used Sharp's service menu to finetune performance further still, but straight-from-the-box the set's gray scale imperfections were slight. Ditto for great out-of-the-box gamma results, where the set read 2.26 tracking, just a hair off the ideal of 2.2. The result: images with accurate skin tones, neutral blacks (thanks to proper color temperature), and detail that is retained in dark areas. Drawing another example from Sideways, the brightly lit scene with Miles and Jack in the vineyard was crisp, with the sunlit faces and row after row of the emerald vines accurately rendered—not washed out.
The geometry was near perfect with just a very slight amount of pincushion distortion. Color matrix (component video) was very good with no red exaggeration. Focus was highly uniform and the lenses created very small amounts of chromatic aberrations at the edges and corners, with artifacts typically less than a pixel wide, which is excellent. Detail tests, however, gave very mixed results. The set was able to pass a "one pixel on/one pixel off" test (vertical rows) using a native 720p RGB test signal, fed through the PC input, though the pattern was more gray/white than black/white, indicating that the bandwidth was not as wide as it should have been.
Even more interesting was the crosshatch test, which has a white one-pixel dot in the middle of each square—or at least that is how it should appear. The latest TI DLP chips (720 and 1080) display square pixels turned 45° from vertical, so the "square" dots appear as single diamonds, but these are still basically the same single-pixel dots. However, when I would input any type of signal other than 720p RGB (480i, 1080i component, etc.) many of the single diamonds would turn into twins, with side-by-side, overlapping pairs of diamonds showing in most parts of the screen. Over-and-under pairs appeared in some cases, and in still other areas, an "X"-shaped grouping of diamonds. I speculate this phenomenon is a scaling artifact and contributes to an apparent lack of image sharpness on all signals (other than 720p RGB signals, which bypass the scaler). For example, on HD Net's Designing America, which is a pristine HD program videotaped in 1080i, the Sharp— unlike other similarly-sized HD displays I have tested—lacked sufficient detail to reveal individual hair strands of reporter Kim Schlegal or hostess Jocelyn White.
On the deinterlacing test, which contains a SMPTE 133- type pattern of mixed horizontal white lines (odd scan lines) and black lines (even scan lines), the 1080i whiteline/ black-line pattern should appear just at it does when displayed at the Sharp's native 720p resolution, but in this case, it did not. Instead, the test pattern displayed as a square that strobed back and forth between solid black and solid white. What was happening? I talked with a Faroudja spokesperson (this set includes Faroudja deinterlacing) who confirmed my assessment that the display was "bobbing" from one 540 line-field to the next one, displaying an upconverted 720p white field (made up of the white/odd horizontal lines) for 1/60th of a second, and then a black field (black/even lines) for 1/60th of a second, and so forth. Simply stated, this set was upconverting only half of the frame picture data that is within the 1080 line signal. The spokesman added that there are other Faroudja chips available that "weave" the two 540 line fields into one 1080-line frame prior to downconversion to 720p. A Sharp spokesperson also confirmed my observations using the same SMPTE 133 test pattern.
I viewed a substantial amount of high-definition programming including 1080i from a JVC D-Theater DVHS tape player, as well broadcast and satellite content. My overall impression was it appeared somewhat soft. On HD programs such as CBS' Two and a Half Men and Without a Trace, the Sharp's picture just did not have the depth and snap that I have observed with other HD displays, such as the 34" Sampo direct view I had on hand. This is too bad, given that the optical qualities of the 56DR650 are some of the best I have seen on any rear projector.
Sharp's 56DR650 provides solid and reasonably flexible display functions, but it omits three features that may be important to you now or in the future. Specifically, the set is not CableCARDready, does not provide an interactive on-screen programming guide, and has no ability to send tuner signals from the set to an HD hard drive or upcoming HD DVD disc recorder (via IEEE 1394). The 56DR650 delivers mixed performance results. I liked the set's remote control, and its tuner performed well. The set allows dot-for-dot viewing of native 720p RGB and performed exceptionally with this signal. Its Faroudja chip did a fine job of upconverting 480i signals with freedom from jaggies, and it performed well when detecting the 3:2 cadence of film-based content. On the other hand, it did only a fair job handling noisy signals and avoiding false contouring on 480i DVD discs. The level of black was very deep and neutral out of the box and white was very bright, without crushing very light detail. The big let down, though, was the set's 1080i performance whose image sharpness fell short of what I expected from a display with such otherwise fine basics as rich color, accurate gamma, and excellent optics.
I imagine typical TV viewers would be tickled pink by the performance of the 56DR650, but I think serious videophiles may desire more.