THE VIPER's SHOOTING MODES

The beauty of the Grass Valley/Thomson VIPER "Filmstream" Camera System is its extreme versatility. In a sense, it is like several very different cameras rolled into one. It has 4 distinct shooting modes, it has a choice of 2 aspect ratios (both native), and it offers a choice of HD formats, either 1080p or 720p. This allows for several possible configurations of the Viper, each very different from the other. Within these configurations, there are further options like A: pure data-streaming modes or conventional processed video modes, B: full 4:4:4 color space or more conventional 4:2:2 color space, C: Single-link or Dual link output (SMPTE 292M or 372M) , D: 10-bit Log or 10-bit Linear output, E: a choice of frame rates, and F: a choice of external recording decks / formats. Sounds versatile ....... well it certainly is, which is the genius of the design. But with so many options available, it can be extremely difficult (and often very confusing) to decide how best to set up and deploy the camera for any individual production.

The 4 Basic Modes

The fundamental choice is first to select one of the 4 possible shooting modes. They are: (1) DATA 4:4:4 (called "Filmstream" mode), (2) DATA 4:2:2 (called "HD Stream" mode), (3) VIDEO 4:4:4 (called "RGB" mode), and (4) VIDEO 4:2:2 (called "Y,Cb,Cr" mode). This last mode used to be called "YUV" in older software versions, but this shorthand is technically incorrect, and was subsequently dropped. The first two are unprocessed, or relatively unprocessed, data streaming modes, and the last two are more conventional processed video modes (white balanced, gamma corrected and so on).

In describing these four shooting modes, let me draw some parallels that might help in understanding them a little better. "Filmstream" mode is like a 35mm film scanner, but with a lens attached. There is no direct parallel with "HD Stream" mode that I know of, but it is like "Filmstream" mode, but with some in-camera processing to convert it to 4:2:2. The Viper's "RGB" mode is very similar to a Sony HDW-F950 camera. And finally, the Viper's "Y,Cb,Cr" mode is similar to a Sony HDW-F900 CineAlta or a Panasonic AJ-HDC27V/F/H Varicam. All modes are 10 bit: two are 10-bit Log, and two are 10-bit Linear. Remember that the F900 and Varicam are strictly 8 bit to tape (defined by their respective formats), although it is possible to get 10 bit out of these cameras externally.

'Filmstream' mode

Thomson's FilmStream mode is rather special, and I believe unique to 2/3-inch HD cameras. Let me describe how this mode works in a little more detail. In FilmStream mode, the Viper's CCDs go through 12-bit A/D (analog to digital) processing which converts the signal to RGB data values using logarithmic calculations. The end result is then converted to 10-bit data values (still log values) and transferred to an external recorder as a pure data stream using a dual HD-SDI link as a carrier (SMPTE 372M). Full resolution is maintained with true-progressive 1920 x 1080 pixels for every color. There is no color sub-sampling, no color space conversions, no irreversible video manipulations, no further quantizations, and no compression. Basically, 'FilmStream' mode streams all the data output by the CCD's preserving most of the information every individual CCD image sensor has to offer. This mode is especially beneficial for blue/green screen applications and intensive CGI-effects work (i.e. heavy computer processing and manipulation).

As remarkable as the 'Filmstream' mode is, the resulting images appear very flat-looking with a strong greenish cast when viewed on a dual-link, 4:4:4 monitor. The Viper provides a parallel channel (called the Viewing Channel) that allows the signal to be fully corrected with all the normal video-processing tools, and viewed on a regular 4:2:2 HD monitor. Although this signal is not (usually) recorded, it provides a normal-looking image to work by on set by giving a good indication of how the final images will appear after it goes through the post process.

Increased Dynamic Range

Simultaneously checking the images with a 4:4:4 monitor is still recommended however, as only this monitor will show the increased dynamic range / exposure latitude being recorded (which can be significant - an extra 2.5 stops in the highlights). This is because the use of 'Filmstream' mode allows you to drive the CCD sensors to their maximum ability to capture brightness. In conventional digital/video cameras, the top 70% of the CCD range is compressed (highlight compression). But with 'FilmStream', this extended part of the CCD exposure range can now be used utilized, dramatically increasing over all dynamic range of the imagers to 9 stops, or more. A 4:4:4 monitor can also be useful in double checking that both the A and B links are working properly. The absolute integrity of the cabling to the external recorder is of paramount importance when shooting with the Viper (be it single link or dual link). This needs constant monitoring and careful attention.

For the D.P., Camera Operator, and Focus Puller, the Viper in 'Filmstream' modes works in a very similar way to working with a film camera. As with film, they must concentrate on lighting, framing, focusing and setting accurate exposure. So their method of working strongly parallels that of film.

All 10 bit

The Viper FilmStream Camera uses a 10-bit logarithmic format per color (30 bits in total). '10 bit' means that digital information is represented by 10 digital bits, or 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 = 1024 code values per CCD. This is significantly better than the more-limited 8-bit system used by most digital cameras, which has only 256 code values. There are two major reasons for using the logarithmic system (.....paraphrased below from Thomson's brochure for the Viper).

1. The human eye responds to light in a logarithmic way. The human eye can see more color values, and is also more discerning, the darker the scene. Therefore, it is preferable to assign more bits for the darker regions of the CCD range than for the brighter parts. A logarithmic transfer curve achieves precisely this by assigning smaller steps for the darker part of the range (more code values), and bigger steps for the brighter part (less code values). Moreover, the logarithmic data format can express more color tones. In fact, 10-bit logarithmic values would require more than 12-bits of linear values to maintain the same visible color space in order to express the full contrast range of the CCDs. By digitizing the CCD signals with logarithmic encoding, optimum use is made of the finite pool of available bits (a total of 1024 code values). The bits are used in the area where they are needed most in terms of the eye's sensitivity.

2. The logarithmic data format also guarantees that the quantization steps are always significantly smaller than the noise level. In CCD cameras, there are two major sources of noise. The first is independent of the light exposure of a pixel. The output amplifier of the CCD causes the bulk of this noise. Even though it is equivalent to no more than 10-20 RMS electrons, it is a dominating factor for the darker parts of the scene. The second major source is photon shot noise. The mechanism of photon shot noise can best be explained by comparing photons to rain drops. When you put two glasses in the rain and remove them after a while, you find that they differ in the number of raindrops they have gathered. Actually, the RMS fluctuation of the number of raindrops is equal to the square root of the average number of raindrops. So, the absolute RMS fluctuation increases with the amount of rain. In CCDs, exactly the same thing happens with the number of photons that reaches each pixel. So, above about 1000-4000 electrons in a pixel, the photon noise becomes the dominant noise factor. In the higher part of the CCD range, the photon-shot noise is several hundreds of equivalent RMS electrons. By using smaller quantization steps for the darker parts of the CCD range, and bigger steps for the higher range, the quantization steps are better matched to the varying noise levels of the CCD. In this way, the quantization noise can always be kept several dBs under the other noise levels. To get accurate and reproducible results, especially between cameras, the Viper FilmStream Camera uses on-line digital calibration technologies that insure that both the black and saturation points of the CCDs are exactly aligned to the correct digital data values regardless of ambient temperature, humidity, or camera age (the above extract taken from Thomson's brochure for the Viper).

The table below summarizes the pertinent points that distinguishes the Viper's four shooting modes:

The Viper's HD-DPM+ Image Sensors

The Viper 'FilmStream' Camera uses three 9.2 million pixel HD-DPM+ CCD sensors, developed by Thomson engineers. Thanks to the special design of these HD-DPM+ image sensors, the Viper camera can be switched between several different formats, including 1080p at 24/25/30 frames per second, and 720p at 24/25/30/50/60 frames per second. Remarkably, full resolution is preserved and the horizontal viewing angle stays the same. So these very special CCD's are at the heart of the 'magic' of the Viper design making switchable HD formats and switchable "native" aspect ratios possible.

Variable Pixels

The HD-DPM+ CCD sensor consists of 1920 horizontal pixels and 4320 vertical sub-pixels (effective pixel count). By grouping the vertical sub-pixels, the vertical line count can be changed (Wow!). This makes the following configurations possible:

1080P

When four vertical sub-pixels are combined per scanning line, the total line count becomes 1080 lines (4320/4 = 1080). So, a native 1920 x 1080 image sensor is obtained with a 16:9 aspect ratio. Full resolution is preserved and the horizontal viewing angle stays the same.

720P

When six vertical sub-pixels are combined per scanning line, then the total line count becomes 720 lines (4320/6 = 720). So, a native 1920 x 720 image sensor is obtained with a 16:9 aspect ratio. Again, full resolution is preserved and the horizontal viewing angle stays the same. To maintain the 720p standard, the Viper converts the output to 1280 x 720 for external recording.

Note: 720p mode has the advantage that higher frame-rates can be captured (up to 60p). This makes it possible to generate true, slow-motion effects in post-production (similar to the Varicam, actually using the same technique).

Cinemascope Aspect Ratio

When three vertical sub-pixels are combined, this results in 1440 vertical lines (4320/3 = 1440). By using the middle 1080 lines, a 2.37:1 aspect ratio is made possible without the need for anamorphic lenses, while still maintaining full 1920 x 1080 resolution. Hey presto! the Cinemascope aspect ratio is achieved without the need to shoot with anamorphic lenses. With the Viper 'FilmStream' Camera there is no need to crop the image and lose resolution to get this aspect ratio (as with all other 2/3-inch HD cameras). It should be noted, that in this 2.37 "wide" mode, the dynamic range and sensitivity is somewhat less than in the 16:9 aspect-ratio mode. This is because of the smaller vertical size of the pixels (smaller pixel area = less light sensitivity).

What is effectively happening is that the vertical aspect ratio of the individual pixels themselves is being changed. Looking at it this way, might help you in understanding how this system works. The individual pixels can become taller or shorter, making it possible to generate different aspect ratios and HD formats natively from the same CCD array. In other words, the individual pixels do not stay the same shape and size, as with (most/all?) other designs. It is certainly an ingenious concept, and unique to Thomson as far as I know.

The Viper's CCD imager is the Frame-Transfer (FT) type, which is well suited for progressive image capture. Thomson says that because of the Viper’s mechanical shutter, CCD exposure has a similar ramp to film cameras, which in turn produces similar motion characteristics in the resulting captured images. The Frame-Transfer principle has parallels to the exposure method used in film cameras. During a certain exposure time (shutter angle in film cameras), the light is projected on the photosensitive surface. Then, the mechanical shutter wipes in front of the imaging area, effectively blocking the incoming light. During the time in which the shutter shields the photosensitive surface, the image is transported to a shielded area where the image is read out, line by line. So the complete surface of the sensor is effectively used to capture images, which helps to suppress fine-detail moire patterns. Other designs (FIT) incorporate this storage area within the imaging area itself, but this is not the case with the FT design.

Shooting for Slow Motion (Over-cranking)

When the Viper is set in the 720p mode at 60 frames per second, slow motion effects can be achieved during post-production similar to the Varicam (using the same techniques). During post-production, these 60 frames per second can be slowed down to 24 frames per second thus giving a 2.5 times slow motion effect (60/24 = 2.5). Other slow-motion ratios are possible, but some of these require more computation and processing in post-production.

The 1280 x 720 slo-mo effects shots can be converted to 1920 x 1080 if that is the format used for the rest of the production. Although there is some loss of resolution in the 720p 'slow-motion' mode, the visible loss is remarkably small since the CCDs keep running at 1920 pixels horizontally.

Also 'Video' RGB and Y,Pb,Pr Modes

While the 10-bit Log Data Streaming modes are unique to the Viper in this class of camera, the Viper also has more traditional 'broadcast video' modes available as well (both 10-bit Linear and either dual-link 4:4:4, or single-link 4:2:2 output). Full video processing is available onboard similar to most broadcast-style cameras. This includes gamma adjustment, black stretch, adaptive knee functions, color matrices, and traditional image enhancement (detail) circuitry. The multi-format capability described above is available in these shooting modes as well. Once again all these modes and functionality are summarized in the diagram above. It is definitely not an understatement to say this is a remarkably versatile camera system.

Recording Format Options

To add to the systems versatility, there is a range of recording formats and 'decks' available. You can record to either tape, or non-tape alternatives like flash memory or hard-drive arrays. Recording options fall into two broad categories, either 10-bit capable, or 8-bit capable. Everything else being equal, a 10-bit recording system is preferable to preserve the higher-quality output that the Viper is capable of. But course, 10-bit recording systems are more expensive and the workflow a little more sophisticated. The best 10-bit tape option is the HDCAM-SR format (440Mb/s or 880Mb/s) with two deck options. The SRW 5000 studio deck, or the SRW-1/SRPC field-pack recorder, both using very mild (imperceptible) compression. You can record without any compression at all to the S.two DFR (Digital Field Recorder) which is a hard-drive array recording system optimized for recording with the Viper. You can also record to Thomson's own Venom Flash Pak, which is an onboard, solid-state memory recording system, primarily designed for untethered Steadicam shooting. All these recording systems are dual-link 4:4:4 capable, so they are the highest-quality options with very little, to no compression used at all.

Next step down is a 10-bit, single-link, 4:2:2, recording system. The Panasonic AJ-HD3700 D-5 HD studio recording deck which is 10-bit, 4:1 compression, 235Mb/s data rate, 1080 60i/24p, fits the bill nicely.

The remaining recording options are all single-link, 4:2:2, and 8-bit only. They are the Sony HDCAM format (140Mb/s) with decks like the Sony HDW-F500, and the Panasonic DVCPRO HD format (100Mb/s) with decks like the Panasonic AJ-HD1200A etc. But these decks will convert the 10-bit signal down to 8-bit for recording.

As if the choice of shooting mode is not hard enough, to this you have to add a complimentary recording format/system. Lots of options, so lots of decisions (phew). But at least you have a choice. Format and shooting mode decisions like this can be a real brain teaser, as so many variables have to be weighted up and taken into consideration. Historically speaking, format/shooting mode decisions have always have been tricky, but perhaps more so now in an emerging digital world.

Biggest Gotcha

There is an inherent 'weakness' (if it can be called that), with external recording. Or perhaps inherent risk is a better describtion. I strongly recomment using the Viper's Breakout Box and multi-core cable system, rather than a simple BNC cable hookup to the recorder. BNC connectors are not bad when they are used in stationary installations, but they are not designed for the cables to be continually moved around, and jostled about, and tugged, and so forth (in other words, like what happens constantly on a typical movie set). BNC's will fail eventually ....... usually quicker than you think. But mostly not catastrophically at first - well, unless the cables are kicked out completely. Catastrophic failure is not so bad as the problem is obvious and you can fix it quickly before shooting the next take. But rather, BNC cables usually go out more slowly producing a dirty signal that progressively degrades over time. It can be hard to see this problems at first, partly because the problem might only be a momentary event lasting only a fraction of a second in time (a glitch). The multi-core connectors, on the other hand, are much tougher, and more robust, and generally speaking better designed. They are indeed designed to have the cable dragged around as the camera is repositioned, including people tripping over the cable etc. The biggest risk you take using an external deck is the risk of a compromised recording due to bad cabling, especially BNC cabling. You have to check and guard against this constantly. It is much easier with fiber optics as there are monitoring systems and alarms to tell you about the precise integrity of the signal (number of errors). With copper and BNC's, all you have is your waveform monitor and a HD picture monitor to try and detect (i.e. see) any problems as they develop. If you have no other option but to use BNC cabling, then you may also want to change out your BNC cable between the Viper and the deck periodically, even if it seems to be working O.K. This might help prevent potential problems before they get a chance to occur. Never use a BNC "barrel", or BNC "elbow", or similar BNC adapters in this critical cable between the Viper and the external recording system. That is just asking for trouble. I think anybody would be lucky to get through a whole shoot without some problems if relying on BNC's alone. It only has to save you reshooting a shot or two because of glitches or similar problems, to make renting the breakout box and multi-core cables worth while economically speaking (not to mention in terms of time and schedule).

Just to give you an idea, here is what comes out of the Breakout Box......

Breakout Box Connections:

Multicore Input
HD-SDI Link A Output (x2)
HD-SDI Link B Output (x2)
Viewing Channel HD-SDI Output (x2)
CVBS Out (= Composite Video Blackburst & Sync = downconverted NTSC Composite Video)
Trilevel Sync Input
Playback Input
Audio Out (I think this is for a camera-mounted microphone only)
9-pin Control Interface (has tally light and VTR start/stop etc.)
Power In 12 Vdc
Power In 24 Vdc

Peter Gray
(in California)
P.O. Box 5132
Pine Mountain Club, CA 93222
United States of America
telephone: +1 (661) 242-1234

dp@petergray.org