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.

Sync modern multiple DPL projectors for stereo viewing?

In 2022 tech, how do I sync multiple DLP projectors (at least 3 or 4) to display a continuous, circular (270 to 360 degrees) 3-D stereo image ???

We would prefer to use RF shutter glasses if possible for multiple players/operators.

I am in the educational gaming field, specifically the Entertainment Technology Center, at Carnegie Mellon University and we generally use Unity to play & render animations.

We are building a 20 to 30 foot circular wall work-room for the students with central ceiling projectors.

The users will operate inside the room with various hardware.

The corrective projection/surface mapping for the curved surface is no problem, but we can not get multiple projectors to stay consistently synced, for stereo 3-D viewing, by any of several means that we have tried thus far.

Thanks in advance for any guidance that you can provide.

 - D. Purta

Senior Engineer

  • Hello David,

    It would be good to know what you have tried so far to synchronize the projectors that did not work.  From there our video folks may be able to provide suggestions from their experience regarding synchronization.  You have rightly perceived that it is not a trivial exercise.

    If you are not comfortable sharing this in a public forum, please send me a friend invite and we can discuss this privately.

    Fizix

  • I will compile a list. and get back to you ASAP.  THANKS !!!

     - David

  • My name is Steve and I'm working on this with D. Purta (the OP).  I can add some details that should help with understanding what we've tried.

    We're using DLP projectors that are 3D stereo capable and use DLP Link for glasses synchronization (Optoma GT5600).

    The 3D stereo feature of this model projector works great........one projector at a time.

    We have a single PC with a single Quadro GPU, and from that GPU a single output to a Matrox videowall appliance (QuadHead2Go).  From there out to each of these projectors.  We are sending a typical OpenGL Quad Buffered stereo image from Unity.  This set-up is our best attempt at "passive sync" in that there shouldn't be anything that needs to be sync'ed.  If we were to use multiple GPUs, or multiple PC's, then my understanding is that we would need hardware sync cards and projectors with dedicated sync inputs.

    In my first test I have two of these projectors placed side by side and edge blended to make a ~32:9 display.

    Here is the problem:

    When viewing stereoscopic content on this double-projector display, only one side will display correctly at a time.  If you look at the left side, the glasses will pick up the left projectors DLP Link signal and sync to it, but in your periphery you can see the right side projector not in 3d anymore.

    If you turn your head to favor the right side, the glasses will then suddenly sync to it (within < 0.25 sec) and now you can view that side in stereo 3d.  But, then in your periphery you can now see that the left side is no longer in 3d.

    What this tell us is that the two projectors are not in sync, but, we have no reason to believe the two signals coming out of the Matrox videowall appliance are not in sync.

    The best we can conclude is this:

    The two display signals are in sync, but extra information that the projectors *add* to the signal, for DLP Link, is running on its own internal schedule.  And we don't know how to make that sync with the input video signal.

    We didn't get this idea until we notice that what I just described was not the only artifact.  Sometimes both sides are in 3d stereo, but one side is discolored.  Usually purple or yellow.  Some times both are fine but one exhibits color banding.  The effects seem to cycle, somewhere on the order of minutes.

  • Steve,

    Thank you for the detailed explanation.  The video team is talking about this.

    Let me see if they have any suggestions on this.  This is a very interesting idea, but may not be trivial to solve.

    I will see if we can get something back by Tuesday evening.

    Fizix

  • Steve,

    Thanks for the additional information.  Unfortunately, DLP Link was never designed to be used in a multi-projector scenario, so there are bound to be issues in trying to use it in that manner.

    To answer some of your questions:

    The DMD is driven by what we call a sequence.  The sequence is a set of load and reset instructions that tell the HW when to load the DMD's memory and when to reset the data onto the mirrors.  It also contains signals for external devices, such as laser/LED enables, the DLP Link pulse, etc, so that these devices can sync with the data on the DMD.  The sequence is phase and frequency locked to the colorwheel (CW) on a lamp/laser-phosphor system, and to Vsync on a system with no wheel (LED or direct laser).  The CW is phase and frequency locked to Vsync.  The sequence is HW controlled with tens of nanosecond resolution.  There is some slight variation in the sequence timing as it locks to the CW/Vsync, but this should be in the <10us range or so, depending on the wheel/Vsync stability.

    For 3D modes using active glasses, dark time in the order of 1.5ms-2ms is placed in the sequence either at the beginning or the end of the 120Hz frame.  This dark time gives the glasses time to detect the DLP Link pulse placed in the dark time, calculate which eye is being displayed, and switch the shutters to the appropriate orientation.  As the glasses are triggered by the sequence, they are always sync with the data on the DMD.  There could be some possibility that the L/R are swapped, but they should never be getting switched in the middle of a frame. 

    For what you are attempting, every projector would have to be perfectly synced in both phase and frequency.   The Matrox website says the QuadHead2Go output is synced, but I couldn't find any spec or material that explains exactly what this means (tolerances, etc).  I have no reason to doubt that they are synced, so assume something else must be going on.

    Your description "Sometimes both sides are in 3d stereo, but one side is discolored.  Usually purple or yellow.  Some times both are fine but one exhibits color banding.  The effects seem to cycle, somewhere on the order of minutes." would seem to indicate that not only are the projectors not phase synced, but that they are running at slightly different frequencies.  If one was running at 119.9Hz and the other at 119,95Hz, the Vsyncs of one would slowly drift through the frame when compared to the other.  Your description would seem to indicate this is the case as it takes minutes to cycle through the artifacts.

    To see what is actually happening, it would be good to get a couple of high speed photodetectors and a scope to display their output.  Put a photodetector in front of each screen.  You should see 6ms or so of light and then 2 ms or dark time for each frame.  The dark time of each projector should line up exactly.  If they are not, but are stationary relative to one another, they are not phase synced (obviously).  If one is drifting when compared to the other, they are not frequency synced.

    One test you could try to see if this would ever be feasible is to send the exact same data to side-by-side projectors through an HDMI splitter and check the photodetector signal per the above.  This is the best case scenario where they should absolutely be synced.

    Check that both projectors are in the exact same mode.  Not sure of what modes your projectors have, but you might check that they are using the same settings (display mode, color temperature, etc).

    Another issue in all of this is that our customers implement their own designs.  It could be that this system has a frontend that is somehow altering the incoming Vsync signal in some manner (frame rate conversion, etc).  The above side-by-side test should show whether this is happening or not.

    Obviously, the optimal solution would be to use passive 3D (polarization, wavelength, etc).  For passive 3D systems, the projectors would no longer need to be synced at all as the glasses would automatically filter out the data as necessary no matter when it occurs during the frame.  However, these would probably be very expensive, and I don't know of any systems using this method except cinema systems.

  • Duck thank you for that extremely detailed and informative reply.  This helps illuminate things considerably.

    Yes, the test you suggested, with the same output to both projectors via splitter, was going to be our next test.

    And I agree that this would be easier if we used passive + polarized, but, that has it's own problems from our end.  We still may end up there if we can determine that polarization-preserving screens will work with both ultra short-throw (way off-axis) and curved screens.  We're worried it will create bad hotspotting.

    D. Purta can look into a photodetector for the other test you suggested.

    Based on what you also said about the phase and sync locking, LCD projectors are looking like a better option overall.  It's too bad as these Optomas we have are great otherwise.

  • Steve,

    No problem.  Let me know if you come up with any more questions, or want to discuss the results of your tests.

  • Steve,

    I am going to mark this as "TI Thinks Resolved".  It will remain open to adding follow-up questions for a couple of weeks, even if you click on "Resolved".

    Fizix

  • That's fine. Thanks.

    We won't have a chance to conduct another test with this set-up until early next week.

  • Steve,

    Even if you miss the window, you can start another thread and point back to this thread.

    Fizix