This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Increasing frame rate of DLP using LED dimming

Can someone explain (with example number calculations) the concept of modulating an LED light source to speed up the frame rate of the DLP.

Specifically using the .7 XGA chip and its inherent limitations today. Any help is appreciated. If there are TI docs on this already or papers in the community

that would help also.

  • Hi Christopher,

    I hope all is well today.  Thank you for your participation on the E2E site.  Below is some information and answers regarding your inquiry above.  I hope this helps out. 

    When projecting simple binary images (typical of Discovery setups) there is no way to increase frame rate by modulating the light source.  The max frame rate for binary images is dictated by how fast the hardware can load the DMD array.  For producing grayscale images with a DMD, a series of time weighted binary images is displayed in rapid succession.  The minimum realizable duration of DMD mirror switching is what dictates the max possible frame rate, as follows…

     Assume that the fastest possible off-on-off mirror transition interval is 20 usec (approximately correct).  To create an 8-bit grayscale image, you need to project at minimum a series of 8 binary images with the following durations:

    LSB=20usec, LSB+1=40usec, LSB+2=80usec, etc... etc… LSB+7=MSB=2560usec

     Add them all up and you get a frame time of 5100usec (which is not coincidentally equal to 255*20usec).

    To render a complete 8-bit grayscale image any faster, you need a shorter LSB time.  You can create an “effectively” shorter LSB time by dimming the light source during the LSB mirror interval.  Cut the source intensity in half during the LSB and you have an “effective” LSB of 10usec.  Alternately, the LED can be strobed with a pulse duration of just 10usec.  In either case, the DMD mirror must still remain in the on-state for 20usec, because it cannot switch any faster, but the integrated brightness during the LSB interval will be cut in half.  Now the realizable frame time is something like 255*10=2550usec. 
    Actually, the realizable frame rate is somewhat lower than would be suggested by the preceding simplified calculation because of DMD loading constraints, but this is approximately correct and illustrates the major limiting factor.

     Please let me know if you need clarification on any of this or if you have any other questions:

    Justin Lemon
    Digital Light Innovations
    512-615-4630
    jlemon@dlinnovations.com

     

  • Can you tell me what the limitations of the LightCommander are.

    What is the highest frequency/shortest duration it can modulate/pulse the LED?

  • Hi Justin and others,

    May you discuss more detail about the scheme above? You can cut the LED time by half but you still need to wait for the mirrors to flip. So that the time is still 5100usec for one frame.

    It seems I do miss something. If I want to play a grey (8 bits) video at 120 fps, I need the DMD operating at frequency of 120*255 = 30600 Hz. This is higher than the current DMD with HD resolution. But I see this performance in many places. Information below is even super high with a single chip. Can you please let me know how it is possible? Do the cinema-DLP manufactures use special chips?

    Thank you very much,

    Example on super high performance:
    This one is super high frame rate for RGB video: www.barco.com/.../F50.pdf
    120 fps * 3 colors * 255 = 91800 Hz.