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.

  • Resolved

DLP5500: Seeing strange vertical lines

Intellectual 2250 points

Replies: 10

Views: 190

Part Number: DLP5500

I have a customer using the DLP5500, DLPC200, DLPA200 on a PCB for 3D image detection.

This product was launched already, but they are seeing returns due to strange patterns in the displayed image.

They have two of these products on hand:
- One is a return showing the strange pattern as shown in the attached documentation.
- The other is a working version showing no issues

They tried replacing the DLP5500 from the good product to the bad product, but they saw the same "strange patterns"
Thus, they think there is a problem with the PCB/setup, and not a device malfunction.
They believe there might be a problem with the solder joints causing the LVDS transmission lines impedance to go out of spec.

Is there any idea what might cause these "vertical lines" as shown in the attached .ppt?
There was another post that discussed impedance and maybe wiring issues causing "vertical lines" issue: https://e2e.ti.com/support/dlp/f/94/t/585587

Any help is hugely appreciated.

Regards,

Darren

DMD Lines.pptx

  • Hello Darren,

    This particular pattern very strongly suggests a poor ohmic contact on some of the data lines or possibly on the ground.  It looks like something is occasionally not getting the 1 and shows up as a light gray band.  If it were narrow vertical black lines that would almost certianly be an LVDS line that is not making contact.

    In any case there is clearly something that is making poor or capacitively coupled.

    We assume this is their own design.  Care must be taken in design of the interconnect cable if they use one and also check the socket the DMD is insterted into.  If it is a LIF (low insertion force) socket, they are not very reusable and more susceptible to damage on insertion.  If they are a ZIF (zero insertion force) socket, they can still be damaged, but less likely.  Extract the DMD and inspect the socket for bent or missing clips in the socket.

    FIzix

  • In reply to Fizix:

    Hi Fixiz,

    Response from the customer:

    We checked the LIF socket the DMD was inserted into, but didn't detect any damage (bent/broken clips) upon visual inspection.
    We could use X-Ray to examine the solder joints, but we don't have the capacity to check every single data line for the DMD given our equipment.
    So, based on the image(attached in previous post), are there any data lines that have a higher chance of being the culprit, that we could investigate first?

     

    Thanks,

    Darren

  • In reply to Darren (FAE):

    Darren,

    The fact that these are wide bands instead of a single completely dark line 1 pixel wide is why I suggested looking at the DMD board main ground.  This looks like a possible ground bounce.  If it was simply an LVDS positive or negative line it would show up as thin stripes.  These bands look to possibly be associated with one bus (A or B).

    Check the clock lines for sure to make sure one of them does not have a poor connection.  Can you tell me how wide the bands are in terms of pixels?

    Fizix

  • In reply to Fizix:

    Hi Fizix,

    They did the pixel calculations based on the image projected to a surface.
    Since the projected image is 1024 pixels horizontally, and the width of each "problematic vertical stripe" is about 1~1.5% of the entire projected image, the pixel width of the problematic stripes are calculated to be around 10~15pixels wide.

    If there is any other way to identify the pixel width besides this ☝, advice would be appreciated.

    Darren

  • In reply to Darren (FAE):

    Hello Darren,

    I would venture to guess that they are probably 16 pixels apart.  This is the width of the data input bus.

    One of the simplest ways is to display a vertical line one pixel wide that repeats every 16 pixels.  If the banding matches the line (in stays the same distance from the vertical line across the 1024 pixels of each row, then the separation is 16 pixels. 

    If it appears to occur at a different rate than the vertical one pixel line, then it is not 16 pixels.  Then make a pattern 13, 14, 15, 17, 18, & 20 until it matches the banding.  

    If the width of each band is NOT 16 pixels, then that would tend to rule out an LVDS issue.

    Fizix

  • In reply to Fizix:

    Hi Fizix,

    Due to some difficulties in their ability to modify the settings of the displayed image, they used the method shown in the attached document.

    It looks like the pattern repeats itself every 32px

    Hmmm...*scratches head*

    pattern_measure.pptx

  • In reply to Darren (FAE):

    Hi Fizix, I forgot to attach the document...O_o...

    So I added it to the above post. Sorry about that.

    -Darren

  • In reply to Darren (FAE):

    Hello again Darren,

    Thank you for the image.  Almost certainly related to the data bus.  I am not certain how to help further at this point.

    This will take more investigation on your side.  I will think about it over the holiday weekend to see if I can come up with a way for you to test.

    Fizix

  • In reply to Fizix:

    Hi Fizix,

    I marked the above as resolved for now, but have you had any thoughts regarding a possible efficient way to test the databus...? Or perhaps which lines might be a bigger culprit...?

    Thanks,

    Darren

  • In reply to Darren (FAE):

    Darren,

    Because the bands are a bus wide, I would look at signals that affect the whole bus.  Things like the associated clock bus, or SCTRL lines.

    Fizix

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.