Originally Posted by qoopy
Greetings BattleAxe, thank you for your excellent posts.
1. It comes as a surprise to find out that down scaling is used by the 2716x1528 XPR chip to display 4k content. Was always under the impression that it worked under the same principle as in the .47 XPR, but using a 2-way shifter instead. Any reason given why point sampling is not used?
2. Is there a reason why there's no paring of Optotune’s XPR
shifter with the larger native 2k .66 DMD?
Many thanks and best regards,
1. You can't use point sampling during a non-one-to-one input-to-output map, mathematically it just doesn't work.
Going from 3840 x 2160 to two shifted 2716 x 1528 frames, implies that each shifted image is sampled at (0,0) and (0.5, 0.5) pixels offset, respectively. Then there is the non-integer scale factor of 1 / ~1.4, both of which mean that if you point sampled the 4K image you wouldn't get the right colours at the right places, and it would introduce all kinds of artifacts. Of course, you can always do point sampling and take four samples instead of one, then manual bilinear filtering (4 input samples to 1 output sample) during the downscale, but that amounts to the same thing as bilinear sampling in hardware. Who knows how these chips work, but downscaling is done in hardware mostly these days.
2. There are native 4K DLP projectors using XPR now to do 4-way 8K. The one from Barco can accept 4K 240hz too, which is wicked for gaming. I don't see the point why they'd ever release a native 2K one when cinema DLPs use the 4K chips (without XPR since there is no 8K DCI content yet AFAIK
). Shifting once diagonally from a 2K DLP instead of a 2.5K DLP would only get you halfway to 4K resolution, or 4M pixels instead of 8M. Shifting twice would require 240hz operation which the older chips cannot do, at least not in 1-chip mode. Although the DCI variants are all 3-chip designs so they can do 120hz no problem, maybe even 240hz since the 4K native DMDs do 240hz now.
The most interesting thing I discovered about all this, tech-wise, is how XPR differs from e-shift. e-shift is superior for 2-way shifting, actually cannot function in 4-way shifting, because of how it uses a single value, the circular polarization angle, to control the amount of shifting that a birefringent crystal does. That's why no 4-way e-shift, because stacking two would make alignment a pain. It's also true in the other way, e-shift is better in the sense that it has no mechanical moving parts to it, it's just an LCD and a crystal wedge, but LCD means the input light has to be pre-polarized otherwise you'd lose half your lumens. So that's the real reason why DLP uses XPR instead of e-shift, at least for the 2x shifter variants.