• Blogs
• Forums
• Tradeshows
• Partners
• White Papers
• Events

CMOS Latitude and More

Nov 11, 2006 2:04 PM

In the last installment, we looked at Sony’s implementation of CMOS technology, which it calls 3ClearVid. In its implementation, each CMOS element is rotated 45 degrees.

Through a complex process, three 1-megapixel ClearVid chips generate three 2-megapixel Virtual Sensors. This “interpolation” process increases each chip’s effective horizontal resolution by 175 percent while decreasing effective vertical resolution by 75 percent. The result is a measured resolution of approximately of 800 TVL/ph and 700 TVL.

Although 3ClearVid’s resolution is clearly interesting, it is important not to think of resolution is the most important aspect of the new HVR-V1—or of any camera.

Latitude

Latitude is defined as the range of F-stops that can be accurately recorded on film or by an electronic image sensor. To understand latitude, however, look at the image below from a single-chip camcorder.

The white dog’s fur, the woman’s skin, and even the grass have lost both color and detail. For a decade, 1/3in. camcorders have suffered from the inability to handle highlights as shown above.

Look closely at the image below from a Sony Z1. The bright blue sky looks perfect. Likewise, note the faces of people in the foreground. The Z1 was the first 1/3in. camcorder that correctly handled highlights. Now look at the doorways on the left side of the street that are not in direct sunlight. In the image we see them plunged into deep shadow so most detail is lost. To my eyes, of course, the west side of Ninth Avenue did not look this way. Although the buildings and people were not in direct sunlight, in the early afternoon of a very sunny day, they were fully visible.

In both images, we see that the camera has been unable to accurately capture the scene. By taking multiple exposures of very high-contrast scenes, I have experimentally determined that most 1/3in. camcorders need 2 to 3 stops of additional light latitude than they are able to provide.

Camcorders with larger chips are able to provide this additional latitude, which is why pro shooters prefer them. To match the latitude of film, an additional few stops of light range are required. This capability is available in the very expensive 2/3in. CCDs used in HD camcorders. These chips have the ability to capture a logarithmic 11-bar grayscale chart accurately, as shown below.

When this capability is not available, the result is compromised highlights, as shown below.

Or the result is compromised deep shadow, as shown below.

Gamma

To this point, I’ve focused on the image chips as being responsible for a camera’s latitude. However, recent advances in DSP technology enable cameras to manipulate an image to provide a more natural look. The primary DSP function is to support “gamma.” Gamma determines the relationship between light input and signal output. Gamma can be represented as a curve as shown below.

The DSP can also process the upper and lower light ranges independently. The upper range, called the “knee,” shapes the relationship between light input and signal output in a manner designed to minimize burned-out highlights. This function can be set to Auto or set manually — typically between 80 IRE and 100 IRE. A lower value enables greater apparent latitude.

A Black Stretch or Compress circuit handles dark grays and near-black. When Stretch is engaged, the relationship between light input and signal output is altered to prevent the loss of shadow detail. This also enables greater apparent latitude. (As you would guess, Compress makes shadows appear even darker.)

The diagram below shows the signal output from DSP as a function of light input from a sensor. It has a “lazy S” shape with two segments on either end of a more linear segment: (A-B), where Stretch/Compress occurs (Stretch shown), and (C-D), which is the knee.

The ability of DSP to manipulate the relationship between input and output makes it difficult to distinguish between an imaging chip’s inherent latitude and the additional latitude computed by the DSP.

As the series of screenshots below demonstrates, video recorded by the V1 has exceptionally wide latitude. Keep in mind that I cannot directly link the V1’s latitude to its use of CMOS rather than CCD chips — let alone to its use of 3ClearVid chips.

The “parade” display below shows a full range from about 2 IRE to 106 IRE.

The latitude of the scene below exceeded the range of my light meter. Moreover, the scene does not even include any really dark shadows. (It’s not clear that web photos can convey the video image.)

The HVR-V1’s wide latitude makes using zebra a bit old-fashioned. Below is a shot where, with zebra set at 100 IRE, the window area would be expected to have either no color or even no detail. Yet the camera captured the overexposed detail well.

Because zebra no longer provides adequate information, the V1 includes a histogram that displays information that is far more useful.

Contrast

The decade-long quest to create a “film look” has, at its origin, four aspects of video that many, but certainly not all, video professionals find displeasing. These include low latitude, high contrast, overly enhanced edges, and unnatural color. We have already discussed a lack of latitude that is manifest as blown highlights and/or crushed shadows. The second characteristic is the high contrast of video when compared to modern low-contrast film stock. As the image below demonstrates, the V1 is able capture images that do not have the “contrasty” video look.

High contrast increases apparent sharpness. For those who want increased contrast, Black Compress can be used.

An even more dramatic image can be created by enabling CineGamma 1, as shown below.

Edge Enhancement

Nothing screams “video” more than an overdose of edge enhancement. The whole point of HD is to provide an image that naturally carries fine detail. Thankfully, the HVR-V1 adds very little edge enhancement, so it is not necessary to dial down sharpness. Of course, some may misinterpret this as lack of resolution.

Colorimetry

To date, DV and HDV shooters have had to choose between a pleasingly warm picture and a more accurate but cool picture. This has often been called the “DV look.” JVC’s first-generation HDV camcorder, which used a unique Hybrid Complementary-Primary CCD, eliminated this look. Now, the V1 carries this a step further by creating an image that is neither warm nor cool — it's simply natural.

Latitude and More

Assuming, as I do, that Sony’s 3ClearVid CMOS sensors are responsible for the HVR-V1’s high imaging qualities, then CMOS has a very bright future. One can only imagine this technology implemented in larger chips.

© 2008 Penton Media, Inc.