The rise of large-sensor cameras
Over the last couple of years there has been a revolution going on. The camcorder is being ousted by the DSLR in some sectors of content creation. NAB 2011 was full of owners and all manner of booths supplying aftermarket accessories.
Many budding directors of photography (DPs) have left film school to find the only camera they could afford to build up a portfolio was a prosumer camcorder. These cameras have three 1/4in or 1/3in sensors with an integral zoom lens. Designed for the wedding videographer and corporate communications market, they provide good pictures for the price, typically around 3000€.
In film school they may well have used Super 16 or more expensive digital cameras. They have soon discovered that although the low-end cameras do what they are designed to do, they lack many features necessary for the more creative DP.
Camcorders had followed a progression; the more you spent, the bigger the sensor, from low-cost 1/4in up to 2/3in for the full broadcast specification camera. Similarly, if you spent more, then the camera had an interchangeable lens mount. Simply put, larger sensors cost more, as do lenses to cover the larger format.
Some years ago, consumer manufacturers started to add basic video capture to compact digital cameras. In 2009, Nikon released the D90 with video capability. The real catalyst for the change came when Canon was approached by a news agency to add a video facility to a DSLR. This would allow photojournalists to shoot brief video clips using the same camera and lenses that they were using for stills coverage.
This was the eureka moment for indie filmmakers. If you couldn't aspire to the RED, a movie camera could be had for around 5000€ after adding a few accessories. Many owners already had a good collection of still lenses that could be pressed into service as primes.
The DSLR offers two advantages over a similarly priced camcorder: an interchangeable lens mount with a large selection of lenses available, and a large sensor, (41.3mm diagonal for a Canon 5D versus 6mm for a 1/3in sensor) which gives a small depth of field at large apertures. (See Figure 1.) The latter is important for “filmic” shooting, where a small depth of field is used to give depth to the two-dimensional scene and to focus the viewer on the subject.
Depth of field
Depth of field is the region of the object that has adequately sharp focus. The boundary is where a point in the image becomes sufficiently blurred to be perceived by the viewer of the final image as a circle of confusion rather than a point. Depth of field is not an absolute measure, but subjective. It is inversely related to the aperture of the lens, and to the focal length. A small aperture will create a larger depth of field than a wide open lens, to the limit of a pin-hole camera where all is in focus. A long focus lens will have a smaller depth of field than a wide angle (short focal length). Note that depth of focus refers to the image plane at the sensor.
The focal length of a lens is selected to give the desired angle of view. The angle of view is a function of focal length and sensor (image) size. The smaller the sensor, the smaller the focal length for a given angle of view. This means that a small sensor will have a shorter focal length for a given angle, with attendant larger depth of field. As large sensors will use longer focal lengths, then inherently the depth of field is smaller.
The resolution of the eye is around 0° 1' (1 minute of arc). For a 42in screen, at a viewing distance of 1.7m (to give a 30° viewing angle), the eye can resolve 0.5mm. The screen height is about 520mm, 1080 lines, so the eye can resolve about one pixel, hence the recommended viewing distance.
Depth of field can be derived from the hyperfocal distance. This is the closest point that appears in focus when the lens is focused on infinity. (See Figure 2.) Table 1 on page 34 gives hyperfocal distances for various sensors at the same angle of view.
Note that for the full-frame sensor, at f2, objects are in focus only up to 66m distant from the camera, whereas for the 1/3in sensor, everything is in focus up to around 10m. This indicates the problem with the small sensors: Nearly everything is in focus.
Television cameras have traditionally used a beam-splitting prism with three sensors: red, green and blue. This system was developed back in the days of tube cameras, and the principle carried on when CCD and later CMOS sensors replaced tubes.
Single-sensor cameras use a color filter array referred to as a Bayer filter, after Bryce Bayer, who patented the concept in 1976 on behalf of Eastman Kodak. Green filters, serving to represent luminance, occur every other pixel, with blue and red alternating, giving a quad of two green pixels, one red and one blue. (See Figure 3 on page 35.) The mosaic of color sites are demosaiced to give red, green and blue values for each site using interpolation algorithms. The value for each color is derived from weighted values of the nearest neighbors.
The resolution of the array will be less than the absolute pixel count as a consequence of the interpolation process. Contrast this with the three-sensor array where the full resolution of the sensor is delivered.
The beam-splitter was not practical for consumer cameras, especially the DSLR, where the large sensor size would have resulted in a bulky body. The long back focus would also have ruled out backwards compatibility with existing film lenses. The first commercially available cameras appeared in the early '90s.
The same need to avoid the optical complexities of the prism meant that cinematographers adopted the single-chip Bayer filter sensor for digital cameras with a 35mm film-sized sensor. The Panavision Genesis was one such early development, and proved popular. More recently, the Sony F35, the RED One, and the ARRI D-21 and Alexa have provided the digital cinematographer with a choice of cameras that can now rival film as a capture medium. However, these cameras are beyond the budgets of the indie filmmaker, and it was this pent up demand for a large sensor and interchangeable lenses that led to the ready acceptance of the DSLR as a lower-cost alternative.
The small depth of field has a related consequence: Greater precision is needed in focusing on the key subject. A telephoto DSLR lens for a full-frame camera can have such a small depth of field that if the subject's eyes are in focus, the end of the nose can be soft. Using the reflex optical viewfinder, the stills shooter can check focus before pressing the shutter release. However, when the mirror is up for video shooting, the optical viewfinder is of no use. The cameras provide a live view on the LCD display, but this is not full resolution, and it is difficult to see in full sunlight.
DSLR shooters can address the viewfinder issue in two ways. A simple solution is to shield the LCD display from light with some form of hood. Another is to use an external viewfinder driven from the HDMI output of the camera.
Cinematographers long ago developed ways around the problems of achieving sharp focus. Film actors are trained to hit the mark, a taped spot on the studio floor a measured distance from the camera. Unlike the television camera operator, who adjusts both zoom and focus, the film camera operator has a second person, the focus puller, to operate the focusing. The large focus wheel can be marked with key points of focus during a shot. The accurately calibrated lenses can be relied on to focus at the set distance. Contrast the DSLR lens, where the focus scale is more of a guide, and for most users autofocus provides a sharp image.
A film lens has a large travel from the closest focus to infinity, around 330 degrees of rotation. For a DSLR or video lens, it may only be 90 degrees. Film lens focus is always in one direction; with still cameras, it varies from manufacturer to manufacturer.
Various aftermarket accessories can be used to add film-style geared follow focus to the DSLR lens and even to reverse direction of rotation to the conventional.
If budget allows, rather than trying to adapt DSLR lenses, proper film lenses can now be purchased designed with the necessary image circle for the full-frame sensor, and with N and EF mounts. The film lens has long focus travel with an accurately calibrated focus scale, with gearing for the follow focus attachments. Films lenses are highly corrected, with low breathing and low flare.
The DSLR shooter now has the choice to add focus control for the cameraman or to use a focus puller.
The DSLR has application from the wedding videographer, who wants pleasing out of focus backgrounds, and to shoot with the same gear as the still pictures, all the way up to production companies shooting episodic television.
The audio facilities of most DSLRs are limited. For most shooting, the audio must be recorded with a separate device, film-style. Without time code for sync, traditional methods of the slate or clapperboard must be used, although sound can be synced in post using special software that aligns the camera sound on the timeline with the separate high-quality sound recording.
Camera manufacturers responded quickly to the demand for large-sensor cameras that could accept interchangeable lens.
Panasonic released the AG-AF100 with a Micro Four Thirds (MFT) sensor. This format was developed by Olympus as a smaller configuration than the full-frame DSLR, which would permit smaller, lighter cameras and lenses, yet include a larger sensor than a compact camera. The image circle is about twice that of a 2/3in sensor, but smaller than the APS-C used for most consumer DSLRs. The AF100 camera uses the Four Thirds mount, but a wide range of adaptors are available for other mounts including to PL. Video is encoded as AVCHD at rates up to 24Mb/s. HD-SDI and HDMI outputs at 4:2:2 8-bit sampling allows external recorders to be used for other codecs and with less compression.
Sony has the PMW-F3, which uses a Super 35-sized sensor and accepts PL-mount lenses. Forming part of the XDCAM EX family, it uses MPEG-2 long-GOP encoding at 35Mb/s recording to SxS memory cards. The camera has options for HD-SDI output as S-log and 4:4:4 sampling, which lends itself to compositing applications.
The F3 is now joined by the lower cost NEX-FS100, which uses the same Exmor sensor as the F3, but carries Sony E-mount lenses. The E-mount was developed for mirrorless stills cameras and the NEX camcorders. The flange to focal plane is about half that of DSLRs (as is the MFT), as no allowance is made for a reflex mirror. This allows for more compact cameras. The FS100 uses AVCHD or AVC coding at rates up to 28Mb/s. An HDMI output provides 4:2:2 uncompressed video with SMPTE time code for external recorders.
At NAB, Ikegami showed a prototype of a camera with a Four Thirds-size sensor, provisionally named the HDS-F90. This camera uses the PL mount. Video is encoded as MPEG-2 compression at 50Mb/s long GOP or 100Mb/s I-frame, and stored on the GFPAK SSD. Interestingly, the sensor is 4K and downconverted to 1080. This is claimed to give a better MTF characteristic for the sensor.
These cameras start at prices not much more than a high-end DSLR, but they record proper audio, have proper viewfinders and all the other facilities you expect from a video camera, plus they can record longer than 10 minutes (the limit of many DSLRs).
What future for the DSLR?
Will these cameras replace the hybrid DSLR? In some applications yes, but the video capability of the DSLR was originally developed for the cross-media journalist, and that requirement still stands. Newspapers and consumer magazines are now expected to carry video on their websites and tablet apps, so the shooter has to bring back video and stills. Single-chip camcorders only provide low-resolution still images. The RED EPIC introduces the concept of the Digital Still & Motion Camera (DSMC) and can capture 5K still images while shooting video, but that carries a higher price tag and is not going to replace the humble DSLR.
For the video shooter, these large-sensor camcorders overcome many of the drawbacks of the DSLR. Compelling features for videographers include an optical low-pass filter matched to video resolutions, to properly control spatial aliasing. Moire and aliasing on moving edges proves a problem for DSLRs. Another feature is the uncompressed video outputs for external recorders. This allows the videographer to choose the codec for the desired picture quality and to match their workflow. The codecs in a DSLR are a compromise in a device optimized for still image capture and low power consumption.
There has never been a wider choice of camera for the DP to choose from. The three-chip camera may remain as the mainstay for many broadcast applications from ENG to studio and sports production, but for episodic television, drama and commercials, the filmic look of the large sensors offers advantages to the DP, along with the ability to use high-quality film primes lenses.