Lately it seems as though every component in a fully tricked-out home theater system wants to dink with the video — the DVD player, the receiver, the TV. Usually whatever is being done is described as some sort of upconversion. What does that mean, though? And perhaps more important, is it always a good thing?

The types of video processing commonly described as upconversion fall into three basic categories:
• Transcoding. Changing one signal format to another, such as composite-video to S- or component-video.
• Deinterlacing. Converting an interlaced video signal to progressive-scan, such as from 480i to 480p or 1080i to 1080p.
• Scaling. Changing the signal from one display resolution to another, such as from 480p to 720p.
Understanding how each of these works will help you decide when upconversion is necessary or desirable (and when it's not) and where in the signal chain the processing is best applied. At the end, I'll give you some basic rules for figuring all that out.

Transcoding: The Background
Although video transcoding might seem to be a recent innovation — a feature built into some A/V receivers — it actually has a long history; every TV set ever sold has had to do it just in order to work. That's because video has never been transmitted in the same format in which it is acquired and displayed. TV cameras and displays are fundamentally RGB devices, meaning that they use three primary colors — red, green, and blue — in various proportions to represent all the others, including black, white, and all the shades of gray in between. We see fine detail mainly in grayscale, or black-and-white, however, so using full-bandwidth color signals to represent it isn't very efficient.

The solution is component-video, which consists of a full-bandwidth black-and-white, or luminance, signal, and two narrower-bandwidth "color-difference' signals, created by subtracting the luminance signal from the red and blue signals. TV sets have reciprocal color-decoder "matrices" that add and subtract these three signals in various proportions to regenerate an RGB signal to drive their displays. The letter used to designate the luminance signal is Y, so the color-difference signals are Y-R and Y-B, which are abbreviated to Cr and Cb for digital video and Pr and Pb for analog video. Hence, digital component-video is called YCrCb and analog component video is called YPrPb.

Component-video has two key benefits. The first, which is applicable to all forms of video transmission and storage, is efficiency: A component-video signal requires much less bandwidth than the RGB signal from which it was created to deliver a picture of equivalent perceived quality. The other goes back to the development of color television in the 1950s. RCA used component-video to piggyback color onto a regular black-and-white TV signal in such a way that B&W sets would not recognize the extra information while new sets could extract it to display full-color pictures.

And there's where we get to the processes that led to the other types of analog video connections we have today and eventually to video transcoding in A/V receivers and preamplifiers. (You were starting to wonder, weren't you?) RCA's engineers had to figure out a way to attach the color-difference signals without disrupting the standard B&W signal. They achieved this by phase-modulating the color-difference components onto a single subcarrier, creating a combined chrominance (C) signal. The subcarrier resides at 3.58 megahertz on a standard 4.2-MHz NTSC B&W signal, which together are known as a composite-video signal. An NTSC broadcast signal is created by frequency-modulating the audio onto another subcarrier at 4.5 MHz and then amplitude-modulating the entire thing onto an RF (radio-frequency) carrier for transmission. An ordinary color television set pulls that all apart in order to deliver the picture and sound.

Unfortunately, every step along the path from component-video to NTSC broadcast signal entails some sacrifice in quality, so when non-broadcast video sources, such as VCRs and laserdisc players, came on the scene, it was natural to look for ways to avoid some of those steps. First up was sending audio and composite-video via separate connections instead of combining them into an RF signal. Then came the S-video, or Y/C, connection, which keeps the luminance and chrominance separate instead of combining them into a composite signal. And with the advent of DVD, which carries video in digital component format, came the analog component-video connection. Analog component-video is created from the digital original by running the latter through video digital-to-analog (D/A) converters, but the most recent addition to the connection zoo is HDMI, which can carry digital component-video in its native form — perfect for digital component-video sources such as DVD, HD DVD, and Blu-ray Disc players, HDTV tuners, and satellite and digital-cable boxes.

Transcoding: The Bottom Line
The upside to all these different types of video connections is that they enable you to eke the last ounce of performance out of your various video sources. The downside is that it's a lot of connections. Just look at the back of a contemporary A/V receiver. It's nice to be able to run one cable from your receiver to your TV instead of half a dozen or more. Transcoding, more commonly called component-video or HDMI upconversion, makes that possible. A receiver or preamp with component-video upconversion will transcode composite-video to S-video and S-video to component-video, so regardless of what type of analog video signal comes in, it will be available at the component-video output. HDMI upconversion takes the process one step further, running the analog component-video signals through analog-to-digital (A/D) converters to yield digital component-video at the HDMI output.

What's most important to remember about video transcoding is that it's a convenience feature, not a performance upgrade. Converting a composite-video signal to component-video will not make it as good as it would have been if it had never been reduced to composite form. And it is not likely that a receiver will do a better job of the conversion than your TV would if you made all the connections there.

Deinterlacing: The Background
Another relic of our analog TV heritage is the technique known as interlaced scanning. In NTSC TV, a video stream carries 30 complete frames per second (fps), a frame being a complete still picture created by 480 active scan lines running horizontally across the screen. (The total number of lines in an NTSC frame is 525, but that includes lines in what is known as the vertical blanking interval, or VBI, which carry no picture information.) Each frame is split into two fields, each of which contains every other scan line. The fields are transmitted and displayed sequentially, one every sixtieth of a second, so that the first field of a frame is completely scanned, and then the lines of the second field are scanned between those of the first. This is known as interlaced scanning.

Interlacing is surprisingly effective and serves its original purpose of getting more resolution out of the available transmission bandwidth. But if you compare a scene shot and displayed in standard analog format, now known as 480i, with the same scene displayed with the same number of lines using progressive scanning — in which all the lines in each frame are displayed sequentially instead of being divided into two interlaced fields — the progressive-scan version will look cleaner and smoother. That format is called 480p, the number indicating the active scan lines and the "i" or "p" the scanning method.

The process of converting video from interlaced to progressive scanning — 480i to 480p or 1080i to 1080p — is called deinterlacing. Originally introduced as an option for improving picture quality, especially from DVDs, deinterlacing is now a necessity for most HDTVs, few of which support interlaced scanning at all. That's a consequence of the transition away from CRT-based sets to "fixed-pixel" display technologies, such as plasma, LCD, DLP, and LCoS, which work by flashing a complete frame on the screen at a time; what used to be scanning lines are now pixel rows. So today deinterlacing is less a feature that can improve your TV picture than an essential process that can make the picture worse if it isn't done right.

Although it seems straightforward enough to combine two fields of a frame and display them all at once, there are a couple of gotchas. The first is that when video is shot in interlaced format — which is how most TV and video cameras operate — the fields are shot sequentially, with the second field of a frame acquired a sixtieth of a second after the first. Any motion that occurs between the two will cause "jaggies" if the two fields are just slapped together and displayed simultaneously. Consequently, a deinterlacer must incorporate sophisticated motion-compensation techniques to achieve good results with typical video-originated material, and some are distinctly better in this regard than others.

The second potential issue arises with material originally shot on film. When film is transferred to interlaced video (for broadcast or DVD mastering), it must be converted from its native frame rate of 24 fps (frames per second) to the 30 fps employed for TV in most of the world via a method called 2:3 (or 3:2) pulldown, which pads out the sequence with a repeated field every other frame. For example, imagine four film frames, A, B, C, and D. When this 24-fps sequence is transferred to 30-fps interlaced video, each frame is split into two fields, which are organized in a sequence like this: A1, A2, B1, B2, B1, C2, C1, D2, D1, D2, and so on.

The beauty of video originated from film it that it can be deinterlaced perfectly to 60-fps progressive-scan format — but only if the deinterlacer accurately detects and compensates for the 2:3 pulldown. Then each field is reunited with its mate to restore the original film frame, and, following our example above, you wind up with a sequence like this: A, A, B, B, B, C, C, D, D, D, and so on. If the deinterlacer doesn't handle the pulldown correctly, however, it will create some video frames out of fields from two different film frames, causing an ugly artifact called "combing" if there is any motion between those original frames. (A deinterlacer might be able to fudge this by not switching to film mode at all, but even if it had excellent video-mode performance, resolution would suffer.) This is why 2:3 pulldown compensation is so important in progressive-scan DVD players and HDTV sets.

Deinterlacing: The Bottom Line
Deinterlacing is one of the most critical video-processing steps in today's HDTVs, all of which incorporate circuits for this purpose. But virtually all DVD players available now also have deinterlacers, as do an increasing number of A/V receivers and preamplifiers. You may therefore be faced with the question of where in the chain to perform deinterlacing for certain sources. Since most HDTVs have good deinterlacers, leaving the job to your TV is usually a reasonable default choice. For DVDs, the job is better handled in the player if, and only if, it has a very good deinterlacer. Most cheapie progressive-scan models are mediocre in this department, however. Look for an indication that your player uses a deinterlacing chip from a company known for performance in this category, such as Faroudja, Silicon Image, Silicon Optix, or Gennum. Deinterlacing in a receiver or preamp will seldom make sense unless it is also scaling the image.

Scaling: The Background
An HDTV set has to handle at least four basic video formats: regular old 480i standard-definition (SD) for conventional analog broadcasts and videotapes, 480p SD (mainly from progressive-scan DVD players), and the two widescreen high-definition (HD) formats, 720p and 1080i, which provide much greater picture detail. An HDTV set should, therefore, be able to accommodate inputs in a number of scan formats and in both 4:3 and 16:9 aspect ratios for standard-definition signals (4:3 is not used for high-definition broadcasts).

It's possible to design a CRT display to handle all of those formats directly, but since it's cheaper to convert some formats to others than to make a full-bore multi-scanning monitor, most rear-projection and direct-view CRT sets take the conversion approach. And in the case of fixed-pixel displays, such as LCD, DLP, LCoS, and plasma, all incoming signals must be converted to a progressive-scan format that exactly matches the display's pixel array. Most CRT-based HDTVs work at 480p and 1080i and convert every other incoming signal to one of those native formats. That usually means 480i gets bumped up to 480p and 720p gets converted to 1080i. (Because 720p actually has the highest data bandwidth and horizontal scan rate, it is easier from the display-design standpoint to convert it "up" to 1080i than to step 1080i "down.") The process of converting between scan formats is known as scaling, and interlaced signals must be deinterlaced prior to any other processing, which is one reason deinterlacing performance is so critical in HDTVs.

When converting to a resolution higher than that of the incoming signal — from 480p (852 x 480 pixels, horizontal by vertical, in a 16:9 frame) to 720p (1,280 x 720 pixels), for example, or either of those resolutions to 1080p (1,920 x 1,080) — a scaler has to interpolate between pixels in a frame to create new ones. It is not a simple stretching process. In fact, it usually will result in a completely new set of pixels mathematically generated from the originals. Good scalers use sophisticated alogorithms to arrive at optimum pixel values not only in the context of the current frame but also with respect to preceding and succeeding frames, so that motion is smooth and natural looking. Among the biggest challenges to a scaler is noise in the signal, such as from analog videotape or a poor-quality cable or broadcast TV transmission, which upconversion can accentuate. One of the marks of an outstanding scaler is good performance with noisy sources.

Not all scaling is upconversion, however. For example, if you send the 1080p output from an HD DVD or Blu-ray Disc player to a display with a lower resolution (720p, 1,366 x 768, 1,024 x 768, etc.), the set will have to scale the video down. This involves a sort of inverse-interpolation process known as decimation, the goal being to reduce the number of pixels while preserving as much information as possible. With more and more devices offering video scaling, including DVD players, A/V receivers and preamps, and of course HDTV sets, opportunities for yo-yo scaling — upconversion followed by downconversion followed by upconversion again — are growing. Generally speaking, it is best not to scale a video signal more than once unless absolutely necessary.

Scaling: The Bottom Line
Scaling is hard to do, and bad scaling can look really, really bad (especially if it starts off with mediocre deinterlacing). Historically, good scalers have been very expensive, even if all they did was line-double 480i to 480p. And the very best standalone scalers, from companies such as Faroudja, Key Digital, Runco, Pixel Magic, and Anchor Bay, are still pricey. The good news is that the growing need for video scaling has led to substantial progress further down the food chain — a trend that will surely continue. Still, before you buy an HDTV set, cast a critical eye on how it looks with a variety of input signals. Pay special attention to what the set does with ordinary analog signals from cable or broadcast TV, which tend to give crummy scalers the biggest fits. Look particularly at what happens around the edges of moving objects. (Problems are often most apparent on slowly moving objects in the background.) Jagged or fuzzy edges or halos around objects are a bad sign (although these can also be caused by excessive compression in digital cable or satellite signals). If you already have an HDTV set that you think is not so hot in this regard, you might be a candidate for an outboard scaler or a receiver or preamp that features high-performance video scaling (look for names such as Silicon Optix, Gennum, or Faroudja).

Rules to Process By
Here are some basic guidelines to help you navigate the wilderness of video upconversion:

• Always start with the best connection you can. HDMI (or DVI) usually is best, followed by component-video, S-video, composite-video, and RF, in that order.
• Remember that video transcoding in an A/V receiver or preamplifier is a convenience, not a magic picture-improver. Converting a composite-video signal to component-video or HDMI will not make it better.
• You can't get a good picture on today's fixed-pixel (plasma, LCD, DLP, or LCoS) HDTVs without good deinterlacing. Cheap progressive-scan DVD players may do a worse job of it than your TV, so unless you know what's in your player and are confident of its quality, try it with both 480i and 480p output to see which looks better on your set.
• Avoid unncessary scaling. This is a particularly important consideration if you have an upconverting DVD player or a receiver or preamp with built-in video scaling capability. If you set the output on one of these devices to anything other than your display's native resolution, the display will have to re-scale the signal. And if you have, for example, a 1,366 x 768 plasma and your upconverting DVD player or receiver, as is common, supports only ATSC-standard resolutions such as 720p and 1080i, that will not be possible. If you can't match the upconversion resolution to the display resolution, don't go beyond simple deinterlacing — let the display handle the scaling.
• Never convert a 720p HDTV signal to 1080i unless you absolutely have to. Interlacing a 720p signal loses, irretrievably, its most special quality. If you are using an HD cable or satellite TV box, set its video output to "native" mode if it has one. That will send video out in the same format it was broadcast. If the box requires you to select an output resolution, choose 720p unless you have a CRT-based HDTV; then you should choose 1080i. If you have a 1080p display, you might want to try both 720p and 1080i and see which looks better to you on most of the programs you watch, or you could switch the box's output resolution based on the channel you're watching (720p for 720p networks such as ABC, ESPN, and Fox, 1080i for 1080i networks such as NBC and CBS).

Standalone Video Processors
2:3 or Not 2:3?
Inside a Video Scaler
The Progressive Tradeoff
Basic Cable: Choosing the Right Cables for Your System
Back to Homepage
What's New on S&V