Samsung Plasma PS51D6900 - not starting - smoked DC-DC converter BD9153MUV on logic board - Panel defect - address buffers kill logic board

This is a long story with no happy ending, but many interesting findings.

This pretty Samsung plasma machine caught my eyes on eBay classified. The reviews were very good at the time. It almost matched the image quality of the Panasonic GT30, which I have and love. I always wanted to take a look into a Samsung plasma, so I got it for a fair price. From the outside it was in mint condition. Not a scratch.

The device would click once plugged in and the stand-by LED blinked six times. Then silence. Just a slight buzz from a transformer in the SMPS.

I browsed the net for the six blinks, and apparently this doesn't mean anything. It's the normal startup blinking. As always, this Samsung had no proper self diagnose error codes. Suckers.

First, I had to extend my "lab" table to make room for the monster.

It is chock full of boards, no comparison to an LCD TV of the same size! Plasma screens require hardware.

No bloated capacitors. Standby voltage and all other voltages to the main board where all ok, the processors were warm, no findings on the regulators. PS-ON signal was low, which means "start". However, the power supply did not start any of the voltages for the scan/sustain boards.

I browsed the net and found this very useful instruction movie from Samsung, which goes through a number of power up sequence checks:

Power Up Sequence Diagnose

I stopped at the VS-ON check. This voltage was supposed to go high from the logic board. It didn't.

So, up next: logic board check. Fuse ok, but no voltages at the switching inductors. Upon closer inspection it was obvious why. The DC-DC converter was smoked. 

And a look through my 10x magnifier showed damaged traces, too. The chip is a BD9153MUV step-down converter. 

This board was toast. No point in soldering a new converter. Also, I had no clue why it burned up, maybe some chip had a short and killed it  - not likely though, those converters normally have protection circuitry. The chip was shorted on both switching output lines (S1, S2)

The good thing about Samsung TVs is that spare parts are available in numbers and cheap. A shop in Great Britain offered one for 25€ including shipping and a 30 day return policy. Deal!

Well, after I had mounted the new board, the TV seemingly booted. The logic board was blinking once per second, as explained in the video. Vs and Va voltages were ok. The converter chip was not getting hot and produced 3.3V and 1.18V. So that was looking good.

However, still no image.

I switched it on and off a few times and one time I lifted the set up from the table, to check if the menu would come up - dzzzzt, smoke, the chip burned up again! Argh! WTF? This time, only the S1 output line was shorted and no traces melted.

Time was too short to try the logic board self test screen patterns, which run without the mother board. This would tell me whether the panel, sustain, and logic boards work together properly.

So now what? What was killing this chip, why did it take a couple of minutes, and why wasn't it getting hot when I checked?

I removed the chip from the original board and checked for shorts on the output lines. None! All in the safe 10k+ Ohms range. This is really strange. Has the chip destroyed itself?

I ordered 10pcs via Aliexpress and replaced the DC-DC chip on the second board. That took hours and many failed attempts. Those tiny flat pack chips are a nightmare to solder, because the pins are flat underneath the package, unlike normal ICs where they stand out.

Just look at that tiny bugger:

Positioning it tested my patience. Hot air attached the chip on the pad, but the pins just wouldn't pick up the solder, most likely because I had too much solder on the pad and the pins did not reach down enough. I had to use the iron. Unless each pin has a shiny solder blob on it, the thing will not work. I think the art is to apply just enough solder on the traces and pad, and then reflow the damned thing until it sinks in. Or maybe: fix the chip on the pad first, then apply paste to the edges and then reflow. I think I'll try that next time.

Back in the TV the chip produced 1.2V  and 3.3V and the heartbeat LED morsed the OK code. Still no image.

Ok, logic board self test! As described in the video, first jump-start the power supply by shorting PWR-ON to ground:

Then, put a jumper on the upper two pins of the test jig:

After plugging in, the TV should start itself and display various test patterns.

Nothing on the screen. However, I noticed an alternating buzz from the scan boards. The logic board was producing something and the scan machinery did its job. Thus, the processor on the logic board was able to run a program. Could I conclude that it is 100% ok?

The following possibilities were left:
0) The processor is partially dead on its output circuitry.
1) Some shorted address buffer is pulling everything down.
2) None of the address buffers are working.
3) The software on the logic board doesn't match the panel.
4) The panel is generally FOOBAR.

Check 0)
How would I know? The processor runs its self test alright.

Check 4)
Can't do :-(

Check 1)
The address buffers have three connectors to the logic board. They are independent. When you disconnect one buffer board, the other half of the screen should still be ok. I checked one connector after another and got no picture.

Check 2)
I removed the cover from the address buffers to take a look.

There is a fuse for the Va voltage. It checked good and Va had its specified value.

I fired up the TV and checked if any of the driver chips in the flat cables, which connect the panel, get hot. Nope.

It is not likely that the address buffers are the problem. Cover back on.

Check 3)

There seem to be a number of versions for this board. While browsing eBay, I noticed a variety of labels on the U2002 SDRAM chip. It suspect this documents the software on the board, which is stored in the Flash memory chip U2012. The flash chip is too small to put labels on.

The label on the original board:

The label on the second board: 

It is also possible that there was another sticker on top of this one and it got lost. I think the first line is critical. I found the DS/DE label printed on the address boards, so that seems to match the 51'DS. Does FF really fit?

What now? Swap the FLASH ROM chip!

Done. Even louder buzzing from the scan boards. Still no image.

Swap the EEPROM U2007!

Done. No image.

Fuck this. Found another board on eBay with the exact same label and ordered it. If that's not going to help, we have case 4: FOOBAR panel.

So, the second board arrived quickly. Plugged in. No difference. I measured all voltages again and VA was not stable. It started with 58V and intermittently jumped up to 74V during the self-test. This can not be good! The problem search closed down on the address buffers.

Then I lifted the screen a little and found some dancing red pixels on the bottom edge. To get a better look I put the TV in a vertical position. Suddenly it switched off and the power supply would not restart (click, click, click). I replugged the power line - pffffft - DC DC chip blown again.

Cross check with the repaired board and the TV laid flat: same red pixels, which I did not notice last time.

This TV was susceptible to gravity! Once it stood vertically, it apparently kills the logic board via its address buffers. Crap! The only reason left was the panel. Major bummer!

Before I finally dismantled the whole device I ran a last check on the VA voltage. I unplugged the VA power supply connector to the address buffers, ran the self test and measured. At the power supply, the 58V were now stable. However, on the buffers, there was a voltage! It seemed that scan or sustain voltage was creeping into the address lines and from there into the buffers, the logic board, and the power supply. Some internal isolation in the panel must have been broken. That's why the power supply refused to start in vertical position. The voltage was far off and the overvoltage protection jumped in. The poor logic board had no defenses and died.

It think it is very unusual for a panel to fail completely. Normally at least some parts are still working and only a few address buffers or sectors are dead. Here though, the whole image was reduced to some red pixels noise on the bottom.

This was an expensive, yet educating experiment. I now own a large piece of junk. At least the scan, sustain and main boards are ok for sure and can be sold.