Trying to help the neighbors with their TV that acted up...

Big LCD TV with LED backlight. In the electronics everything is on one PCB - All the power supply, LED driver, and display processing chips. There is nothing obviously wrong with the PCB

The backlight is supplied with 2 cables, each with 2 conductors (Red and Black). Each is responsible for lighting half of the screen. Surprisingly its not a strip along the top and strip along the bottom as i'd expect, but left half and right half of the screen

The problem : The backlight flashes for split second and goes off. There is voltage in the backlight related circuitry all the time, looks like some Enable signal is flashing once and then disappearing, switching the backlight off



I put the multimeter to maximum detection & hold (there is such feature in it) and measured the voltage each array gets during the bried flash

76V
76V



I made an isolated (not Earth referenced) high resistance DC supply :
UPS (230V squarewave output) ----> rectifier & big capacitor ----> 340Kohm resistor ---->

The thing put out ~370VDC

Powered each of the LED arrays with it

I ~1mA
Each of the LED arrays (tested seoarately) glows dimly. Measured Vf 75V in each.

Changed the resistor to ~8K ohm

I ~40mA
Each of the LED arrays (tested seoarately) lights up well. Measured Vf 83V in one / 100V in other



Looks like they are not the same, so one is failing ? So the original circuit is responding to that ?

But how it detects the problem if the Vdrop is 76V in both when it drives the arrays ?

Ash do you have the make model and chassis number of the TV, and I'll see if I can dig up a schematic for it, because with out the schematic or trouble shooting guides they are basically impossible to fix unless you fix TV's all day long and know what to look for.. 

Chinese no name (sold locally as Fujicom). I have not looked further although maybe i would find something with the part number of the PCB or variations of the TV model in general

By the V/I characteristic of the LED array (as measured) it clearly is something like ~25 LEDs in series x unknown in parallel, and by the brightness at 40mA, its proper working current is probably about 50..60 mA. There are no obviously dark spots anywhere so no part of the array appears to be out

Check the transformer/s leads, or any inductor used on the PCB they have had problems with the soldering and sometimes they break because of a cold solder joint, if that is the case rework the solder joints (I like to use 63/37 even if it's not lead free, it lasts way longer) and it should be good to go, I would also glue the transformer to the PCB once it's all working again to reduce any risk of it moving but that is optional..  Also check the wire wrapping on the post from the coil to the post going into the PCB while your inspecting it, they have also been known to have issues..

All transformers are connected well. The backlight secondary is up to voltage and stays at stable DC voltage. Its the actual LED switching transistors that are not getting drive, as if something in the circuit is not getting enable signal

How do the CAP's look on the power supply and are they the 85 deg ones ?

Mixed bag in there, but its not a bad elcos issue

What is suspicious: One string 80V, other 100V. That suggest that one has failing LEDs inside, so the ballast is likely shutting it down (load overvoltage)...
Normally I wouldn't expect such difference among two identically designed string assemblies...

There is the interesting part :

Powered at 1mA - A=75V B=75V

Powered by its original driver, at what looks like full power - A=76V B=76V

Powered at 40mA, at what looks like below full power - A=83V B=100V



Why at the high power setting, on the original driver they are the same and with DC+resistor drive they are not ?

Does the "max hold" feature of a Fluke 115 DMM (which i use to get a reading in the brief power on) catch the absolute peak voltage (i.e. LED true forward voltage in the times when the PWM is ON) or some average voltage (the PWM average), which might be allready regulated by the controller ?

What sort of fault would lead to increase in Vf of a string of LEDs of just few Volts ?

The "MaxHold" takes the maximum reading the meter acquired as a result of the measurement conversion procedure, nothing else. It is not any peak detector or so, so you should not confuse it.
The question is, how the measurement conversion work, so how it respond to fast changing voltage.
With a DC meter (set to a DC voltage measurement) it usually means applying the normal voltage integration (either dual slope, or continual integration, depends on the exact system used), so in fact it means an average voltage there.
So the "MaxHold" may represent a peak value, but only if the voltage is varying very slow, with maximum frequency component significantly slower than both the integration time and the readout refresh rate (be aware, many meter structures have physical integration time multiple times the display refresh period).

For real peak value measurement you really need to use an oscilloscope.

What the ballast does is a question: It may sense an overvoltage and temporarily shut down. Then after some time restart and again go to the overvoltage (and repeat that for the second or so, when some suprvising diagnostic decides to shut it down completely). This may lead to quite large AC component, but shift the average voltage reading by the DC meter down.

What may be the fault with the LED array that makes the Vf go up ?

Loose bondwire and so arcing there, or some bad metalization on the LED chips (so the current has to flow through higher resistive semiconductor material). Note, this wont affect the emited light: The active junctions get all the same current.
This would even conform to the user observation of flicker and flashing - these effects are typical for failing bond wires on LEDs before they really fail for good.
But there may be more reasons, it even depends on the technology, how the individual LEDs are connected together (soldering vs glue, copper traces PCB or some silk printed conductive paint,...)

In any case why would be there any difference, if both modules were OK? So that really means the culprit is one of these illumination modules (it could be even some breakdown problem, so leakages, on the "lower voltage" one) and so it needs to be replaced (or if the fault nature allows it, maybe even fixed).

I mean this :

The higher Vdrop on a specific LED in an array would lead to accelerated destruction of that LED so it ends open circuit. The original backlight circuit may shutdown fast enough to save it from burning open at once, but wouldnt ~20 sec operation be enough ?

(i could not do it for longer because in the 40mA experiment i used 7 1Kohm 1/2W resistors in series, i pushed them over 3x their power rating)

Alternatively, if the failing array is the one with low voltage, then would it be ok to just keep supplying it with the normal current ? If any LEDs are shorted they dont dissipate much power anymore, so they would be quite "stable"..

Accelerated destruction yes, but the question is, what timeframe are we talking about. The prime destruction mechanism is heat, but the extra 1W  on only one LED may not heat it hot enough to really fail within just few hours, so it may operate like that easily few 10's or even 100's hours. Dont forget if the LEDs are rated 20khours at Tj of 80egC, on free air that may be just 60degC, so the expected life may increase to nearly 100khous. And when the fault shortens the functionality to 1/100th, we are still talking about 1000hours of operation...
The fact it is not reliable with such local overload (i.e. the reason why they didn't load the LEDs in the first place) does not mean it wil always fail soon, it just mean the percentage of the units becoming faulti is too high...

So there is no way to let it stay stable with the failing LEDs left inside. I'll then have to really open the panel and take the LED array out, manually short the dead LEDs, (add a diode string in series to bring the overall voltage back up so the controller sees correct voltage)