View Full Version : DIY high speed video capture (1000fps)


David Bythell
October 16th, 2007, 07:33 AM
I'm new to the forum - and realise that this is probably a dumb question, however I feel I'm out of options after a long search, and need some practical advice to consider if its worth looking into further. Any help would be appreciated...

I have a budget of £500 and would like to build/buy a high-speed colour video camera (720fps or above) of reasonable quality (miniDV and above) to shoot short 3sec clips (or less). I'm willing to put in all the graft it needs to build it or to process the output. I don't mind lugging round a laptop, or lumpy custom solutions, as long as it remains reasonably 'portable'.

Renting is not an option because I need time which is $$$ and would outweigh the costs of buying one outright.

My current attempts are with homemade film cameras and a basic miniDV at 25fps with lots of editing - which looks horrible when slowed down, and doesn't have the granularity that i need. I've thought about shooting a subject simultaneously with multiple video cameras to achieve higher frame rates, and also building a camera from the ground up with CMOS/CCDs - but I now feel that these solutions are not really practical.

I'm willing to compromise on any of the requirements if a solution almost meeting these needs is possible - e.g. sacrificing quality or colour, or upping the budget slightly for bigger gains. I looked at a cheap and cheerful Sony (i cant locate the model) that was suitable for slowing things down by 30% or so for golf and sports analysis (250fps?) - but I need something at least 3 times slower.

I'll consider any solution and appreciate any input!

cheers

Dave

Daniel Ross
October 16th, 2007, 12:39 PM
Sounds cool, but I don't know if you'll find anything on that budget. Not much is made like that.

Cole McDonald
October 16th, 2007, 02:36 PM
if you shoot ntsc at 60i, you can convert that to 60fps reasonably well...it's documented around this site fairly extensively...just do a search for slow motion or 60p. This solution will get you a 50% reduction of speed. and can be had for off the shelf budget.

Jim Andrada
October 16th, 2007, 02:43 PM
Only 1000?

About 40 years ago we succeeded to capture over 1,000,000 frames per second.

Now the real story. I was working in an explosives research group at the time and they were trying to film the combustion of gunpowder. And they succeeded.

This of course was a military research facility and the trick was to use a rotating prism and stationary film arranged in an arc around the axis of the prism. Technology does march on however.

I think you'd need significantly more than 500 pounds - maybe a couple of more zeros on the end would do it though.

http://www.photron.com/content.cfm?n=products&id=66

http://courses.ncssm.edu/hsi/splashes/video/video.htm

Marcus Marchesseault
October 16th, 2007, 05:04 PM
I think there might be cameras made for recording manufacturing processes that may be close to what you want. You might want to check the "alternative imaging" forum since some of those guys are converting industrial cameras into HD camcorders. They would know if there are some intended for high speed. I do know that they can do variable framerates but I don't know the upper limit.

Josh Laronge
October 16th, 2007, 05:45 PM
It's not 720fps but 300fps is closer:

http://www.dpreview.com/news/0708/07083101casio300fps.asp

I don't know anything about the above camera however, sports still photographer Robert Hanashiro linked up 3 10fps cameras shooting them 1/3 of a second off each other to get 30fps. Perhaps you could do something like this.

http://www.photo.net.ph/blogalicious/2007/09/24/how-to-document-record-breaking-feats-30-fps-using-3-canon-mark-iii-and-3-600-f4-lenses/

Boyd Ostroff
October 16th, 2007, 05:55 PM
Thread moved to alternative imaging.

Jim, your story reminds me of my childhood when my father was responsible for the pyrotechnic systems in several McDonnell Douglass projects, including the Mercury and Gemini spacecraft. They used Fastax cameras to capture high speed footage of explosive bolts and flexible linear shaped charges. Some pretty cool stuff :-) Here's a link to one that shoots 5600 fps but I think other models went up to 16,000 fps - http://owyheesound.com/fastax.html

I saw some specialized cameras at NAB for high speed video, and the demo reels were really impressive. Building your own would be a challenge for sure... good luck David. I'm sure the folks here at Alternative Imaging will have some ideas for you.

David W. Jones
October 16th, 2007, 08:22 PM
David, you have unreal expectations shooting over 700 fps on a budget of £500. Your lighting alone would exceed that.

Jim Andrada
October 16th, 2007, 09:28 PM
Boyd,

We were working on optimizing the extrusion shapes for the propellant used in 16" Naval Rifles on battleships. It was somewhat old tech at the time, but we also worked on designing shaped charges to separate the front of an aircraft so it would float around if the jet flamed out on launch from a carrier.

I was working for the Naval Weapons lab in Virginia as a computer programmer at the time (1962) I seem to remember that the fastest camera of the time could capture at the rate of 6 million frames per second, but obviously only for a small fraction of a second

Take Vos
October 16th, 2007, 11:47 PM
David,

I am reading the technical manual of the Allied Vision Technologies Pike F-032 which is a 640 x 480 industrial color camera with firewire 800 interface. The higher resolution cameras in the Pike range are slower.

There is a table on page 213 that gives us the frame rate as a function of image height. Changing the width does not help to speed up the the read-out of the CCD.

640x480 208 fps
640x300 314 fps
640x240 372 fps
640x150 551 fps
640x120 640 fps
640x60 941 fps
640x30 1358 fps
640x10 1778 fps

So, if you want to shoot at a 2.40:1 aspect ratio you can shoot 240x100 pixels at around 700 fps. On a standard definition television this would be scaled up to around (in reallity TV has non square pixels) 720 x 300, so each pixel on the camera becomes 3 pixels on TV. If you would shoot at other aspect ratios like 16:9 or 4:3 the picture would be upscaled even more and thus would look less sharp. An aspect ration of 2.40:1 is also interesting if you are shooting horizontal action, as you have more pixels to play with.

It is a bit of a low resolution but you said you wanted to compromise.

As for software, you can either roll your own or maybe the included software (I don't know if it can record video) or "Stream Pix" will work for you.

Cheers,
Take Vos

Robert Hruska
October 17th, 2007, 06:16 PM
David, you have unreal expectations shooting over 700 fps on a budget of £500. Your lighting alone would exceed that.

I agree. It's not a question of shooting video and slowing it down - as you have found already, taking a 24fps clip and slowing it down to that level just results in frames being copied. You don't get 700fps, you get 24fps stretched out, so each frame is 5 seconds long.

The problem is throughput. Even if your camera's sensor is capable of capturing at 1000fps, its storage subsystem isn't capable of writing it to disk or tape. My XH A1, for example, can only manage to store 24fps HD frames to tape by compressing the images. It can put 25 MB per second through its system. A VHX200 can store faster because its storing to hard memory, at about 50 or 100 MB/second (can't remember off-hand), but even it's throughput is only 60fps or so, I think.

1000fps, at the same size and compression as my Canon, requires a bandwidth of roughly 40 times as much as 24fps, or 1000 MB/sec.

Look at it this way: A garden hose and a fire hose do the same thing, but at different rates. Both systems can put 100 gallons of water on a fire, but the garden hose does it in 10 minutes, while the fire hose does it in 1. You're trying to send fire-hose level data through a garden hose.

As for your budget, you can get a consumer-level HD camera for about £500. But I don't know of any camera for less than, say, £5,000 that is capable of what you want. I don't know that you can even buy the components and put it together yourself. Systems capable of storing 40 times as much data as MiniDV are bound to be expensive.

As for building one... well, if you can find a way to do it, I bet you could sell a lot of them.

Steven Mingam
October 18th, 2007, 05:20 AM
The Micron SXGA 500fps sensor (http://www.micron.com/products/partdetail?part=MT9M413C36STC) cost already at least 1500$ and output 660MB/s.
You need a really fast storing facility (well with 2GB of DDR-SDRAM you could shoot 3 sec...) and it's not going to be cheap either...

Igor Babic
October 18th, 2007, 09:39 AM
http://www.epixinc.com/products/index.htm

http://www.epixinc.com/products/system_3cam_16g.htm

they have all you need for start, and they have a lot of info how to build system like this and how to see if your system is good for this task. Their site is very informative but a lot of info is spread across their products so be patient and read all.

here is link from their vision section.
http://vsd.pennnet.com/articles/article_display.cfm?Section=ARCHI&C=Feat&ARTICLE_ID=218273

here is list of MV cameras and their characteristics
http://damien.douxchamps.net/ieee1394/cameras/index.php

Enjoy