• TI Thinks Resolved

DLPC410: Projected image periodically shifts at subpixel level in diagonal direction

Prodigy 40 points

Replies: 6

Views: 69

Part Number: DLPC410

Hello TI,

I'm currently trying to use a light engine fitted with 0.96" WUXGA DLP9600 & DLPC410 for 3D printing application.

For starter test I printed a straight hollow column, therefore the light engine was supposed to project the same cross-sectional image at exact same location for each layer.

The light engine was securely locked in place but here's the printed part:

   

The projection shifted in diagonal direction by a small amount every 40ish minutes then back to normal in about 5 minutes and repeated itself.

I checked the image files on PC screen but no abnormality found. I examined the printed part again and it seemed that the offset was smaller than a pixel.

In order to determine the amount of offset I performed another test. Using a Mylar film as projection screen with marked masking tape on top as reference, I was able to record 600 of consecutive projections under a macro camera which was mounted on a rock steady tripod. Let's take projection 227-228 as an example:

The thick bar in the middle was 6 pixels in width. There were two bars and I deliberately put the left one just outside the edge of the frame by fine tuning camera position.

Everything looked normal till projection no.227 then the image suddenly shifted to the right in projection no.228, with part of the left bar entering the view. 

With the help of Guideline function on Photoshop, the amount of offset was measured 1/6 pixel:

 

The image shift happened in both X and Y simultaneously (i.e. diagonal) but I only managed to get it down on tape in one direction due to the lack of professional monitoring equipment.

I repeated the test quite a few times and the offset amount varied from 1/7 to 1/4 pixel, instead of a constant number.

The image shift was just too regular to be input signal or mechanical structure related. 

So, what are the reasons of such a behaviour and how this problem can be fixed?

Thank you in advance.

Daniel

  • Daniel

    Welcome to E2E

    Thank you for the detailed description of the issue you are seeing.

    This is a question to ask your DLP light engine provider / 3D rendering software vendor.

    I would investigate the bit plane files before and after the event to see if you can identify the shift in the data.

    There is nothing in a DLP DMD that can cause a shift like you are seeing.

    Carey Ritchey

    Texas Instruments

    Industrial Business Development Manager

    DLP® Enterprise & Cinema Display

     www.ti.com/dlp

  • In reply to Carey Ritchey:

    Hi Carey,

    Thanks for your reply.

    I've already contacted the vendor but I have to say it is a remarkable waste of time so I figured I'd try my luck here.

    My first guess is that this image shift has something to do with unstable power supply or static causing insufficient rotate angle of the micromirrors.

    Anyway, thank you for your advice again.

    Daniel

  • In reply to user6215894:

    Daniel

    Sorry to hear you did not have a good response from your supplier.

    The DMD micromirrors are fully bi-stable as in they only have the two tilt angles (+12 deg, -12 Deg)

    The issue certainly appears to be in the pattern data being loaded into the DMD.

    You image showing the shift at 227, 228 looking at the pixilation of the middle 2 in 228 the shift appears to be full pixel shift.

    This can occur in the pattern generation software as it maps the data to the bit plane

    You can try a test where you load the same single pattern for a full build

    Carey Ritchey

    Texas Instruments

    Industrial Business Development Manager

    DLP® Enterprise & Cinema Display

     www.ti.com/dlp

  • In reply to Carey Ritchey:

    Carey

    The shifts as seen here were actually less than a pixel.

    It would be much easier to see the pixelation with a High Pass filter added in Photoshop:

    So I ran another test today following your suggestion. The build job was a sequence of identical pattern and here is a comparison of projected image before and after a shift:

    Vertical red line in the middle represented the edge of masking tape used as reference in displacement measurement. And the vertical red line on the right indicated that both camera and projection screen were absolutely still because the markers lined up perfectly.

    Each of the light bars was 2 pixels in width so relatively the projection shift was about 1/4 pixel to the right, which fitted well with my previous test results.

    Daniel

  • In reply to user6215894:

    Daniel the last thing we can think of is a thermal shift in the  light engine or printer mechanical mount ?

    Is the exposure time and time between each layer build exactly the same ?

    Carey Ritchey

    Texas Instruments

    Industrial Business Development Manager

    DLP® Enterprise & Cinema Display

     www.ti.com/dlp

  • In reply to Carey Ritchey:

    Carey

    Yes, all settings including exposure time, waiting time and Z-axis movements were fixed and both recorded video and printer log confirmed that.

    In fact, thermal shift was the first thing came to my mind, yet I failed to figure out a feasible way to measure any physical displacements that small.

    The throw distance on my set up was about 10 inches, meaning that mechanical displacement / rotation of the light engine (or any components inside) should be sub-micron level to result in a 10 micron shift on projection screen.

    Our room temperature was maintained at 25°C and temperature sensor placed right next to the light engine read 28°C constant as shown on printer logs.

    I believe any thermal shifts occur under this circumstance would most likely be random, yet the 40ish minutes regular time interval doesn't seem to be related to anything.

    Here in my workshop I also have a German made DLP printer with seemingly weaker and simpler mount but no projection shifts detected so far.

    Could it be thermal shift caused by insecurely mounted DMD or DMD heatsink or TIR prism?

    Another weird thing I noticed today was that the shift seemed to be smaller with air conditioner and any other high power consumption equipment turned off when I was trying to rule out the effects of unstable power supply.

    Daniel