Shimadzu’s million-frame-per-second video camera

[Steve]

Thought SLI was fast? Think again. Of course, here we're talking video capture, not rendering, but it's a remarkable framerate nonetheless.

It may only shoot in monochrome at a size of 312 x 260 pixels and take up to 100 frames at a time, but Shimadzu’s HyperVision HPV-1 camera can record at a million frames a second (you can step down through a range of slower rates to a crawling 30 fps should you wish), enabling you to capture those precious split seconds of blindingly fast motion in all the loving detail you need.

[engadget story]

Thanks to Lee for the news tip!

[chrestomanci]

Interesting, but I suspect that high speed film will continue to dominate for most purposes for quite a while longer.

312 x 260 pixels x 8 bit grey x 1 million fps is 81 Gb/second or about 1000 times the speed of a hard disc, and relatively hard to do with solid state storage. (An Opteron HyperTransport bus runs at 22.4 Gb/sec btw) That is why the camera is limited to 100 frames, as that is the size of it's fast internal buffers.

A conventional film based high speed camera can store image data at a comparable rate, but the only limit on how many images is how much film can fit on the spools. The image resolution will probably be better as well.

High speed film cameras work by running the film very quicky past the lens without stoping it for each frame or using a shutter. Instead a high speed strobe flashing at the required rate is used to freaze the action for each frame.

The main limiting factors are:

• How quickly the strobe can flash
• Making each flash bright enought to expose the film
• Making each flash short enough to freaze the action
• The mecanical strenght of the film so it won't break when it is wound at 10,000 m/s

Overall, I think this is interesting technonoly, and probably breaks a speed barrier that is not achevable with film, but the 100 frame limit will prevent it's widespread use.

[Steve]

A very insightful post chrestomanci and I agree with you.

Of course, over time this technology is likely to become more widespread, but it will depend on bigger buffers, faster backing storage and possibly very nippy on the fly compression.