Single CMOS sensors displace three-chip CCDs.
You could say there has never been more choice of cameras. And this year, NAB saw several new camera releases. NHK's Super Hi-Vision project to develop 4K and 8K UHDTV systems is driving innovation, much as its HD project led development decades ago. Although the current project started back in 2000, it is only in the last year that evidence of the research and development around the project is starting to bear fruit as products. The first steps are the increasing support for 4K imaging systems. This includes the sensors and lenses that can resolve such resolution.
It is not only spatial resolution that is increasing in the new cameras, but temporal resolution also is increasing with 60fps being commonplace and 120fps being supported by some cameras (60fps forms part of the Super Hi-Vision standard).
At first sight, 4K may seem a format more suited to theatrical distribution and not for broadcast. However, 4K is already being viewed equally as valid for HD production, especially for premium content. Just as SD programs look better when downconverted from an HD master, HD programs look better when downsampled from 4K.
Although the Super Hi-Vision is driving the top end, low-budget producers have created the whole DSLR revolution that has driven demand for large-sensor cameras to achieve the cinematic shallow-focus effects. These two influences, along with the general requirements for digital cinematography, have broken the 3-CCD mold that has defined video cameras until recently.
Three sensors or one?
Television camera design has a direct lineage from three-tube analog cameras, with the vidicons replaced by CCD as equipment moved to digital processing, and latterly as CCD has made way for CMOS sensors. There were some single-tube cameras, using striped color filters on the face of the tube, but the trisensor design with a beam-splitting block dominated the broadcast market.
In contrast, digital still and cinema cameras developed from film cameras. The film gate was replaced with a single digital sensor. The same lenses could be used with either image capture method.
Film uses three layers, the tripack, each sensitized to a different band of the spectrum, so there was no need for beam splitters. The challenge for digital cameras was to capture red, green and blue signals from a single sensor.
Early attempts at a digital color imaging system used field-sequential color filters, much like early television, but this only works for still subject material. Another approach was the use of color stripes, as had been used with single tube cameras for consumer products. In 1975, Bryce Bayer at Eastman Kodak patented a color filter array (CFA), a concept that forms the basis of today's single-sensor products. The three-layer concept of film has been emulated with the Foveon sensor, which is used in Sigma still cameras.
Bayer's patent describes a means of deriving a luminance and two chrominance signals, with a green filter being used to create the luminance-like signal, and blue and red for the chrominance signals. The filters are interlaced with two luminance sensors (G), and the red and blue, for a 4:2:2 sampling arrangement. This arrangement remains the most popular today, although some sensors use cyan, magenta and yellow filters (like film negative). The interleaving provides for uniform sampling in the horizontal and vertical directions, which is not the case with the striped filters.
The drawback of the Bayer sensor is that a 4K display has only 2K of green resolution. To counter this, manufacturers interpolate the raw RGBG sensor information to derive an RGB value for each sensor site. Picture quality from a color filter array (CFA) is dependent on the algorithm used for the interpolation. A basic algorithm will generate false color at the edges of objects as well as softening the image. Subjects with high spatial frequency information will also alias with the CFA creating moiré artifacts. These are especially visible if the fine detail is in motion.
The combination of the modulation transfer function (MTF) of the lens, the optical low-pass filter (LPF), the sensor, and the interpolation algorithm that demosaics the raw data all contribute to the final image quality.
Is the three-sensor camera on the way out? Time will tell. It is expensive to build a camera larger than 2/3in with a beam splitter and three sensors. The long flange-to-sensor depth complicates wide-angle lens design.
The single-sensor camera is not without issues. The interpolation from the Bayer pattern to three co-sited RGB pixels can lead to artifacts. Sophisticated algorithms and an optical LPF before the sensor will minimize the artifacts.
Rather than trying to interpolate to achieve a final RGB image the same size as the sensor, some manufacturers have taken different approaches. Canon uses a Bayer array in the CF300 and CF500, but it doesn't demosaic for the HD output. Instead, the 4K sensor is read out as RGB, 1920 × 1080. This avoids the artifacts of demosaicing and produces clean images on test patterns, free of the aliasing and moiré found with the demosaic process.
Sony's F35 uses RGB stripes with a 12-megapixel sensor. Again, no interpolation is necessary to generate the 1920 × 1080 grid of RGB pixels.
The PL mount is becoming as popular for video cameras as the B4 and an indication of the crossover between digital cinematography and television production. PL has the advantage of a wide choice of lenses from rental houses. Many cameramen may already own PL lenses from their film days.
The DSLR revolution has brought the EOS mount to the fore, with a huge choice of low-cost lenses designed for still photography. The Nikon F mount and Four-Thirds are also popular.
So why pay for a broadcast lens when an SLR lens can be purchased for under €1000?
Let's compare and contrast DSLR lenses with cine and broadcast lenses. (See Table 1.) Television cameras may have integral or separate lenses — usually wide-range zooms. DSLR and digital cinema cameras use fixed focal length lenses, and only occasionally short-range zooms. However a cine lens is very different from an SLR lens. Aside from higher performance (and price), cine lens are designed for operation by a focus puller via fixed-size geared rings. The dimensions of the cine lenses are matched across the range to make the interchange of lenses much quicker, coupling to the matte box and the focus mechanism.
| ||DSLR ||Television ||Cinema |
|Cost ||Low ||High ||High |
|Performance ||Medium to good ||Good ||Very good. Highly corrected for breathing and aberrations |
|Focal length ||Fixed, short-range zoom ||Zoom up to 80:1 ||Prime fixed, short-range zoom |
|Focusing ||Autofocus/manual. Small focus rotation angle. ||Manual via servo. Focus by eye. ||Manual. Focus by focus puller, graduated focus scale, large focus angle. |
|Dimensions ||Nonstandard. Changing lens needs realignment of matte box. ||Box or field zoom. ||Standard, for use with matte boxes and rigs. |
|Robustness ||In line with price ||High ||High |
Table 1. Contrast and comparison between DSLR, television and cinema lenses
The current round of large-sensor cameras has tended to opt for the PL mount, giving a wide choice of cine lenses from rental houses, although the EOS mount is also in evidence.
The last few years have seen remarkable progress in increasing the sensitivity and lowering the noise floor of the sensors. It is possible to shoot with ambient light, even at nighttime with the latest cameras. That is not to say that lighting is no longer needed. Controlled lighting is all part of creating the image. The need for less light has led to a much more natural look, of great advantage to drama productions.
It's been a long time since Zeiss produced very fast lenses for the Kubrick production “Barry Lyndon” in order to shoot scenes by candlelight, but that natural look is very much in vogue today. Crime series, among other dramatic genres, use much less obvious lighting and a more “natural” look. Observational documentaries also stand to gain from the new, more sensitive cameras.
The increase in sensitivity is accompanied by an increase in dynamic range — highlights to shadows. The wide dynamic range is achieved by a number of techniques for reading the data from the sensor. The current high-end sensors can produce a dynamic range approaching or equaling film negative.
Tape transports used to set the minimum size for a camcorder body. Using a memory card makes a much smaller camera body possible. A single-sensor design means no beam-splitter, so the camera can be even smaller. There are ergonomic limits. A palmcorder is all very well for the consumer on vacation, but a professional camera needs all the controls for set up, flexible mounting options, means to attach wireless microphone receivers, eyelights, shotgun microphones and all the other paraphernalia for shooting. The DSLR shows that you can improvise with rigs, but for broadcast applications the conventional shoulder-mount layout provides for good stability when hand held, with the eyepiece falling in a convenient position. The smaller cameras must be held in front of the body, calling for rigs to provide the same stability as a shoulder mount.
Many cameras now include a smartphone-sized LCD screen for set up menus and as a viewfinder to complement the conventional electronic viewfinder (EVF). Although the LCD screens are useful to confirm focus, sunlight viewing can be challenging. The EVF has never had the resolution of the optical finders of film cameras. Now that large monitors are often used on set, the camera operator may find that he or she has the lowest resolution picture yet is attempting to focus a wide-aperture lens. The shallow depth of field of the large-sensor cameras compounds these issues, especially if the production style or budget doesn't include a focus puller.
DATA AND MEDIA
A 124-minute HDCAM-SR tape stores around 400GB of data. Contrast this with a 50GB of a Professional Disc, or 64GB typical of memory cards. Although the Sony F65 can use SRMemory packs with a 1TB capacity, most operators must juggle rapidly filling memory cards.
The smaller capacity of memory cards when compared with tape has led many vendors to offer high compression for the camera codec. At the low end, 8-bit AVCHD at 24Mb/s is common, and 8-bit 50Mb/s long MPEG for XDCAM.
This has led to the field recorder, a separate unit that records an uncompressed HD-SDI or HDMI output from the camera to a hard drive or SSD. Not only does this improve image quality, but some cameras also provide a 10-bit output, lending the images to more color grading than would be possible with an 8-bit signal. Many field recorders can code to DNxHD or ProRes, creating an edit-ready file that neatly avoids the heavy compression of a camera codec along with the concatenation artifacts of re-encoding at ingest to the NLE.
Panasonic has taken a different route; the AVC-I codec and P2 cards offer a direct-to-edit solution. Advances in semiconductor technology allow Panasonic to offer the P2 workflow in an SDHC footprint, microP2, rather than the old PC Card form factor.
Selecting a camera for a production used to mean choosing between 35mm or television, and if television, either a 2/3in or 1/2in sensor. The choices are blurred now — CCD or CMOS, three-chip or single sensor, 1/3in up to Super 35 and larger, and now HD or 4K.
The decision may well be decided by weight, cost, the capture codec and the recording medium. The latter two impact on the complexity or otherwise of the downstream workflow. It is common to see inappropriate choices made as production companies come to learn the hard way about the optimum camera for a shoot.
Although CCD technology dominated the market for video cameras, all the research now seems to be going into developing the quality from CMOS sensors. CMOS is often derided for the Jell-O effect, but this is minimized in some of the more sophisticated cameras through fast readouts from the sensor, and in the case of the F65, a mechanical shutter.
All this choice is turning the world of camera buying upside down. Sub-€10,000 cameras are being used for newsgathering. Even the low-cost cameras have XLR audio inputs, and the HD quality exceeds cameras five times the price from five years ago. DSLRs are popping up in all manner of shoots, much to the chagrin of those in post. The rule book has been torn up, but this is art. The debate about which camera to use will run and run, but no one can complain they have little choice.
David Austerberry is editor of Broadcast Engineering World Edition.