Sensor war

The level of complexity that goes into making a high-end camera requires a lot of background knowledge and experience. As a result, new entrants to the camera market will fail to make a real impact, while established manufacturers will be those driving the technology forward, argues Klaus Weber of Thomson Grassvalley.
Thomson Grassvalley uses the Xensium CMOS sensor in its Infinity Digital Media Camcorder (DMC), which it claims provides performance that rivals CCD t
Thomson Grassvalley uses the Xensium CMOS sensor in its Infinity Digital Media Camcorder (DMC), which it claims provides performance that rivals CCD t
Today?s HD cameras have pixels that are a quarter of the size of pixels found in SD cameras, which means one needs four times as much light to achieve
Today?s HD cameras have pixels that are a quarter of the size of pixels found in SD cameras, which means one needs four times as much light to achieve

Share

The level of complexity that goes into making a high-end camera requires a lot of background knowledge and experience. As a result, new entrants to the camera market will fail to make a real impact, while established manufacturers will be those driving the technology forward, argues Klaus Weber of Thomson Grassvalley.

Quite unlike anything else in our industry, the television camera is an extraordinary instrument. Its level of complexity requires a sizable bank of background knowledge and experience, which is why new entrants to the camera market fail to make a real impact - and why established manufacturers drive the technology forward in response to demands from users.

As content creators continue to challenge the creative limits of HD production, each major stage of the television camera system - imagers, digital signal processing, and transmission - will see continued development over the next few years.

 

"While you could build a 4k CMOS sensor, it would ‘cost’ you at least a couple of stops down in sensitivity. That would require throwing a lot more light at any production — directors would hate the restrictions, and producers would hate the additional costs. The financial and creative price is simply too high for 4k CMOS to be a viable option. - Klaus Weber, Thomson Grassvalley."

In 1987, the frame transfer (FT) CCD imager was introduced by Thomson at a time when other manufacturers were using interline transfer (IT) CCDs, but we were convinced that the FT chip was superior.

Frame transfer technology gives users the substantial benefit of lossless switching inside the sensor itself. This dynamic pixel management provides the only practical solution for perfect switching between 4:3 and 16:9 aspect ratios.

Now, HD DPM+ allows us to shoot natively at any standard-definition (SD) or high-definition (HD) resolution on the same chip. Frame transfer CCDs have a higher dynamic range than other imagers and better anti-aliasing performance, but it is this high-density resolution (and the ability to switch that resolution on the chip) that is the real winner.

Critics contend that the required use of a mechanical shutter is a detriment to FT CCDs, but such arguments are not justified. In fact, we have seen thousands of FT-based cameras in service over a long period of time with no significant issues.

Now, there is another imager to consider. There has been plenty of discussion about using CMOS sensors to replace CCDs, but the technical issues are enormous. For example, it is generally accepted that CMOS sensors need a mask that is 20 times finer than the required pixel size.

A 1080p sensor in a 2/3-inch camera has an individual pixel size of 5µm square, which means the mask has to be 0.25µm x 0.25µm, a level that has only been achieved in the last few years.

Dynamic range will especially be a key issue in development. The latest CMOS-based cameras from established manufacturers have a dynamic range equivalent to eight or nine f stops. The latest film stocks still offer much more, at around 14 f stops.

Of late, though, research and development teams have close to 20 f stops of latitude with a CMOS-based digital camera. Once this progress can be integrated into commercial products, it will totally transform the way television is produced as video will have more dynamic range than film.

Such a vast dynamic range requires intelligent signal processing downstream of the imager, but that is not likely to be a significant challenge. Cameras already apply multiple layers of processing to images from the sensor, as well as adjustments for creative and technical corrections, to deliver professional quality pictures in real time.

One of the issues on any sensor is consistency between each photosite. Typically, an imager will exhibit different black levels from pixel to pixel, which appears as fixed pattern noise.

One technique that can be used for CMOS sensors is to read each pixel twice, once at the beginning of the frame exposure and once at the end, and then use those two readings to cancel out the fixed pattern noise. This method reduces the consistency problem by about a factor of four.

Photosite output will also vary, with some producing over voltage (leaking pixels) and some under voltage (dark pixels). These effects can be caused by a variety of factors, from operating temperature to the ageing of the imager.


 

Another area that needs to be addressed is camera customisation. Today, manufacturers are adding a large number of colour matrices into the camera, from skin tone compensation to colour correction for modern fluorescent lighting systems, so a cameraperson can set up the exact look they want.

We cannot defeat the laws of physics completely, so some correction is always going to be necessary. That said, digital signal processing is not a workaround for poor imagers. A professional camera should only use carefully selected imaging chips that are as close to perfect as possible.

Better imagers not only provide better pictures, but also extend the active working life of a camera. One should feel confident that their camera will provide excellent service for at least 10 years without the need for expensive optical block replacement.

The delivery stage is much more than a simple cable connection. It also has to carry audio, intercoms, sound, tallies, control signals, power and much more, often over very long distances.

Five years ago, more than 95% of cameras used triax. Studios and venues like sports arenas invested heavily in triax infrastructure. Then, along came HD with its larger bandwidth requirements, which threatened to make existing triax installations obsolete.

While some camera manufacturers moved immediately to fibre, others recognised that users would rather not replace their installed base of triax cable. A new triax system was consequently developed to allow users to get 30 MHz of bandwidth with reasonable cable lengths of a kilometre or more.

As a result, standard HD cameras can still use existing triax cables.

Unfortunately for triax, the future for cameras includes 3 Gb/s data for 1080p work, as well as 3X (and perhaps more) super slo-mo cameras. A new fibre system had to be developed that could handle the increased bandwidth requirements over long cable runs.

The first fruits of this new, data-centric fibre development can be seen in use this summer, connecting 3X super slo-mo cameras at the UEFA EURO 2008 football group playoffs, quarter finals, semi-finals, and the final match as well as at the Beijing Olympic Games.

Fibre technology will continue to evolve, but we may not be the end of the road for triax. Our own research shows that the potential of transmitting digital camera signals for 1080p50 operations and more using triax over distances that exceed even the cable length of the existing analogue triax systems is there.

Today's HD cameras also have pixels that are a quarter the size of pixels found in SD cameras, which means one needs four times as much light to achieve the same output from the chip. Sensor quality is certainly improving, but sensitivity and signal to noise ratio are still worse in an HD camera.

As a result, the engineering challenge is to minimise sensitivity and noise issues.

One solution would be to increase the size of the sensor, but that is not really practical. The industry has standardised on 2/3-inch sensors and has a huge inventory of lenses to fit.

Broadcast lenses assume a three-chip optical block; the standard for back focus assumes that the three colours will be in different planes. Plus, even if you could persuade users to replace them, a sports zoom lens for a 1-inch camera would probably weigh close to 100 kg.

While you could probably build a 4k CMOS sensor, it would 'cost' you at least a couple of stops down in sensitivity. That would require throwing a lot more light at any production - directors would hate the restrictions, and producers would hate the additional costs.

The financial and creative price is simply too high for 4k CMOS to be a viable option.

Over the next few years, television camera development will most likely focus on achieving the absolute best performance from HD, particularly 1080p, while respecting established industry equipment standards.

Continuing research - perhaps completely moving to CMOS imagers, deeper and faster signal processing, and more flexible interconnections - will concentrate on cleaner, sharper, and more controllable HD imagery.

Klaus Weber is director of Product Marketing Cameras, Thomson.

Most Popular

Editor's Choice

StarzPlay’s tech powers India’s Lionsgate Play
Provides managed services, developed by StarzPlay, to mark successful launch of Lionsgate Play ...
Subscribe to Digital Studio Middle East
The latest editor of Digital Studio Middle East, the foremost and most respected platform for ...
Dubai's Faisal Hashmi working on feature film
10 short films in 10 years has earned Faisal Hashmi international plaudits at film festival, ...

Don't Miss a Story