Faint lines on LCD display, any ideas?

[jeremy]

Hi all,

I'm really stuck with this repair, and I could use some advice.

I recently picked up a wacom cintiq which wouldn't power on the display. All of the LEDs and were lit, and the wacom touch surface worked, but the display wouldn't turn on.

I popped it open and the problem was that an 0603 decoupling cap had failed short (it was covered in cracks under the microscope), and that had caused the very small mosfet (SOT23) which switches on power to the entire display to literally explode. I replaced the cap and the mosfet after using my multimeter to measure the display current (500mA@12V, I used a mosfet which can take 4A at that voltage). The display came alive and all was well.

Looking at it more closely over the past few days, I'm getting strange vertical lines/bands down the display. It's hard to explain how it looks, so I have attached pictures. I can only describe it by saying that the green and red sections of the pixel seem to be unable to completely turn off; the problem is invisible with a white background but very clear on black.

The strangest thing is that if I leave it off for 24h and turn it on again, the good and bad areas of the display have randomly moved and are different in size! Sometimes it is mostly good with only a few stray lines, sometimes (like when I took these photos) it is mostly bad. On top of this, if I leave the display on for 30 minutes, it mostly goes away (sometimes completely). And if I turn the display on and off quickly or unplug the display cable, the lines still show up on the OSD (see pic) in the same place.

My guess was that this is a thermal problem somewhere, but I've tried the can of cold air on just about every IC I can find. My little thermal camera reports nothing hotter than 50C except a 1.8V linear reg which is running at something like 70-80C.

Some more info:
The display is made of two PCBs; a DVI/VGA to LVDS board, and then an LVDS receiver board on the LCD panel itself made by NEC (PCB says NEC 213 pw 303). The first board has an ST display controller and the switch mode power supply for the entire device. The second board has a Doestek dtc33ln08brh (google returns no hits, I assumed LVDS recevier) as well as a Sharp IR3E11M1 which seems like a high current voltage reference. Basically everything else is passives and small switch mode power supplies. I'm pretty sure the DVI board generates the OSD.

The panel is connected by hot bar soldered FFC cables on the LVDS recevier board. There are about 8 or 9 of these along the top of the screen and about 6-7 down the left side. I initially suspected there might be a bad joint which was warming up, but after having a gentle go with a 100C air gun on these, I couldn't see anything change on the display.

Things I have tried (all with the wacom touch panel disconnected):

-> Continuity testing the LVDS cable: all good
-> Changing the resolution of the display: lines remain in physically the same place
-> Use VGA instead of DVI: no difference
-> Change DVI cable: no difference
-> Use HDMI to DVI cable: no difference
-> disconnecting and reconnecting cables: no difference
-> cold air spray on the chips mentioned above and bar solder tabs: no discernable difference
-> 100C hot air gun on the above chips and bar solder tabs: no discernable difference
-> poking and prodding the tabs with my finger: no discernable difference

Any ideas here folks? I'm actually considering trying to build an LVDS recevier with an FPGA to see if it is the first PCB or the second which has the issue. And I'm guessing this isn't a power supply thing, but if anyone can think of a good hypothesis as to why it could be, I'm all ears.

[jeremy]

Edit:

After some googling, I think I've found the exact same display module: http://www.ebay.com/itm/NEC-SA-SFT-LCD-PANEL-NL160120BC27-14-21-3-/251389073853

It's the NL160120BC27-14

[Psi]

I wonder if, when the MOSFET failed and killed the power, if the datalines still had signal power and this damaged the inputs on the lcds col/row drivers.

In their damages state maybe the input esd diodes are randomly causing latchup.

Just a theory

[amyk]

Have you tried light tapping of various parts on the boards to see if it changes the lines?

A clear, focused photo that shows the subpixels clearly between the good and bad areas might help too.

[jeremy]

I have tried tapping, yes. I've also attached about the best photo I can get using my current camera.

Interesting, this morning when I turned it on, it was completely covered in lines. Perhaps it is getting worse. Unfortunately the panels go for about $1k, even nowadays, so it's going to be difficult to get a replacement.

[jeremy]

Some more research from today:

I found the following datasheet for an equivalent LVDS recevier: http://pdf.datasheetcatalog.com/datasheets2/22/224393_1.pdf

After looking at that latest photo closely, it seems like it is only every second row which has the problem. I'm wondering if this has anything to do with the even/odd pixel split on the LVDS interface? And now I'm even more confused, because this looks like a problem with the horizontal drivers rather than the vertical, but there are vertical sections of the display which still seem to be ok. It seems quite strange that every single odd horizontal driver has gone bad, so perhaps it is the connection to the LVDS receiver that is the culprit?

Thoroughly confused now 

[bktemp]

My guess: There is something wrong with the AC driving signal. It looks like this display uses dot inversion in a checker pattern: Every alternating pixel gets the inverted voltage to reduce flicker. And this gets inverted in every frame. It is an odd failure but my guess is, the POL signal (or whatever it is called) has a bad connection on some source driver ics so the pixels get a dc instead of a ac voltage.
Do the affected colums get worse over time when displaying a constant image?

[jeremy]

Actually, all of the columns get better over time as the display is on. But what you say would make sense. Is this normally implemented at the driver level? I'm assuming that the final drivers are COG type and are buried in the FFC tabs (but this is just a guess) because I don't see anything elsewhere that could be doing it.

Had another quick look this afternoon, now I'm getting some pink, yellow and cyan vertical lines when the OSD is showing (they are still there without the black OSD screen, but they are much harder to see; it's likely they could have been there all along). Not red, green or blue though. And yes, 15 minutes later the display is perfect and back to normal.

There is also a whitish border on the edges of the panel. Could be the CCFL backlight, but it really looks like it is in the pixels. This goes away after about 5 minutes, with the outermost pixels taking the longest.

High res pics of the lines: here and here
High res pics of the white border: here and here
They are too large to upload to the forum.

PS: no wonder these NEC displays cost so much; the colours are incredible (when it is working).

[bktemp]

The source driver ics are either mounted on the display glass or on the FFC. Here is a datasheet for an older source driver, but it shows the typical operating principle:
http://www.datasheetcatalog.com/datasheets_pdf/N/T/3/9/NT3980.shtml
On the new pictures it looks like several colums have bad contacts.
Try pressing or knocking around the screen to see if those lines are changing.
Maybe a picture on the screen with a higher contrast will show the bad colums better.