Between The Lines

ARTICLE

BETWEEN THE LINES
Improving Large-Screen Video Images

by Peter H. Putman, CTS

As the computer and video industries continue along their slow, unsteady and often stumbling paths to media convergence, it's helpful to step back and consider what has been accomplished in the field of video production to date. Lightweight, high-resolution cameras let us record images under all kinds of lighting. Powerful, low-cost desktop workstations make it possible to edit high-quality video and stereo sound. Disc- and tape-based component video recording systems deliver the best picture and audio quality.

And where does all this hard work wind up? Played back into a color television monitor that hasn't changed very much in the past twenty years. Sure, our cameras can capture 600 to 800 lines of detail. Our editing equipment can sample the components of a video signal at a high bit rate, preserving much of that quality. New formats like Digital Video Disc deliver video and audio from the component domain, using MPEG-2 compression. It makes for a better picture, but the end product still comes up short when viewed as a composite signal with about 430 lines of resolution.

What else is there? The HDTV powers that be are still debating aspect ratios, interlacing vs. non-interlacing, channel spacing and audio issues. Rather than sitting around and twiddling our thumbs waiting for the next high-resolution paradigm, why not investigate some of the newer display technology that can help "soup up" the quality of our video images.

VIDEO WITH SMARTS

The marriage of composite video and computers has been an odd fit from the start, primarily because one system (NTSC video) is limited by a maximum bandwidth and a single scan rate, while the other (RGB Video) is not. If computers had to conform to the standards that broadcast TV pictures did, we'd still be playing around with 8088 microprocessors and CGA color displays.

Since there are so many parts to a composite video signal, it takes a microprocessor with lots of memory and processing speed to analyze and break down those components, then accurately reconstruct them. Fortunately for us, microprocessors and ASICS (Application Specific Integrated Circuits) get faster, smaller, and cheaper every year. We can put those same pieces of silicon to work for us by playing all kinds of tricks with video signals.

The key to improving video image quality is an understanding of the design and limitations of NTSC signals. Like an overstuffed van, this 4+ MHZ-wide signal carries luminance (black & white) and chrominance information in the same carrier. The decoding circuitry in a television monitor has to strip out the color information from the black & white information and perform a series of mathematical calculations to wind up with red, green, and blue picture components. At the same time, the horizontal sync and vertical refresh rates must be extracted from the composite signal.

Many of the picture artifacts we've come to know and love in composite video images are created during the color decoding process. Since the burst signal containing color information lies at 3.58 MHz, filtering it out also removes some of the luminance information at that frequency, which contains 300 to 400 lines of picture detail. (The more lines of detail, the higher the frequency and the wider the system bandwidth must be.) Comb filtering won't remove as much detail, but may result in crawl between light and dark transitions.

If our circuitry is smart enough and powerful enough to cleanly execute the decoding process, we've now displaying better-looking video than the old 20" monitor in the rack. Better yet, if we have access to a multisync monitor or video projector with component video inputs, we can execute more black magic and improve the apparent resolution of the image.

DOUBLE YOUR FUN

The second step to improving video images is to convert the interlaced lines in an NTSC signal to progressively-scanned lines, kicking up the horizontal scan frequency in the process. The resulting signal should have no flicker, a finer perceived line structure, and less color moire and crawl. That's precisely what a line doubler does, converting the horizontal scan rate of a composite signal to 31.5 kHz from 15.75 kHz, while breaking out the red, green, blue, and sync information into separate channels.

The concept of line doubling was developed several years back in an attempt to create an in-between step between composite video and HDTV, known as Improved Definition Television (IDTV). IDTV was not terribly successful, due to the poor quality of NTSC decoding and line doubling in the IDTV sets sold at that time. But that was then, and this is now.

Outboard line doublers are available today from numerous manufacturers, and range in price from as little as $1,500 all the way to $20,000, with commensurate steps in quality. Most LCD and DLP data/video projectors use some form of line-doubling, often just deinterlacing the signal and using tricks like parallel line addressing to eliminate flicker. More sophisticated line doubler designs look at odd and even scan lines to interpolate motion changes from field to field, thereby eliminating motion artifacts such as stair-stepping.

Both projectors and projection monitors are now coming to market with plug-in line doublers. I tested Sony's new VPH-D50Q CRT projector for several weeks in my studio, and it's equipped with an optional EXB-DS10 plug-in line doubler that can be activated from the Sony remote control. The advantage of this plug-in line doubler is that it works from the existing composite video input, while keeping the projector's component input connections free for other uses.

The difference between an interlaced and non-interlaced video image is quite noticeable with the EXB-DS10. Using a Snell & Wilcox #2 Zone Plate test pattern as a reference, picture flicker was eliminated and the line structure of the picture appeared less distinct. The EXB-DS10 does a good job of retaining image sharpness during the doubling stage, which was a tricky job for earlier line doublers to pull off. No color artifacts were observed, and motion artifacts were minimal. The only problem I saw with this doubler was high-frequency "ringing" around black lines on the test chart.

Although I used a CRT projector for my tests, any monitor or projector that supports a 31.5 kHz horizontal and 60 Hz vertical scan rate can be used with a line doubler. This is also the scan rate for PC-based VGA (640 x 480) video cards, so if you have access to a component RGB switcher, you can switch back and forth seamlessly from video playback to computer images and back again. This technique is used by video staging companies to blend electronic speaker support and video modules at large meetings, but it works just as well with a multiscan computer monitor equipped with audio playback.

Line doublers come in all shapes and sizes. Most provide both S-Video and RGBS outputs, but the degree of video signal processing will vary from brand to brand. Pay less money, and you'll probably see more artifacts in your pictures. Pay more money, and you'll have finer control over signal levels and better-looking pictures. No matter which way you go, start out with the best source material you can (read: component video) and you'll be happier with the end result.

IF YOU CAN'T SEE 'EM, MAKE 'EM UP...

Since microprocessors and ASICs are so smart and nimble, why not push them even further? Instead of just interpolating motion between odd and even video fields, why not interpolate lines of information that don't even exist between scan lines? That might seem like a tall order, but it's exactly what happens in a line quadrupler. Two additional picture scan lines are added to a line-doubled image, which has the effect of not only increasing picture detail, but boosting screen brightness as well.

The microprocessor circuitry in a quadrupler looks at both odd and even video fields in a frame buffer, along with all the odd and even scan lines. Then, artificial scan lines are created to fill in between each odd and even line, and displayed progressively as before. Now, consider this: To scan 262.5 picture lines in 1/60 of a second in an interlaced system, we need a horizontal scan rate of 15.75 kHz. To scan all 525 lines in 1/60 of a second requires 31.5 kHz. So, to scan twice that number of lines (1050) in 1/60 of a second, we need to double the horizontal scan rate one more time to 63 kHz. That's really getting up there!

To put a line quadrupler to work, we need a multiscan monitor or projector with a comparable scan rate, plus the ability to resolve about 1100 lines of detail. When you stop to think about it, we've essentially created a system for displaying high-resolution video images that approximates 1125-line HDTV, without the 16:9 aspect ratio - sort of a back door approach to the high-resolution video problem, except that we are still working with 100% analog source video - not digital.

Line quadruplers and multiscanning CRT projectors work together particularly well when improving the look of letter boxed videotapes, laserdiscs and DVDs. Depending on the widescreen aspect ratio used, the letter boxed portion of the image represents as little as 270 to 330 lines of the total picture area, so projecting this type of image on a large screen makes the scan lines very noticeable. Using a quadrupler doubles the horizontal line resolution to 550-660 lines, which is a dramatic improvement.

(As an aside, it is possible to connect a line quadrupler to an LCD projector. But you'll need to have a 1280 x 1024 native resolution to see the entire picture - XGA displays don't have enough pixels to accommodate all 1050 horizontal lines. If your XGA projector has defeatable image scaling, you will get a picture underscanned to the TV safe action area.)

For my tests, I used the same laserdisc source into Extron's Sentosa line quadrupler, which also generates RGBHV signal connections to the VPH-D50Q. Separate color saturation and hue adjustments are provided on the Sentosa's control panel for fine-tuning. A contrast control handles the peak white level, while a position control centers the image.

After calibration, the line structure on the quadrupled image had all but disappeared, but pictures were noticeably softer than the line-doubled versions. Using the zone plate to measure detail, I found the Sentosa started to have trouble with line flicker around 300 horizontal lines of detail. Contrast was a bit lower, and there was a shift in white balance to a warmer image.

No color artifacts were seen on the Sentosa, but I did notice a few motion artifacts from time to time, particularly when the camera passed objects with distinct angular shapes. The overall look resembled a photograph with coarse grain and soft focus. I have seen sharper and cleaner images from other line quadruplers (notably Faroudja's VP-400), but they cost four to five times as much as the Sentosa.

ONE SIZE FITS ALL

Another approach to enlarging and resizing video - scaling - is now emerging. The need for scaling has increased with the influx of flat-screen projectors and monitors that have odd-sized native resolutions, like 800 x 600, 832 x 624, and 1024 x 768. Although all three formats measure 4:3 in aspect ratio, none have a pixel count that works out to an even multiple of the 525-line NTSC standard.

Several companies have now introduced video scalars for commercial and home theater applications, including Faroudja's Presentation Plus Scalar, Snell & Wilcox's Interpolator and Communications Specialties' Deuce. These systems connect between a standard PC video card and the display monitor or projector, automatically syncing to the horizontal scan and vertical refresh rates. Interlaced video then fed to the scalar can be overlaid on Windows screens, as well as resized horizontally and vertically to different aspect ratios.

Early attempts at video scaling resulted in all kinds of picture artifacts, such as pixelization and aliasing. Many late-model SVGA and XGA LCD/DLP projectors have problems with full-frame video, and a couple produce downright awful scaled images. Unfortunately, only a few manufacturers offer a scan selector switch to allow the display of video at 525 non-interlaced lines. While this results in a smaller picture surrounded by black bands, image quality is improved considerably.

In my October 1997 review in Video Systems of SVGA projectors, I noted that video quality varied considerably among the review models. Several manufacturers have achieved better results by using custom video scaling circuitry manufactured by Genesis Microchips, while others have developed proprietary circuitry. Some of the best projectors displayed full-frame video that appeared slightly soft, but had no noticeable pixel structure. Other projectors had crisper-looking images, but couldn't shake the pixelization/aliasing problem.

For comparison, I ran some tests on Sony's VPL-S500U, one of the three top performers in scaling video to SVGA. The results were quite different than what I saw with the line doubler/VPH-D50Q combination. The resulting image was slightly softer with a thinner, more noticeable line structure than the line-doubled CRT produced. I did notice a small amount of pixelization, but less aliasing around the scan lines.

Color saturation and contrast were about the same as the CRT projector, making allowances for the VPL-S500U's metal halide projection lamp. Some motion artifacts were observed, but they were no more severe than those seen on the EXB-DS10 and Sentosa. Viewed at a more normal distance, the VPL-S500U's video was not as sharp as the VPH-D50Q/EXB-DS10 combination, but was equal to the Sentosa in crispness with the difference being the lack of scan lines on the Sentosa.

CONCLUSIONS

First of all, each of my test platforms produced video images superior to conventional 525-line interlaced pictures, mainly due to the elimination of flicker. Next, there's no question that video looks better when line-doubled, as long as a good doubler is used. Line quadrupling is a bit trickier due to the interpolation, but the highest quality quad boxes do produce a cinema-like image, best viewed using a CRT or ILA projector. If you have access to a multiscanning monitor/projector and line multiplier, use the combination whenever possible to show your work. This is especially true for producers working on large meetings or exhibits.

As far as scaling goes, it's more of a necessary evil that doesn't necessarily improve an image, but lets you take full advantage of your flat-screen display's native resolution. In my experience, outboard scalars do a noticeably better job than built-in models and produce pictures comparable to the best line doublers and quadruplers. You'll pay for that quality, too - both the Snell & Wilcox and Faroudja scalars come with pretty hefty price tags, as did the first line doublers and line quadruplers several years ago.

However, look for scalars (like everything else in the large-screen display marketplace) to come down in price over the next two to three years. Once the problems with pixelization are resolved - and they will be - the ability to resize images to a wide variety of screen resolutions and scan rates, combined with easy interfacing to computer video cards will eventually make video scalars the more desirable product.

©1998 Peter H. Putman / Intertec Publishing, Inc.
This article originally appeared in the February 1998 issue of Video Systems.


						ARTICLE
					BETWEEN THE LINES Improving Large-Screen Video Images


					by Peter H. Putman, CTS As the computer and video industries continue along their slow, unsteady and often stumbling paths to media convergence, it's helpful to step back and consider what has been accomplished in the field of video production to date. Lightweight, high-resolution cameras let us record images under all kinds of lighting. Powerful, low-cost desktop workstations make it possible to edit high-quality video and stereo sound. Disc- and tape-based component video recording systems deliver the best picture and audio quality. And where does all this hard work wind up? Played back into a color television monitor that hasn't changed very much in the past twenty years. Sure, our cameras can capture 600 to 800 lines of detail. Our editing equipment can sample the components of a video signal at a high bit rate, preserving much of that quality. New formats like Digital Video Disc deliver video and audio from the component domain, using MPEG-2 compression. It makes for a better picture, but the end product still comes up short when viewed as a composite signal with about 430 lines of resolution. What else is there? The HDTV powers that be are still debating aspect ratios, interlacing vs. non-interlacing, channel spacing and audio issues. Rather than sitting around and twiddling our thumbs waiting for the next high-resolution paradigm, why not investigate some of the newer display technology that can help "soup up" the quality of our video images. VIDEO WITH SMARTS The marriage of composite video and computers has been an odd fit from the start, primarily because one system (NTSC video) is limited by a maximum bandwidth and a single scan rate, while the other (RGB Video) is not. If computers had to conform to the standards that broadcast TV pictures did, we'd still be playing around with 8088 microprocessors and CGA color displays. Since there are so many parts to a composite video signal, it takes a microprocessor with lots of memory and processing speed to analyze and break down those components, then accurately reconstruct them. Fortunately for us, microprocessors and ASICS (Application Specific Integrated Circuits) get faster, smaller, and cheaper every year. We can put those same pieces of silicon to work for us by playing all kinds of tricks with video signals. The key to improving video image quality is an understanding of the design and limitations of NTSC signals. Like an overstuffed van, this 4+ MHZ-wide signal carries luminance (black & white) and chrominance information in the same carrier. The decoding circuitry in a television monitor has to strip out the color information from the black & white information and perform a series of mathematical calculations to wind up with red, green, and blue picture components. At the same time, the horizontal sync and vertical refresh rates must be extracted from the composite signal. Many of the picture artifacts we've come to know and love in composite video images are created during the color decoding process. Since the burst signal containing color information lies at 3.58 MHz, filtering it out also removes some of the luminance information at that frequency, which contains 300 to 400 lines of picture detail. (The more lines of detail, the higher the frequency and the wider the system bandwidth must be.) Comb filtering won't remove as much detail, but may result in crawl between light and dark transitions. If our circuitry is smart enough and powerful enough to cleanly execute the decoding process, we've now displaying better-looking video than the old 20" monitor in the rack. Better yet, if we have access to a multisync monitor or video projector with component video inputs, we can execute more black magic and improve the apparent resolution of the image. DOUBLE YOUR FUN The second step to improving video images is to convert the interlaced lines in an NTSC signal to progressively-scanned lines, kicking up the horizontal scan frequency in the process. The resulting signal should have no flicker, a finer perceived line structure, and less color moire and crawl. That's precisely what a line doubler does, converting the horizontal scan rate of a composite signal to 31.5 kHz from 15.75 kHz, while breaking out the red, green, blue, and sync information into separate channels. The concept of line doubling was developed several years back in an attempt to create an in-between step between composite video and HDTV, known as Improved Definition Television (IDTV). IDTV was not terribly successful, due to the poor quality of NTSC decoding and line doubling in the IDTV sets sold at that time. But that was then, and this is now. Outboard line doublers are available today from numerous manufacturers, and range in price from as little as $1,500 all the way to $20,000, with commensurate steps in quality. Most LCD and DLP data/video projectors use some form of line-doubling, often just deinterlacing the signal and using tricks like parallel line addressing to eliminate flicker. More sophisticated line doubler designs look at odd and even scan lines to interpolate motion changes from field to field, thereby eliminating motion artifacts such as stair-stepping. Both projectors and projection monitors are now coming to market with plug-in line doublers. I tested Sony's new VPH-D50Q CRT projector for several weeks in my studio, and it's equipped with an optional EXB-DS10 plug-in line doubler that can be activated from the Sony remote control. The advantage of this plug-in line doubler is that it works from the existing composite video input, while keeping the projector's component input connections free for other uses. The difference between an interlaced and non-interlaced video image is quite noticeable with the EXB-DS10. Using a Snell & Wilcox #2 Zone Plate test pattern as a reference, picture flicker was eliminated and the line structure of the picture appeared less distinct. The EXB-DS10 does a good job of retaining image sharpness during the doubling stage, which was a tricky job for earlier line doublers to pull off. No color artifacts were observed, and motion artifacts were minimal. The only problem I saw with this doubler was high-frequency "ringing" around black lines on the test chart. Although I used a CRT projector for my tests, any monitor or projector that supports a 31.5 kHz horizontal and 60 Hz vertical scan rate can be used with a line doubler. This is also the scan rate for PC-based VGA (640 x 480) video cards, so if you have access to a component RGB switcher, you can switch back and forth seamlessly from video playback to computer images and back again. This technique is used by video staging companies to blend electronic speaker support and video modules at large meetings, but it works just as well with a multiscan computer monitor equipped with audio playback. Line doublers come in all shapes and sizes. Most provide both S-Video and RGBS outputs, but the degree of video signal processing will vary from brand to brand. Pay less money, and you'll probably see more artifacts in your pictures. Pay more money, and you'll have finer control over signal levels and better-looking pictures. No matter which way you go, start out with the best source material you can (read: component video) and you'll be happier with the end result. IF YOU CAN'T SEE 'EM, MAKE 'EM UP... Since microprocessors and ASICs are so smart and nimble, why not push them even further? Instead of just interpolating motion between odd and even video fields, why not interpolate lines of information that don't even exist between scan lines? That might seem like a tall order, but it's exactly what happens in a line quadrupler. Two additional picture scan lines are added to a line-doubled image, which has the effect of not only increasing picture detail, but boosting screen brightness as well. The microprocessor circuitry in a quadrupler looks at both odd and even video fields in a frame buffer, along with all the odd and even scan lines. Then, artificial scan lines are created to fill in between each odd and even line, and displayed progressively as before. Now, consider this: To scan 262.5 picture lines in 1/60 of a second in an interlaced system, we need a horizontal scan rate of 15.75 kHz. To scan all 525 lines in 1/60 of a second requires 31.5 kHz. So, to scan twice that number of lines (1050) in 1/60 of a second, we need to double the horizontal scan rate one more time to 63 kHz. That's really getting up there! To put a line quadrupler to work, we need a multiscan monitor or projector with a comparable scan rate, plus the ability to resolve about 1100 lines of detail. When you stop to think about it, we've essentially created a system for displaying high-resolution video images that approximates 1125-line HDTV, without the 16:9 aspect ratio - sort of a back door approach to the high-resolution video problem, except that we are still working with 100% analog source video - not digital. Line quadruplers and multiscanning CRT projectors work together particularly well when improving the look of letter boxed videotapes, laserdiscs and DVDs. Depending on the widescreen aspect ratio used, the letter boxed portion of the image represents as little as 270 to 330 lines of the total picture area, so projecting this type of image on a large screen makes the scan lines very noticeable. Using a quadrupler doubles the horizontal line resolution to 550-660 lines, which is a dramatic improvement. (As an aside, it is possible to connect a line quadrupler to an LCD projector. But you'll need to have a 1280 x 1024 native resolution to see the entire picture - XGA displays don't have enough pixels to accommodate all 1050 horizontal lines. If your XGA projector has defeatable image scaling, you will get a picture underscanned to the TV safe action area.) For my tests, I used the same laserdisc source into Extron's Sentosa line quadrupler, which also generates RGBHV signal connections to the VPH-D50Q. Separate color saturation and hue adjustments are provided on the Sentosa's control panel for fine-tuning. A contrast control handles the peak white level, while a position control centers the image. After calibration, the line structure on the quadrupled image had all but disappeared, but pictures were noticeably softer than the line-doubled versions. Using the zone plate to measure detail, I found the Sentosa started to have trouble with line flicker around 300 horizontal lines of detail. Contrast was a bit lower, and there was a shift in white balance to a warmer image. No color artifacts were seen on the Sentosa, but I did notice a few motion artifacts from time to time, particularly when the camera passed objects with distinct angular shapes. The overall look resembled a photograph with coarse grain and soft focus. I have seen sharper and cleaner images from other line quadruplers (notably Faroudja's VP-400), but they cost four to five times as much as the Sentosa. ONE SIZE FITS ALL Another approach to enlarging and resizing video - scaling - is now emerging. The need for scaling has increased with the influx of flat-screen projectors and monitors that have odd-sized native resolutions, like 800 x 600, 832 x 624, and 1024 x 768. Although all three formats measure 4:3 in aspect ratio, none have a pixel count that works out to an even multiple of the 525-line NTSC standard. Several companies have now introduced video scalars for commercial and home theater applications, including Faroudja's Presentation Plus Scalar, Snell & Wilcox's Interpolator and Communications Specialties' Deuce. These systems connect between a standard PC video card and the display monitor or projector, automatically syncing to the horizontal scan and vertical refresh rates. Interlaced video then fed to the scalar can be overlaid on Windows screens, as well as resized horizontally and vertically to different aspect ratios. Early attempts at video scaling resulted in all kinds of picture artifacts, such as pixelization and aliasing. Many late-model SVGA and XGA LCD/DLP projectors have problems with full-frame video, and a couple produce downright awful scaled images. Unfortunately, only a few manufacturers offer a scan selector switch to allow the display of video at 525 non-interlaced lines. While this results in a smaller picture surrounded by black bands, image quality is improved considerably. In my October 1997 review in Video Systems of SVGA projectors, I noted that video quality varied considerably among the review models. Several manufacturers have achieved better results by using custom video scaling circuitry manufactured by Genesis Microchips, while others have developed proprietary circuitry. Some of the best projectors displayed full-frame video that appeared slightly soft, but had no noticeable pixel structure. Other projectors had crisper-looking images, but couldn't shake the pixelization/aliasing problem. For comparison, I ran some tests on Sony's VPL-S500U, one of the three top performers in scaling video to SVGA. The results were quite different than what I saw with the line doubler/VPH-D50Q combination. The resulting image was slightly softer with a thinner, more noticeable line structure than the line-doubled CRT produced. I did notice a small amount of pixelization, but less aliasing around the scan lines. Color saturation and contrast were about the same as the CRT projector, making allowances for the VPL-S500U's metal halide projection lamp. Some motion artifacts were observed, but they were no more severe than those seen on the EXB-DS10 and Sentosa. Viewed at a more normal distance, the VPL-S500U's video was not as sharp as the VPH-D50Q/EXB-DS10 combination, but was equal to the Sentosa in crispness with the difference being the lack of scan lines on the Sentosa. CONCLUSIONS First of all, each of my test platforms produced video images superior to conventional 525-line interlaced pictures, mainly due to the elimination of flicker. Next, there's no question that video looks better when line-doubled, as long as a good doubler is used. Line quadrupling is a bit trickier due to the interpolation, but the highest quality quad boxes do produce a cinema-like image, best viewed using a CRT or ILA projector. If you have access to a multiscanning monitor/projector and line multiplier, use the combination whenever possible to show your work. This is especially true for producers working on large meetings or exhibits. As far as scaling goes, it's more of a necessary evil that doesn't necessarily improve an image, but lets you take full advantage of your flat-screen display's native resolution. In my experience, outboard scalars do a noticeably better job than built-in models and produce pictures comparable to the best line doublers and quadruplers. You'll pay for that quality, too - both the Snell & Wilcox and Faroudja scalars come with pretty hefty price tags, as did the first line doublers and line quadruplers several years ago. However, look for scalars (like everything else in the large-screen display marketplace) to come down in price over the next two to three years. Once the problems with pixelization are resolved - and they will be - the ability to resize images to a wide variety of screen resolutions and scan rates, combined with easy interfacing to computer video cards will eventually make video scalars the more desirable product. ©1998 Peter H. Putman / Intertec Publishing, Inc. This article originally appeared in the February 1998 issue of Video Systems.