Chips are not bad for your complexion.

August 21, 2009

Chips are not bad for your complexion.

While many Hi-def camera manufacturers battle it out in an endless debate over resolution and related formats, another, equally significant, resolution battle is being waged in Hollywood, by actors reluctant to sign on to HD productions because they believe HD will make their complexion look bad. Failing to understand that 35mm film has more resolution than any current HD displays, they fear that increased resolution will portray them negatively, making their complexion look old an haggard.

What they don’t know is that when film undergoes a DI (digital intermediate), it is scanned at 4K (4 million pixels).  Actually 4Kfilm scanners scan at 4096 x 3112, or about 12.7 MegaPixels.  However, not all DI’s are scanned at 4K.  It’s evolving in that direction, but most are currently scanned at 2048 x 1556 or roughly 4 times the resolution of a million pixel HD imager. Yes, an even higher resolution than the formats they reject, yet film is still an acceptable medium for them.  While many HD cameras will not produce an image that is satisfying in this regard, lacking that very “something “ that film possesses, in the right hands, many HD cameras will now produce a filmic look.  That’s right, not all HD is created equal and it has little or nothing to do with resolution and everything to do with sharpness, or more appropriately, limited dynamic range.  So the producer’s conundrum becomes “shall we allow camera selection to become the domain of starlets, or do we need to revise our thinking in evaluating camera selection?”

Imagine that you are looking at a resolution chart through the lens of your camera.  The fine series of alternating black and white lines is the maximum number of lines that can be resolved by your camera and they are clear, clean, and sharp.  Let’s also imagine that you are routing the video signal to a waveform scope and that the white lines appear on the scope at 100% and the black lines at 0%.  In reality this would not be the case, but for the purposes of discussion we will use these numbers as theoretical reference numbers.  With a pure white and a black, these lines would appear as crisp as is possible, that is to say, very sharp.

Now imagine that you begin to throw the lens out of focus.  As the lack of focus increases, the alternating lines diminish in tonal value as each approaches 50% on the scope.  In other words, white values lower and black values rise. The edges of the formerly crisp lines blur as the white and black values intermix, each degrading the other.  Eventually (theoretically and on some lenses only), the fine lines become blurred to the extent that the entire screen is a 50% value on the scope, or what we photographers call 18% grey.  This test would produce the same results regardless of the resolution of the camera, however the finer the lines, the sooner the effect would be noticed.

The practical test for resolution of the entire camera system, including the lens, is our ability to count the individual lines.  If you can count them, they are resolved.  However, the test for sharpness is the measurement of the difference between the values of white and black.  That difference is referred to as depth of modulation.  With white at 100% and black at 0% the depth of modulation is 100%, a sharp image, but with white at say 65% and black at 35%, the depth of modulation is only 30%, a very unsharp image.

Consider that even with a lack of sharpness, as in the above demonstration, the native resolution of the chip, measured in pixels, remains the same. Throughout this hypothetical test, these images would have been recorded on the same camera at the same native resolution.  Simply stated, the chip uses the same number of pixels regardless of the image it sees.  If you shoot a flat color, a perfectly even tone, the native resolution of the chip is still 1080×1920 whether the image is only one tone or many tones. Therefore recorded images can be sharp or blurry without native chip resolution being the determining factor.

So what we call contrast or depth of modulation is a way of describing sharpness.  Cameras that have a narrow tonal range appear to be sharper, because they are “contrastier”.  Also cameras that are “detailed up” appear sharper as well.  But don’t be fooled. Again, sharpness is not resolution.

In the initial days of HD, and continuing even today, in many cameras, the dynamic range remains narrow.  This, coupled with the fact that some manufacturers like to ship their cameras detailed up, creates an image that is overly sharp.  The same kind of thinking occurred at the turn of the nineteenth century when carriage makers started building cars.  They weren’t cars as we know them today, but truly they were “horseless carriages”.  Similarly, many video manufacturers are still building “HD video cameras” not “digital acquisition film cameras”.  The difference is more than just the name and the recording media.

While sharpness is not a function of resolution, the visible effect of increased resolution makes it more evident, as finer detail can be observed because it is enhanced by false sharpness, incurred by the lack of a full tonal range and a higher than needed detail setting.

Cameras that have a wide dynamic range, with smooth tonal changes, typically have deep bit depth and a gamma curve to emulate a film response to light.  This means that the image is flatter, has less contrast, and yes, makes skin look smooth as film. Also, the detail settings are set low so that edges are soft, rounded and filmic, not crisp and video-like.  To put it another way, filmic images have painterly volume while traditional video images lean gently in the direction of graphic illustration and line drawing.

Now, with that background as our setting, it is no wonder that several years ago a camera with increased resolution made it’s way onto the scene to a mixed response. Easily the size of an old, blimped Mitchell camera, tethered to a bank of hard drives, it seemed like a step backwards in time, away from cinema verite.  With 4K resolution, its introduction at NAB was largely considered by many to be overkill.  Most considered it useful for scientific applications, but oh dear, what about those actresses?   What will they think of “HD ultra”?  That camera was the Dalsa Origin.

Today, the next iteration will be the Evolution, although most people will still refer to it as “the Dalsa”.  Time has produced many significant improvements to the first Origin and the current version boasts reduced size and weight, still with its optical through-the-lens viewfinder and an Arri-style rotating mirrored shutter, 24 white balance look-up tables, simplified touch screen menus, and a feature that allows for on-set visual image analysis.  Shortly, the camera will be freed from its tether and will support flash memory—512 Gigs to be exact.  These units will be called “Flashmags” and will record 20 minutes in RAW and 40 minutes in a lossless compression at 2:1.

But far and away the most significant feature of the Dalsa is it’s chip, for it is like no other.  It is a Bayer pattern chip.  This is the same style used by NASA in spy satellites and is in marked contrast to most other cameras.

Line One alternates photosites (output to pixels) RGRGRGRG while line two alternates GBGBGBGB.

A typical Bayer pattern

Each frame starts as a 16Mbyte file containing the RAW color information, then using a Bayer to RGB code unique to Dalsa, and with a complex measurement of neighboring pixels (not just the adjacent pixels), Dalsa also measures crosstalk in each of the colors where the other colors “leak” through.  The algorithim then goes about examining the photosites and eliminates color fringing on edges.

Because they make their own sensor in their own lab, they are able to take unique advantage of the silicon response in the CCD that most others are cannot do and DALSA takes every opportunity to exploit this advantage. While other manufacturers use photo diodes (lenslets) for gathering light, Dalsa rejects this chip design (raised circles within rectangles) as inefficient, because it ignores as much as 72% of the usable chip surface. Rather, they use a photosensitive substrate divided by tungsten wires.  By designing a chip in this manner they have accomplished three very important goals.  First, 86% of the chip is used to gather light (14% being covered by the wires).  This is referred to as fill factor.  Fill factor is what gives the Origin II its unique-to-the-industry 13 stop dynamic range.  Secondly, with this much “signal”, noise is virtually eliminated.  And lastly, since there are no lenslets to deflect light, causing white shading and color fringing issues with short wide angle lenses, this chip design eliminates the need to white shade the camera.

Completing the process of imaging, Dalsa employs a frame transfer CCD.  The CCD writes to the recording media an entire frame simultaneously rather than using the traditional interline method, thus eliminating the possibility for corruption of the image with lagging electrons that can increase what is called “fixed pattern noise.”  The film analogy is grain.

This is a very costly chip to manufacture. Recognizing this, Dalsa has followed the “Panavision” business model and does not sell, but only rents, their cameras.  This means that all cameras undergo tight maintenance and as improvements are made, continual upgrading, so that cameras are current and each camera goes out with a technician to support it.

Workflow follows a fairly traditional path.  Video and audio are recorded to an on-set data recorder then moved to a computer such as a Mac from which it is transferred to a post facility’s server as 4K RAW data where 2K HD versions are created for editing and preview.  An EDL from Final Cut or Avid is used to conform the RAW data in the server and then the conformed program goes out to a render farm to encode BAYER to RGB, facilitating the color correction of the conformed scenes and lastly to output to a laser printer to create the film neg and print, or to create a digital cinema package for theaters.

Currently available through their office in LA, they are expanding to other markets and will soon be available in New York.  Recognizing that they have a very unique product, Dalsa has initiated an educational outreach program and have been conducting classes, about 5 hours in length, in LA, Atlanta, Philadelphia and New York to familiarize potential users with their camera.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: