On DMDs, screen-door effects and creating 4K...
In the body of this review I indicated that the problem with much pixel shifting technology is that the pixels are just too big. If a pixel is shifted a half-pixel width to its right, most of its left side will overlap the position where it previously was, and most of its right side will overlap the pixel which was previously to its right. (The previous pixels are no longer being displayed, but the persistence of vision in your eyes will make it look like they are still there. DLP projectors do everything over time, rather than simultaneously, and can only work due to this property of human vision.)
Yet with this projector, as our test photo (left) shows, there is clearly resolution down to the UHD level, even if it’s not all that clean. Let’s try to explain this at the tech level.
The Digital Micromirror Device used in this projector appears to be the Texas Instruments DLP470TE, a 0.47-inch model with 1920 x 1080 resolution. It has a pixel pitch (the distance from an edge of one pixel to the equivalent edge on the next) of 5.4 micrometres, or thousandths of a metre. As it happens, this appears to be a ‘cut down’ version of the DLP660TE DMD, the 0.66-inch ‘4K’ model, which also has a pixel pitch of 5.4m. (It doesn’t really have 4K’s worth of pixels, but 2716 by 1528 pixels, which it displays twice.)
The DLP470TE chip is newly developed; indeed it isn’t yet (as we write) fully listed on the site of the manufacturer Texas Instruments. But it shares its size with the 1080p DLP470LE, while TI’s other 1080p DMD model is the DLP650NE, a 0.65-inch model with a pixel pitch of 7.6 micrometres, nearly 41% larger.
Which brings us to the ‘screen door’ effect. Since 1080p projectors appeared, this hasn’t been much of an issue in home theatre. But with 720p and earlier, sometimes the grid pattern of the pixels would be visible, with thin black lines around each pixel. (The name came from the similarity with looking through fly wire.) These days the pixel density is so high one really can’t see the inter-pixel boundaries unless unrealistically close to the screen.
Now consider what happens when you use a 0.47-inch DMD at 1080p instead of a 0.65-inch one. The pixels are smaller of course, but are the spaces between them smaller as well? Remember, a 1080p Digital Micromirror Device has on its surface more than two million tiny mirrors. They are slightly under 7.6m wide for a 0.65-inch DMD, and slightly under 5.4m wide for a 0.47 inch DMD. Are the gaps less wide? My guess is no, or almost no. Some clearance tolerance must be provided, and there’s little reason to think the tolerances could be reduced. If a powerful enough lens were used to display the image at 1080p to the same size as with the larger DMD, it would likely have a more noticable screen-door effect.
That problem at 1080p lends it to effective use with pixel shifting, since it reduces overlap between standard and shifted pixels. The pixels are smaller, which is just what you want.