tick as you would expect, but when the second hand reaches 12 o'clock, it pauses, and the minute hand advances one full minute. It's a strange effect, and then of course you start asking about what happens during that pause and how it was actually counting time around that dial (it would have to be a bit fast, wouldn't it?). From my limited understanding, this is because the system synchronized every minute at the top of the minute, and the system had been around since at least mid-century. Obviously it's just an amusement on a wall clock or wristwatch.

Using an iPhone with continuity camera as webcam for work. Can't remember the exact reason why the iPhone can not receive network time in my setup but it effectively can only rely on its local circuits to meassure time. Time is ten minutes off just after a few weeks. It really blew my mind when I first saw this.

It's cheap (unlike NTP in a non-internet-connected device and a custom radio circuit), works anywhere where a mains-connected device works (unlike the radio stuff which may not work in a basement/concrete building, or if exported to a different country on the same grid), is unlikely to stop working in the next decade, and is more accurate than quartz over the long term due to the compensation games utilities play.

Hmm, but how do they make sure the strobe is strobing at the right frequency?

Most likely the 50/60hz mains frequency?

As someone else said, the trick is that you can use a slower source to line up the rotation of a faster source to high accuracy. The idea is similar to that of the Vernier scale.

See also the way the speed of light was originally measured: https://en.wikipedia.org/wiki/Fizeau%E2%80%93Foucault_appara...

I was using the same technique back in the day when I was maintaining my Pentax MXs but using a fluorescent lamp as a source - they blink with the AC power cycle - 50 or 60Hz depending on your country.

Part of what makes Destin's videos so good is his obvious and pure love of learning. Do I care how to measure a microsecond of a mechanical shutter? Not really. Even though I'm a photographer, who cares, that tech is long gone. But he is just so excited to learn about these things its truly infectious.

Also helps that he's covered some awesome topics.... like Coast Guard jetboats, nuclear submarines and Saturn V rockets.

Reminds me of trying to take pictures of CRTs with a smartphone camera, setting the shutter speed to 1/60 of a second to capture (just about) a single frame's worth of exposure even if the camera shutter and CRT scanline aren't lined up. I didn't know that mechanical rolling shutters accelerated along the path, rather than sweeping linearly across the image.

Reply on news.ycombinator.com