Production switchers

Are 3-D and 3G related? It seems like one or the other comes up in nearly every advertisement or conversation I have with a vendor lately. Though on the surface they seem to be unrelated, they are in fact, at least theoretically, intertwined.

Defining 3G and 3-D

3G refers to a 3Gb/s SDI signal, generally meant to conform to SMPTE 424. The actual data rate is twice SMPTE 292, or 2×1.485Gb/s = 2.97Gb/s (nominal, conventionally tied to 29.97 frame rate by taking 2.97/1.001, or 2.967). Since S424 is simply a data stream, it can actually carry a number of things, including two independent 1080i30 signals, or one 1080p60 signal, as well as digital cinema 2K images and other formats. The former case is interesting for 3-D. A separate method for doing two 1080i30 signals was also standardized by SMPTE using two S292 interfaces, often called dual link, and can be found in SMPTE 372. But two links are more complicated to manage than a single link.

3-D actually refers to stereoscopic imaging and display. In general, the left and right eyes are independently captured and transmitted. It is useful to note that stereoscopic imaging replicates the eye's sensor arrangement, with depth information created by the distance between the imagers and the angle (vergence) of each taking lens. The details do not matter much in the context of this article except that when one does special effects on a stereoscopic image, some strange things can happen at image transition points. For instance, inserting graphics becomes a pretty complicated affair. Consider where in three dimensions you want the graphics to appear, and how that relates to the background image and potential cues for depth that it contains.

Switcher advancements

Inside a digital production switcher, three things are of paramount importance: interface bandwidth, memory requirements and processing power. The processing speed must be able to handle high-speed signals, but clearly handling either a 60fps signal or two 30fps signals will require double the frame memory whenever a process is being executed. However, one option is to use two separate processing paths with more conventional requirements when there are independent signals, as is done with stereoscopic imaging.

In practical terms, this was the short-term route switcher manufacturers took to achieve 3-D switching. Most devices, such as cameras and replay devices, use separate paths (dual-link S372) for interconnection today, making anything else a bit more complicated than necessary. Envision controlling two switcher processes with one panel. That makes it pretty simple to do a left eye and right eye independently. All the sources must be in both crates, and memory registers must be identical.

It is impractical to use separate switchers for two eyes, so the practical 3-D switcher has one crate with two signal paths. This means, assuming you're working with dual-link systems, limiting the number of “usable” inputs to half of those in a conventional system. But manufacturers have started increasing the number of inputs to account for the stereoscopic nature of signals. There are now production switchers with much larger input capacities to accommodate precisely this need. Input capacities have grown in some designs to 120, or even 160, primary inputs. Internally, a single M/E is split into a primary and slave with two outputs for left and right. In some applications, this dual-layer switching allows for two separate M/Es within each bank of the switcher, but when stereoscopic production is involved, they process left- and right-eye signals identically.

Carrying this one step further, wouldn't it be elegant if instead of dual link, which requires twice the router, twice the patch panel space and twice the number of DAs, one might use S424 with both left and right eye on one cable? In the last 18 months, there has been considerable movement toward replacing 1.5Gb/s interfaces with 3Gb/s interfaces on many system components, including routing, distribution, multiviewers, and even master control and branding systems. For a long time, the cost differential between transceiver chips for S292 and S424 was considerable, but today the cost is almost equivalent. This has made it easier for manufacturers to transition products to the latest S424 capable components. Of course, there is more to it than reduced manufacturing cost, and the additional features still lead to a bit of premium in most products. There is, however, positive return on investment for picking 3G over 1.5G products.

Some aspects of production are not quite ready for 3-D. In switchers, a 3-D digital effects system really works in a single plane with 3-D appearance to the manipulated pictures. A “true” 3-D DVE would have to process the content such that it appeared different as the picture moved in true 3-D space. But our eyes may get confused by the visual cues that creates. For now, you can place a DVE panel anywhere in relation to the screen plane, but it will not move in three dimensions.

On the other side of the coin, in practical terms, a 3-D camera is really two conventional cameras on one mechanical rig. Each has an S292 interface and carries discreet left- or right-eye signals. Marrying the separate signals into one cable costs extra and increases complexity. Until true 3-D cameras are widely available, 3G for 3-D may not make sense.

Beyond stereoscopic applications, there is another huge advantage to 3G interfaces on production switchers. When HD first became real, we struggled to deliver 1080i30 images to the home. 19.3Mb/s was barely enough and on some content not nearly enough. But MPEG encoding technology has improved to such an extent that we receive excellent pictures at home with much lower bit rates. Inexorably, the manufacturing and production communities, along with consumer electronics manufacturers, have been pushing the limits of HD capability, with a lot of emphasis on 1080p60 capability. That can be done with a dual-link system, but it is difficult and was a strong force behind the movement to S424's higher data rate. Since 1080p60 has twice the frames per second, it requires higher bandwidth at baseband, but progressive scan pictures compress more easily than interlaced content. Some say the gain is at least 15 percent and perhaps 25 percent. This means a lower bit rate for the same quality, or better quality for modest increases in bit rate.

Future-proofing a facility

This leads to an interesting possibility. Many experts have said that 720p60 is superior for motion rendition, which makes sense, but 1080i30 has higher spatial resolution and hence is preferred by some. If 1080p60 could be distributed and delivered to the home, it might become a premium service today and a standard feature in the future, which is a better match to consumer sets since they are inherently progressive scan today and they weren't 10 years ago. ESPN and others have started building new facilities with 1080p infrastructure even if they deliver 720p today. It makes good sense for the future.

In this multiformat world, buying an expensive production switcher without future capability to do 1080p may be a less expensive option only for the short term. Add to that the ability to do 3-D, should that become important, and you see the impetus behind what may seem like excess capability today but smart investing for the future.

John Luff is a broadcast technology consultant.

Send questions and comments to: