Ben Winter
July 16th, 2011, 03:42 PM
Somewhere, right now, an Olympus engineer is laughing.
An engineer myself, I would be laughing too at the doubtless dozens of hopeful filmmakers thinking they're clever by using a piece of scientific test equipment as a digital cinema camera.
The i-Speed 3 is a slow-motion camera capable of 1,500 frames per second at 1280x1024 resolution. For the most part, it performs as advertised. And it advertises very, very little.
A thorough search of the internets for anything to do on this camera reveals a lot of people wondering how well it works, some "test shots" on vimeo, and poorly lit videos done by aspirant filmmakers claiming the low light and flickering are "creative decisions." Well, folks, there's a few very good reasons for that (helpful tip: for high-speed you need lots of flicker-free light. Is that a tricky concept?)
At only $500/day, 3-day minimum from a local rental company, I thought this myself: how could I go wrong? At worst, it would be a $1500 experiment. And you know us engineers. We love experiments/playing with expensive toys.
The shoot was a continuation of the WTA's quest to prove their players aren't actually men, (http://www.nytimes.com/interactive/2010/08/29/magazine/womens-tennis.html) and this particular shoot was missing the mirror polish of a $5000/day budget for a Phantom camera and operator. While it had started as a print shoot only, the decision was made to see what slow-motion could be captured with minimal investment. Enter the mistake, the i-Speed 3.
The first problem is the Nikon mount. Who uses Nikon for motion capture? The obscurity combined with the flange focal distance renders impossible any hope of adapting to any other mount. Try finding a Nikon body to PL lens mount. I dare you.
The second is that there is no way to save uncompressed video from the camera. The option is there, but for no good reason at all, Olympus engineers decided the maximum file size you should be allowed is 2GB, which equates to just 600 uncompressed frames. This leaves you with either dividing up your shot into 600 frame slices manually, a time-consuming venture, or saving to the "Very High" quality format offered: a proprietary, antiquated flavor of MJPEG compression, which can't be converted into any other usable format unless processed through Olympus's i-Speed software suite. Even then, you're lucky to get your footage out in one piece provided the software doesn't crash. And did I mention the time it takes to download footage from the camera's internal high-speed memory? If you enjoy counting from 0 to 100% very, very slowly, this camera is for you.
The image quality problems continue, unfortunately, when you take into account the camera's sensor. It shoots 1280x1024, yes, but that's the actual pixel dimensions of the active area. The Phantom Flex for instance, crams 2560 x 1600 pixels into a 1920x1080 image, so clever downsizing and debayering renders a nice, clean image. With the i-Speed, the massive 21 micron pixels stumble around the image like a fat drunk trying to find the door, and the result is a pixelated, unsubtle digital edge to anything you shoot.
But this is just the opening act for the sensor's major downside. You see, due to complicated engineering restrictions that I won't get into, the sensor's active area is actually split into four squares, each with its own readout circuitry. Minor, unavoidable manufacturing discrepancies between the circuits gives rise to slightly differing behavior in each circuit, with different quantum efficiency, sensitivity, and dark current. Basically, this means that of the four squares that comprise the final image each look slightly different from each other, giving you a faint, but noticeable checkerboard effect. Calibration of the sensor's white and black points makes no difference. The only possible savior is a trick I learned while working at NASA's detector characterization lab: put a cap on the lens, take an image, and subtract this "dark current image" from the rest of your shots using After Effects. The MJPEG compression nearly ruins this solution, but it beats stubbing your toe on the camera while you're using it as a doorstop.
There are a few—very, very few—redeemable qualities about this camera however. For one, it's insanely simple to use. Olympus engineers spared no expense demonstrating how stupid they think you are. The "CDU" control screen has nice Big Buttons that Do What They're Labeled To Do. The kit comes complete with everything you think you'll need, and even a nice little lighting kit with some Lowell omni's and stands. It's only missing the answer to why you're wasting your time with this camera, which is the one question the helpful reps at Olympus won't answer for you.
The camera was designed and manufactured to visualize and debug high-speed manufacturing equipment, and for that purpose, it's clearly a very helpful tool. But it's not a shortcut to quality high-speed imaging, and it shouldn't even be in the same sentence as the word "Phantom."
An engineer myself, I would be laughing too at the doubtless dozens of hopeful filmmakers thinking they're clever by using a piece of scientific test equipment as a digital cinema camera.
The i-Speed 3 is a slow-motion camera capable of 1,500 frames per second at 1280x1024 resolution. For the most part, it performs as advertised. And it advertises very, very little.
A thorough search of the internets for anything to do on this camera reveals a lot of people wondering how well it works, some "test shots" on vimeo, and poorly lit videos done by aspirant filmmakers claiming the low light and flickering are "creative decisions." Well, folks, there's a few very good reasons for that (helpful tip: for high-speed you need lots of flicker-free light. Is that a tricky concept?)
At only $500/day, 3-day minimum from a local rental company, I thought this myself: how could I go wrong? At worst, it would be a $1500 experiment. And you know us engineers. We love experiments/playing with expensive toys.
The shoot was a continuation of the WTA's quest to prove their players aren't actually men, (http://www.nytimes.com/interactive/2010/08/29/magazine/womens-tennis.html) and this particular shoot was missing the mirror polish of a $5000/day budget for a Phantom camera and operator. While it had started as a print shoot only, the decision was made to see what slow-motion could be captured with minimal investment. Enter the mistake, the i-Speed 3.
The first problem is the Nikon mount. Who uses Nikon for motion capture? The obscurity combined with the flange focal distance renders impossible any hope of adapting to any other mount. Try finding a Nikon body to PL lens mount. I dare you.
The second is that there is no way to save uncompressed video from the camera. The option is there, but for no good reason at all, Olympus engineers decided the maximum file size you should be allowed is 2GB, which equates to just 600 uncompressed frames. This leaves you with either dividing up your shot into 600 frame slices manually, a time-consuming venture, or saving to the "Very High" quality format offered: a proprietary, antiquated flavor of MJPEG compression, which can't be converted into any other usable format unless processed through Olympus's i-Speed software suite. Even then, you're lucky to get your footage out in one piece provided the software doesn't crash. And did I mention the time it takes to download footage from the camera's internal high-speed memory? If you enjoy counting from 0 to 100% very, very slowly, this camera is for you.
The image quality problems continue, unfortunately, when you take into account the camera's sensor. It shoots 1280x1024, yes, but that's the actual pixel dimensions of the active area. The Phantom Flex for instance, crams 2560 x 1600 pixels into a 1920x1080 image, so clever downsizing and debayering renders a nice, clean image. With the i-Speed, the massive 21 micron pixels stumble around the image like a fat drunk trying to find the door, and the result is a pixelated, unsubtle digital edge to anything you shoot.
But this is just the opening act for the sensor's major downside. You see, due to complicated engineering restrictions that I won't get into, the sensor's active area is actually split into four squares, each with its own readout circuitry. Minor, unavoidable manufacturing discrepancies between the circuits gives rise to slightly differing behavior in each circuit, with different quantum efficiency, sensitivity, and dark current. Basically, this means that of the four squares that comprise the final image each look slightly different from each other, giving you a faint, but noticeable checkerboard effect. Calibration of the sensor's white and black points makes no difference. The only possible savior is a trick I learned while working at NASA's detector characterization lab: put a cap on the lens, take an image, and subtract this "dark current image" from the rest of your shots using After Effects. The MJPEG compression nearly ruins this solution, but it beats stubbing your toe on the camera while you're using it as a doorstop.
There are a few—very, very few—redeemable qualities about this camera however. For one, it's insanely simple to use. Olympus engineers spared no expense demonstrating how stupid they think you are. The "CDU" control screen has nice Big Buttons that Do What They're Labeled To Do. The kit comes complete with everything you think you'll need, and even a nice little lighting kit with some Lowell omni's and stands. It's only missing the answer to why you're wasting your time with this camera, which is the one question the helpful reps at Olympus won't answer for you.
The camera was designed and manufactured to visualize and debug high-speed manufacturing equipment, and for that purpose, it's clearly a very helpful tool. But it's not a shortcut to quality high-speed imaging, and it shouldn't even be in the same sentence as the word "Phantom."