[David Mendicino]

I am curious as to why projector manufactures have not gone to the extreme to eliminate the rainbow artifact? If they can do 6x, why not speed it up a little more? Is there a limit to the DMD panel that does not allow the mirrors to switch fast enough? The new wheels, which claim to be 5x, are really 6 segment at twice the speed. Why not 12 segment at 3 times the speed? or 24 segments?...or better yet, make the colour wheel bigger, so that the angular velocity of the segments are faster.


That archimides colour wheel seems to be promising. But now that I think about it...would that not cause rainbow in the vertical direction, instead of horizontal? That is, currently, when I see rainbows, it is on my lateral periphery...with the archimedes wheel, we would see them above and below the screen.

 

[JoeFloyd]

The mirrors themselves would not have any trouble, but each time a color wheel segment passes into the light path, the color data must be loaded into the custom ASIC that drives the actual DMD. Loading 786KB of data per color transition X 3 is 2.35 MB per spin of the color wheel. With Xx that would be 4.71 MB per spin. Spinning at 60 RPM that turn into 141 MB/s for a standard color wheel or 282 MB/s for a 2X. These data rates are well within the limits of current technology, but pushing those data rates higher results in higher costs.


There is also an optical efficiency consideration. As the color wheel rotates into position, the DMD's are switched into the off position. This is so the DMD doesn't modulate the wrong color light into the image. As more color transitions are introduced, the DMD will spend a higher total fraction of time in the off position waiting for the color wheel to be in the correct position. This is one of the reasons that the Archimedes' spiral color wheel has been developed. With this approach there is no lost light from switching the DMD off during a color transition.

 

[David Mendicino]

As for cost, I have a $200 video card that put has a memory transfer rate on the order of gigabytes, as are most modern motherboards within this price range. I guess,what I am saying then, the 200-500 cost that would be required for such processing is trivial when dealing with $7000 projectors? Or, is there something inherently different about projectors that such present day, (relatively) inexpensive technology cannot be applied to it.

 

[PhilB]

David,


Another way to get higher data rates is to make a wider bus. That's probably what your video card is doing. Without going to double-data rate drivers, typical data rates will be in the 50-75 MHz range (100-150 MHz clock rate, single-data rate SDRAM). At that rate you could get 900 MB/s (megabyte, not megabit) by using a 96 bit wide bus. That size bus is not very practicle on a device package scale.


More likely the DMD has a 32 bit wide bus, which could get up to 300 MB/s.


Motherboards also take advantage of wider busses. DIMM modules use multiple 16 or 32 -bit wide parts to make up a (i believe) 64 bit bus.


Joe,


What do you mean about the SCR mirror-off benifits. I thought the benefit of the SCR wheel was that light is no longer absorbed by the wheel but instead bounced back into the light tunnel to be recycled.


-phil

 

[JoeFloyd]

Part of the magic of the DMD device is the color modulation scheme that is based on pulse width modulation. What that means in laymen's terms is that the brighter a given color the longer the mirror will be in the ON position within a given period of time. The period in question is the time in which a given color wheel segment is in the proper position in front of the DMD. (Well the actual optical path is more complex that this, but this is the essence of the situation) Since the wheel spins pretty quick the time that a given color is in front of the DMD is a MAX or 1/4 X 1/60 for a three color color wheel with clear section and spinning at 60 RPM. (The clear segment may not be as large as the other segments. I don't really know what ratio of color to clear appears on a given color wheel) At any rate, the maximum time a given color will illuminate the DMD is .004 or 4 thousandth of a second. In reality, that time is a little less since the color sections are separated by color transition regions that must be accounted for by switching the mirrors into the OFF position until the wheel spins past the color transition region. If this wasn't done the DMD would reflect two different colors for a very brief period of time and that would cause the color and image brightness to be a little wierd.



Archimedes' spiral color wheel takes advantage of the fact that the rate at which colors scroll across the surface of the DMD is a constant. It's one of the properties of the Archimedes spiral and is important to keeping the internal timing consistant within the DMD driver circuitry.


The important difference that comes into play with Archimedes' spiral color wheel is that all three colors illuminate the DMD simultaneously. The DMD driver has been modified to understand how the spiral pattern will map onto the surface of the DMD as the wheel rotates. The basic function DMD is the same, and the color modulation scheme is the same, but the DMD driver logic is much more complex to track the scrolling color bands.


One problem with the spiral color wheel is that there is color contamination in the areas where is mirror is illuminated by two colors. This does affect the resulting image. The only way to address this is to increase the complexity of the DMD driver to account for the color contamination and change how the color modulation system drives the mirror that has two color illumination.

 

[steve5097]

If I remember right, the switching time of a DMD mirror is 10 microsecs. Assume 256 levels of grey, and you arrive at a minimum time required for a segment to be on. I did the math on this a few months ago, but I don't remember the answers and I don't have time to redo it. The bottom line is that there is a maximum refresh rate beyond which the gamma curves are screwed up, then grey level is lost.

 

[David Mendicino]

So what might the theoretical limit on colour wheel speed be, in terms of colours per second? I say this, because the 5x is technically a 6 segment colour wheel (2x) at 2.5 times the speed.