The Official High-Speed Camera Thread

[Hotwired]

I've tried 1/40,000 bursts and 1000fps clips on the flourescent bulbs I've got and I can't see much of a flicker in them.

Which is good news for using them for illumination for high(ish) speed clips.



In other news, my ZR-100 appears to have crashed on the following settings in manual:

ISO 3200
Aperture F5.9
Shutter 1/40,000
Burst speed 40fps
Burst of 30 frames

Had to pull the battery as all controls were dead and the LCD was frozen on the lightbulb image it was taking. It rebooted ok after it reset the extended lens.

Not a good thing to happen only only the second day.

[Insomniac]

Rag, I'm not sure the distortion of the die in that image is what's actually happening. I can't remember exactly which camera you have, but if it's a CMOS sensor then it most likely has a rolling shutter. Those fragments would probably have been spinning pretty quickly after impact and this would make them look very distorted when photograped with a CMOS camera.

That's my $0.02, anyway.

[Hotwired]

EX-FH100 and all the highish speed cameras in here have CMOS sensors.

[Insomniac]

Ah, thanks for clearing up what camera he's using... I knew that all the Casios have CMOS, but I wasn't sure if he'd shot the stills with a Casio. Didn't want to make a fool of mine self.

[Ragnarok]

Rag, I'm not sure the distortion of the die in that image is what's actually happening.

Possibly. It must have been spinning (the 5 was away from the camera before the shot), but I'm not convinced.

At 1/20,000th, even the 240 m/s pellet would have only shifted 12mm in the exposure. Given that the fragments will be travelling considerably slower, I can't see the degree of distortion can represent movement alone - it would have had to be "rolling" slowly (in relative terms, of course) for that much distortion - which, I think would mean that despite a 1/20,000th exposure, it took a few times that to actually read all of it.

I guess we need to find a way to test the camera with something very fast we know isn't distorted. The first idea that comes to mind is taking a Dremel disk, drawing a straight line on it, then taking a picture as it spins. Depending on how straight the line comes out in the photos, we've then got a reference for the scan speed of the camera.

Unfortunately, I'm now away from home, so I can't try it for a few weeks, but I'll get around to it eventually.

[Insomniac]

That sounds like a great idea. I'll be waiting on the results haha.

[ramses]

Nice. Did you ever consider using burst mode with as long of a shutter speed as possible (or a simple long exposure) and use a flash triggered by a laser or sound to freeze the motion?

As far as the camera freezing, my first EX-F1 did that, too, in standard definition. It may be that the card is too slow. I would recommend at least like a class 6 or class 10 card.

[Technician1002]

I've tried 1/40,000 bursts and 1000fps clips on the flourescent bulbs I've got and I can't see much of a flicker in them.

Traditional magnetic ballasts provide lots of flicker. Many of the CF bulbs have a DC supply and high frequency inverter running above 10KHZ so flicker rate would be at least 20 flickers per frame in a 1,000 FPS camera. The high frequency is to eliminate visible flicker and keep them silent.

Because of that they are suitable for high speed photography under 10,000 FPS.

This video shows the starting of a box of the lamps. They flicker for a short time while the DC supply comes up to full voltage.

This was shot at 2,000 FPS.

[youtube][/youtube]

[Hotwired]

As far as the camera freezing, my first EX-F1 did that, too, in standard definition. It may be that the card is too slow. I would recommend at least like a class 6 or class 10 card.

No, the card should not be limiting the camera.

The newer cameras have especially vague wording about recommended cards, simply saying: use a fast card and it's not our fault if you don't get a fast enough one...

The F1 manual gets around to mentioning actual numbers, it says the maximum write speed should be at least 10MB/s and I've read of Casio support saying you're unlikely to need above a class 6 card for their highish speed cameras.

I've got a 32GB Patriot Signature LX Class 10 (all that means its guaranteed to transfer faster than 10MB/s).

Benchmarked it using CrystalDiskMark on sequential read (19.39MB/s) and sequential write (17.01MB/s).

[Hotwired]

Rag, I'm not sure the distortion of the die in that image is what's actually happening.

Possibly. It must have been spinning (the 5 was away from the camera before the shot), but I'm not convinced.

At 1/20,000th, even the 240 m/s pellet would have only shifted 12mm in the exposure. Given that the fragments will be travelling considerably slower, I can't see the degree of distortion can represent movement alone - it would have had to be "rolling" slowly (in relative terms, of course) for that much distortion - which, I think would mean that despite a 1/20,000th exposure, it took a few times that to actually read all of it.

I guess we need to find a way to test the camera with something very fast we know isn't distorted. The first idea that comes to mind is taking a Dremel disk, drawing a straight line on it, then taking a picture as it spins. Depending on how straight the line comes out in the photos, we've then got a reference for the scan speed of the camera.

1000fps recording mode is insufficient to capture any kind of line on a Dremel disk even at lowest speed setting. You get a faint triangular blur.

With burst mode on 40,000ths of a second shutter speeds and high ISO you get a much better picture.

Here were the disks used:



The polishing wheel at lowest speed on the Dremel. At higher speed you get tighter banding.



The cut off wheel with a paper disk fixed to it, also lowest speed, note the retaining screw in the middle:



Lighting was a 32W fluorescent daylight bulb.

[al-xg]

Wow I'm slightly confused by that.

[Hotwired]

It's debatable whether I should have put them as a .gif but on the other hand you have 30 cropped and scaled images in each to look at.

[al-xg]

I'm failing to understand the horizontal marks, I've just tried taking a picture of my CPU fan using the same method, although with a much slower shutter speed (1/1000 as it is dark and I only have a low powered bulb at my disposal), and I just see the blades rotating frame by frame.
In my case the fan is obviously slow enough that only a faction of the rotation is performed between every frame.
But I am still failing to understand why you have horizontal lines and not just randomly spaced radial lines.
Is this to do with the way the sensor captures the picture ?
Also did you rotate the pictures in order to make the line horizontal ? Surely you weren't just lucky enough to start the shooting just when the line was at that angle ?

[Technician1002]

The distortion is indeed due to the way the sensor works. In a CCD sensor, the device is exposed to light and due to light levels current flows in individual pixels charging a set of capacitors. The voltage is then removed from the sensors stopping the current and the "charge remains. The charge is stepped out by walking it cell to cell to the adjacent cell untill all the cells are read. This is how a CCD captures a frame and couples the charge for all the pixels out.

A CMOS sensor is like a memory and all the bits are "addressed" instead of coupled to the edge of the device. In normal speeds, much like the CCD, the voltage is applied and light forms the current in relation to the intensity. The voltage is removed and the cells are addressed and read much like a memory chip.

In high speed mode, the time between frames is not enough to provide enough current to make a significant voltage between frames, so in HS mode, the voltage remains ON. As the bits are read, the cells are erased and begin recording light for the next frame. Since the frame is addressed line by line, the entire frame is recorded at different times dependent on when the cells were addressed in the continuous read operation

In the high speed motion, this continuous scrolling read means each bit of each frame was recored for a different time. Because the time is different, the motion blurr shows the object (the dremel in the example) in different positions throughout the frame resulting in the lines appearing bent.

Moving slit focal plane shutters in conventional photography also exhibit this distortion. Some creative artists use the very slow scan of a flatbed scanner to make artistic scans of objects. To see this exact same distortion, place an analog mechanical clock face down on a flat bed scanner and scan it while the clock is running. Use a large clock to it takes lots of time to fully scan the clock. You will see the bent second had due to motion during the scan.

If you lack a clock, place a CD on the scanner and manually rotate it slowly during the scan.

Example of moved photo during a scan.

Another cool scan.

[Hotwired]

Nothing was rotated. The only things done were to crop and shrink the images.

I've just been taking a few shots on a 12v computer fan being driven by 18v under the same 32W florescent daylight bulb and yes I do get more or less frozen images at about 1000th/s shutter speed.

However at high shutter speed the distortion is extreme.





Probably best if I don't animate the fan sequences, if you take images at 40,000fps for 1 second but only show 30 out of 40,000 images... the animation can be a bit misleading. For example I was seeing the logo on the fan reversing every single frame.