With high definition coming later this year in the form of HD-DVD and Blu-ray Disc, the prospect of having to replace my collection of over twelve hundred DVDs with their high definition counterparts seems like a horrendous financial burden. I know what you're thinking. Hey, didn't he get all those DVDs for free? Not really. I've reviewed only about six hundred films, so the remainder of my collection represents quite an investment. A question occurred to me: might it be possible to extend the life of some of my standard definition DVDs by viewing them in a high definition format, upconverted from standard resolution by a scaler?

Now, I'm not fooling myself. You can't get something for nothing. An image's resolution is strictly dependent upon the picture data that exist within that image. There are some pretty clever algorithms that can accurately interpolate - estimate a new pixel between existing pixels - but such mathematical shenanigans cannot increase image resolution, even in scalers costing upwards of $60,000. What scaling to a higher resolution can do is reduce the visibility of the scan line structure and smooth the existing image, possibly making it more film-like. But it's that tendency to smooth that's always concerned me about scalers. My concern lies in the simple reality that to mathematically deduce the values for red, blue, and green for a newly created pixel, the math must rely on the values of the existing pixels that surround the position of the new pixel.

For example, the simplest such calculation would be an average of the RGB values, weighted by the distances between existing surrounding pixels and the location of the new pixel. And that would be equivalent to a video low-pass filter; the image would get noticeably softer. Fortunately there are more sophisticated mathematical algorithms to perform this task. The downside is that the more complex the calculations, the more powerful the processor it takes to perform them in real-time, and that causes the scaler to become more expensive.

The wonders of very large-scale integrated circuits and microelectronics have reduced such processes to the form of inexpensive chips. Both Faroudja and Teranex processes have been realized in integrated circuits. The advent of such circuits makes possible a slew of DVD players that have scalers built in. (Most of those have HDCP-compliant DVI connections for the scaled output. A few offer scaled video on component video outputs, but my understanding is that those products have been or are about to be withdrawn from the market.) But I can't take advantage of HDCP-compliant DVI outputs with my CRT front projector, and those scaling players are fixed resolution devices. I was looking for more flexibility.

I want to be able to control the scaler's output resolution to optimize the projected image. My CRT front projector is frequency agile and I want to find and use the resolution that corresponds to the projector's sweet spot. Unlike a fixed pixel projector, for which the output resolution of the scaler should be set to match the display device's native dimensions (in pixels), a CRT is a bit more complex. There are vertical and horizontal sweep rate compatibility issues; the scaler's output rates must be within the projector's ability. And the projector's optical focus and electronic focus can reduce the spot size just so much. If the scaler's signal resolution causes the projector to crowd the horizontal scan lines together to the point that the sweeping spot overlaps, instead of increasing resolution, the resolution actually goes down. So the goal is to run the highest possible resolution the projector is capable of accepting but without overlapping the scan lines.

Next, I want to be able to control the frame rate. Those of you who are familiar with the 3:2 pulldown process know that to convert film's 24 frames per second to video's 30 frames per second, it's necessary to repeat visual information to equate film-time to video-time. If you'd like to know more, please have a look at my article of five years ago entitled What the Heck is 3:2 Pulldown? Unsymmetrical repeating of visual information causes a subtle reduction in motion smoothness within the image: temporal asymmetry. In the motion picture theater, film's rate 24 frames per second is shown at 48 frames per second with a double-bladed rotating shutter in the projector; each frame is repeated once. This is temporal symmetry. Well, many frequency agile CRT front projectors can sync to a video frame rate of 48 frames per second, including mine. (A rate of 72 frames per second would also work well, but for high resolution images the resulting horizontal scan rate is beyond my projector's ability.) So I want a scaler that removes the 3:2 pulldown and supports 48 frames per second for standard resolution scaling (480i30 or 480p60 to 960p48 or 1080p48) and high definition scaling (1080i30 to 960p48 or 1080p48). That tends to limit my choices considerably.

As for inputs, I want analog component video, HDCP-compliant DVI or HDMI (that prepares me to scale HD-DVD and Blu-ray Disc to 1080p48), and SDI. SDI is Serial Digital Interface. It is a pure digital signal that avoids conversion from digital to analog as it passes from player to processor. This is a very clean signal path that avoids problems like ringing in the player's anti-aliasing filter. It is not a digitally encrypted interface, so scaler manufacturers are permitted to process the signal and output the RGB-HV signals my projector needs. Alas, even though several DVD players used to be offered with that interface, it would seem that only outrageously expensive professional players offer it today. Fortunately, there are several conversion kits available and several vendors willing and able to install such kits in a limited number of existing models of commonly available players.

I also want aspect ratio flexibility, the ability to change the input aspect ratio on the fly for full screen DVD supplements that should be window-paned in high definition widescreen. The decision tree led me to Key Digital's HD Leeza scaler.

This $3499 scaler is very versatile, with two DVI inputs, an SDI input, and a slew of analog video inputs, including RGB-HV, two HD component video inputs, two SD component video inputs, two S-Video inputs, and two composite video inputs. So an ancillary plus is that the unit can also act as a video switcher. The outputs include DVI, RGB-HV, and component, but as with all such devices, the HDCP-compliant HD Leeza disables its analog outputs when the active DVI input detects an HDCP-encrypted signal.

The scaler will modify the video from interlaced to progressive, and based on the user's preferences, provide a choice of scan rates (48, 50, 60, 72, and 75 Hz), allow the setting of both the output and input aspect ratios (the latter helpful for those pesky non-anamorphic DVD supplements), perform format conversion (the unit acts as a transcoder, accepting one form of video and outputting another, like component to RGB-HV - with the understanding that HDCP material won't appear on the analog outputs), and most significantly, offer the choice of a slew of output resolutions.

The scaler supports a dazzling array of resolutions. Conventional DVD and HD formats of 480p, 720p, and 1080p are supported. For fixed pixel display devices, we have 852x480, 1024x512, 1024x1024, 1280x768, 1440x788, and 1366x768. There are some standard PC display formats of 640x480, 800x600, 1024x768, 1280x1024, and 1400x1050. There is one digital cinema format of 1920x1200. And a few other handy resolutions, like 852x576, 1280x960, 1440x960, and 1365x1024. As I previously mentioned, the goal is to find the resolution that either matches a fixed pixel display or falls at a CRT display's sweet spot. I pessimistically expect conversion artifacts, and since DVD is 720x480, I'm particularly interested in 1440x960, precisely double the pixels in each dimension. That resolution may have fewest artifacts when converting existing DVDs.

Those with frequency agile CRT-based front projectors should be cautioned to avoid driving their displays with signals that are too fast. But if the user's display fails to lock on any scaled images, Key Digital provides a very handy way of forcing the scaler to reset to an unchallenging standard resolution. The user's display should then restore the picture and the setup can continue.

HD Leeza has quite a few available video control adjustments as well: brightness; contrast; image position; image size; 2-D enhancement (uh oh, watch out for halos); gamma; the previously mentioned aspect ratio for both in & out; non-linear stretch for filling a 16:9 screen with 4:3 images; the choice of letterbox or window-pane; and blue and red saturation levels to play with color temperature.

The scaler is quite logical to set up. After making the appropriate connections (the scaler has BNC bulkhead connectors, but Key Digital thoughtfully provides 75-ohm RCA to BNC adaptors), the user sets up the output parameters for the display device and can then independently set parameters for each input. The scaler's memory structure allows the storing of user preferences for each input device. Very handy.

I really appreciate the scaler's ability to specify input and output aspect ratios. This allows me to selectively windowpane full screen DVD supplements while upconverting the video; this not only looks better, but it's a faster and more convenient way of restoring a proper presentation for 4:3 material than switching my projector's display mode. And I only have to maintain the convergence for one screen format.

But rather than take up your time with an extensive description of how to connect the scaler and set it up, for those who may be interested, allow me to suggest that you read or download the product manual.

For my evaluation, Key Digital very kindly also provided a loaner Panasonic DVD player, sans the chroma upsampling error and with an SDI output.

Before I leave the subject of operation, I must mention that the HD Leeza's IR remote control command to turn on is identical to the Lutron Spacer light controller's command to turn the lights on or off. If you'll recall from Part 3, I've installed Spacers in my theater so I can control the lights from my Marantz remote control. So since the last IR command I send before leaving the room is always to turn the lights off, I found that the previously shut down scaler would then spring back to life; this isn't critical, but it is a tad inconvenient. Since I couldn't change the IR command for either unit, I had to find another solution. Fortunately, I had installed a solid-state switch in a power strip; the switch is controlled by the 12-volt DC signal that's generated by my B&K Reference 30 digital audio decoder when it's powered on. I use that switched power strip for my Extron devices (they have no power switch), optical to coax digital audio adaptors (no power switch), and the theater's multiple channels of audio power amplification (no remotely controllable power switches). By plugging the HD Leeza into that switched power strip and leaving the scaler's power switch in the On position solved the problem.

By now, you know how much I abhor edge halos. If a DVD has them, I'll complain; if the disc is clean, I'll praise. And if you read Part 6 of this series, in which I describe replacing my DVD player, you know that I went to great lengths to minimize hardware-generated halos. Since the interface to my projector remains analog, I must rely on the scaler's D/A conversion process and anti-aliasing filter to behave.

So even before I tried to remap any pixels, the first thing I did was to load my copy of the AVIA Guide to Home Theater into the Panasonic player and navigate to the half-black / half-50 IRE test pattern. I examined the standard resolution waveform on my oscilloscope and, in particular, the transition. I found a single half cycle of 16% overshoot that had a duration of only 0.04 usec, about a pixel wide. Impressive. Halos should not be an issue. I then switched to a 100 IRE full screen white test pattern and checked for color balance problems that could modify the color temperature; I found none.

It was time to push some pixels around. I select the 1440x960 resolution with a 60 Hz frame rate, since my projector already had a memory setting for 1080i30, for which the convergence should be quite close. The ringing is even lower in this mode; the overshoot drops to 5.5%, but the duration of the overshoot does not drop appreciatively. And I noticed my first trade-off; the slope of the transition - how fast the video switches from black to gray - took longer in the upconverted video than in original resolution. I measured a little under three pixels. This implies a process equivalent to low pass filtering.

So perhaps checking the frequency response would be helpful in determining how the scaler might affect the images. I switched to the AVIA's stepped video sweep test pattern and I found a hint of what I suspected. High frequencies are very slightly suppressed. I measured the KD Leeza's analog green channel (of the RGB-HV output) while driven by each of the two players using two different inputs. I also compared those readings with the component video outputs of the two players.

DVD's video bandwidth extends beyond 5 MHz, so for the disc's highest frequencies the attenuation would be even greater. So my original suspicion that the remapping algorithm might have a tendency to affect high frequencies seems to be correct. But the numbers imply a modest effect, so the question becomes: are these anomalies visible while watching program material?

I disconnected the RGB-HV signals from my oscilloscope and reconnected them through my Extron switcher to my projector. My interconnects are from BetterCables, so I'm confident that any misbehavior will be from the electronics, not due to cable issues.

As tempting as it is to throw a few reference DVDs into the player, I want to observe how the scaler affects projected test patterns first - the scaler is now in the 1440x960p60 output mode. I've just finished touching up my convergence for the new scan rates. My first impression as I navigate through the AVIA screens is smoothness. When I watch my Denon DVD-9000 player from my seating position, its 480p60 raster is on the threshold of visibility. But this upconverted raster is completely invisible. As I've done with HDTV broadcasts, I walk closer to the screen to see if the raster is perceptible. It's not. Due to the interlacing of 1080i30 HDTV, its raster is more visible than the 960 progressive lines produced by the scaler. But the images are so smooth that I fear that the video will be soft; the test patterns prove my fears to be unfounded.

The sharpness and resolution test patterns are clean, with terrific apparent resolution and bandwidth, right to the frequency limits of the patterns. Halos are of sufficiently low amplitude not to be intrusive, and the smallest detail is clearly conveyed. Small print is cleaner and easier to read. The 6.75 MHz circle looks like it has fine contrast, but upon closer examination, I find that the contrast of the vertical stripes varies slightly with position (implying an interference pattern), a very subtle effect. I can wait no longer; it's time to see how program material looks. I pull out a few superb DVDs for subjective impressions of playback from the Panasonic over SDI to the HD Leeza: Starship Troopers Superbit; Star Wars Episode II: Attack of the Clones; Star Wars Episode IV: A New Hope; and, Ghost in the Shell 2: Innocence.

My initial impression of the images is very smooth but slightly soft, so I set out to compare more thoroughly before and after conversion. I happen to have two copies of Innocence on hand; I load one in the Panasonic and the other in my Denon. I transcode the Denon's component video to RGB-HV and connected those signals to my Extron switcher. I then connect the HD Leeza's RGB-HV output to the switcher's other input channel. I am now able to play both discs, synchronize them, and switch back and forth. Unfortunately, my projector takes a few seconds to recognize a different scan rate and switch to the corresponding memory, so switching from the standard 480p60 and the special 1440x960p60 causes a delay that makes detailed comparisons impossible. My solution is to freeze-frame the same images on both players; comparisons then become easy.

I select a highly detailed scene from Innocence, the elephant-float gently gliding on a canal in the middle of a dense urban jungle. The ornate float is covered with intricate designs. Diminutive figures stroll on walkways on either side of the canal. And I was shocked to find that the float's designs and the tiny expressions on strollers' faces were clearer in upconverted form than they were in the DVD's native resolution. I was amazed.

I continue to play, loading each of my reference discs into the Panasonic player. For each, the scaler creates lovely, film-like images, more convincingly than the progressive video from the Denon. But a very modest - perhaps deceptive - impression of softness remains. The HD Leeza's video enhancement adjustment is not to be used; halos become visible before any noticeable improvement in the images. Ultimately, I had to accept that the effect is likely a product of slightly reduced high frequency response and the reduced slope of video transitions I found. So small object detail is improved, but slower video attack times takes a very modest edge off of very fine textures and transitions.

As mentioned earlier, I am particularly interested in the 48 frames per second progressive video rate, so I adjust the HD Leeza and capture the scan rates, setting up my projector to accept 1440x960p48. I want to see how temporal symmetry looks. Alas, something is amiss. Either the scaler is unhappy or my projector is. The images are slightly unstable, exhibiting a vertical jitter and motion artifacts on horizontal, sharp edges within the images. Disappointed, I have no choice but to switch back to 60 Hz progressive.

This provokes me to look for motion artifacts. I reload the AVIA and check the moving zone plate test patterns; the screen come alive with strobing within all the patterns, quite messy in fact. But when watching program material, with one exception, no such motion artifacts are visible. That one exception is vertically scrolling closing titles against a busy background; some of the edges of the letters sparkle as they move. No such effect is apparent with a black background. Odd.

As for resolution, my projector clearly demonstrates the improved spatial resolution of 1080i30 over 720p60 for HD broadcasts, so I don't feel that I'm pushing the projector too hard by scaling to a 1440x960 format. Regardless, I experiment with other resolutions and discover that for some, Moire patterns appear in the video, depending on the content. For example, at 720p60, the scene in which C3PO talks Uncle Owen into buying him from the Jawas, an interference pattern is visible within C3PO's neck rings in a moderate close-up. The interference pattern is not visible when scaling up to 1440x960p60. Time does not permit the exploration of the many other resolutions. Experimentation will continue after this is published.

I also experiment with HD Leeza's DVI input and output. I visit a friend's home with the scaler and the SDI DVD player. He owns a fixed-pixel HDCP-compliant projection system and HDCP-compliant DVI signal sources. The HD Leeza is set up for 60 Hz progressive video and his projector's native resolution. I use reliable digital cables (DVI, HDMI) from BetterCables to avoid any dropped bits between his HD sources and the scaler and the scaler to the projector. I can't stress enough the need for high quality DVI and HDMI cables. As video bandwidth and associated bit rate go up, the more critical the cables become. Start dropping bits and you'll see artifacts. An investment in good cables is the best one you can make in this digital age, and I continue to admire the value of the cables from this vendor - they reliably and accurately transport signals without excessive expense to the consumer. But I digress.

We connect the HD Leeza's component video output to his projector's component video input. We can now switch back and forth between a pure digital path and an analog path to the projector. But, HDCP encrypted signal sources do indeed cause the scaler to shut down its analog outputs, so comparing how the scaler deals with HD over DVI and component is not possible. We switch HD component video signal source cables around to make comparisons between the projector's internal scaler and the HD Leeza, and although the delays in switching make it difficult, I believe I see fewer motion artifacts with the HD Leeza. Comparing scaled up DVDs is a little easier. After connecting the player's SDI port to the HD Leeza, the HD Leeza's DVI output to the projector, and the player's component video to the display's component input, we're able to switch back and forth. A subtle improvement in low level video noise and the welcome impression of improved image cleanliness gives the edge to the digital signal path, and I believe that the HD Leeza did a better job at scaling than the internal circuitry of the projector.

So are the trade-offs worth the perceived improvements? After living with the scaler for a couple of weeks, I'd have to say yes. Will the HD Leeza extend the life of standard resolution DVDs? Certainly. Will I be able to avoid buying the HD-DVD and Blu-ray Disc formats? Hell, no. As I pointed out at the beginning of this little adventure, you can't get something for nothing. A fine scaler like the HD Leeza has the potential to make an existing standard definition DVD look very good indeed, but it cannot recover finely grained detail lost to DVD's lower resolution.

Is the HD Leeza good enough to find a permanent home in my theater? Absolutely. I'm very impressed with the scaler's flexibility and the quality of the images it produces. The presentation through the scaler comes closer to simulating film than my Denon's progressive video output, even with a very slight suppression of the video's highest frequencies. The HD Leeza is welcome to stay in my video chain and if your system might benefit from the addition of an external scaler, I'd suggest you give this unit serious consideration.