Reply With Quoteloopyloo (16-09-17)
Tap testing (rubber mallet first then an eraser tipped pencil for finer resolution) should narrow it down if it's a bad connection.
But ultimately need to laboriously trace back from the leds to find where it's shutting down.
loopyloo (16-09-17)
Just pulled the back off to look at it.
This is the controller. There's a hell of a lot in it for a little chip.
Last edited by loopyloo; 19-09-17 at 05:19 PM.
From the data sheet:
The PFC circuit has been disabled/bypassed so that HVDC voltage is a bit lower than normal, like 10% or more so need to look into the data to see if there's under voltage shut-down that's being triggered and whether it can be de-sensitized to suit the lower rail voltage. The other approach would be to get the PFC circuit up and running but that's the scene of the original disaster.The LX27901 is an extremely versatile high-performance controller for driving LEDs from the HVDC (output) rail of a standard PFC front-end.
Need to determine if the backlight supply is actually directly supplied from the HV rail too as that would create an isolation issue with respect to the mains input unless there's a transformer between the inverter and the leds.
Check what's going on with the ENABLE (pin2) and FAULT (pin7) inputs when the LEDs shut down would be a start
Last edited by Skepticist; 19-09-17 at 06:41 PM.
loopyloo (19-09-17)
Most likely source of trouble appears to be an open or shorted led as the chip detects these conditions and latches the FAULT output low, putting the chip into a sleep state which is only reset by cycling the power or the ENABLE input.
Capacitors are indicated across the LED strings so a highly leaky cap there could produce an overcurrent problem.
A low value high power resistor between the source of the switching FET and ground is used to detect overcurrent so check that for signs of overheating and check that its value hasn't gone high (would produce false sensing).
Is the screen edge or back lit? (not that it makes much difference but you'll need a large room to fully dismantle this TV to get at the LEDs for individual testing
Yes it's because power cycling restarts the backlights that I'm worried the led's might be the problem. They are fed by transformer by the way.Originally Posted by Skepticist
Most likely source of trouble appears to be an open or shorted led as the chip detects these conditions and latches the FAULT output low, putting the chip into a sleep state which is only reset by cycling the power or the ENABLE input.
Capacitors are indicated across the LED strings so a highly leaky cap there could produce an overcurrent problem.
A low value high power resistor between the source of the switching FET and ground is used to detect overcurrent so check that for signs of overheating and check that its value hasn't gone high (would produce false sensing).
Is the screen edge or back lit? (not that it makes much difference but you'll need a large room to fully dismantle this TV to get at the LEDs for individual testing
I started checking it this morning.
When it's working, the main DC is at 333 volts so pretty good, the LX chip pin2 is 2.5v and pin7 is 2.6v...I'm waiting for it to fail now and has been running for about an hour and a half without problems.
Don't know if it's edge lit but I don't think so as it has 4 connectors for the led's. I have a large bench I built for the shed so not a problem to lay it out for working on the led's if I have to, the big problem would be matching the led's.
Last edited by loopyloo; 20-09-17 at 02:22 PM.
If it lasts longer with the back off heat has to be the major suspect?
Got a small hot air blower to focus on individual components? (like one of those gas soldering iron kits)
Last edited by Skepticist; 20-09-17 at 02:30 PM.
Yep sure have.
Got the freezer spray ready too.
Skepticist (20-09-17)
It finally went off. What happened was, I let it run for quite a while, then decided to change the speakers for better sound. During that time I turn the tele on and off a few times while testing the sound then when finally finished, it was only on for a few minutes before the blackout eventually happened.
DCvolts still good, Pin2 the same at 2.5v, Pin7 dropped to zero.
I also checked the inverter switching output from the main board (Adj,Bk) before and after fail. 4.6v and 4.2v with respect to gnd. No change after fail.
So Pin7 was the only one that changed.
I then grabbed the freezer spray and suddenly realized that won't help to bring it back on because it needs power recycling to turn on and recycling always resets it anyway.
I tried using the heat gun to create the fault but it stayed on regardless
Last edited by loopyloo; 20-09-17 at 08:20 PM.
Seems the inputs that can cause FAULT to be latched are
UVS/SYN pin8 under voltage sensing
VSNS pin9 over voltage sensing
ISNS pin10 current sensing
VSNS and ISNS need to stay in fault condition for a defined number of counts before FAULT is latched.
If ISNS exceeds 930mV, FAULT is immediately latched (high overcurrent) - that's the input I'd be checking first - the normal range for ISNS is 100mV to 600mV and it's measured across a large power resistor (2W or more) so check it hasn't gone high in value because it would run fairly warm under normal operation.
You'll need the CRO to check those
Last edited by Skepticist; 20-09-17 at 09:42 PM.
loopyloo (20-09-17)
Still got more measurements to do but here's what I found so far.
The resistor off pin 10 is only an SMD size 2401 so 2.4k and actually read 2250 in circuit.
ISNS sits around 135mV on multimeter. Following photo is a little blurry, at 50mV , 0.2mS you can see the center of the wave is around the same, with negative peaks dropping below the 100mV mark.
In ref. to post #8 component of board, you will see 2 large resistors, one center right is blu,red,or with gold way down the other end, would be 62k I believe but reads like 6M or something in circuit. Don't know what's going on there. I will have to pull it out to get a better measurement and see if I can work out what it's for. The other one at the top should be 1k, I'll check it too but I think it's ok.
I'll do some more after the circuit bleeds out.
At the other end of that 2.4k resistor there's likely to be a large resistor connected to ground (probably around 0.47 ohm 3W or so) and that's the one measuring the LED current. From your measurement, the LED current is a bit too close to the low current threshold so that measuring resistor is unlikely to be the culprit - if the ISNS voltage is less than 100mV for 256 consecutive cycles, FAULT gets latched and the chip goes to sleep so perhaps there's a string or 2 of leds not working either fully or partially.
You said there were 4 strings of LEDs individually connected to the PCB? (unplugging a bad string should make next to no difference) - maybe measure the voltage on each of those LED circuits to see if there are any obvious anomalies.
loopyloo (22-09-17)
All of the led strings are receiving 79v each at their connectors. Maybe I could rig up another connector to check the current draw on each string but at this stage I think I'd rather check some other things first.
There is a resistor before ground off the pin10 resistor but it's only a small SMD like 3mm long so can't be a very high wattage one, and is marked as "0000" (zero ohm), I think it reads about 0.8ohms ohms. Have to put my ESR meter on it.
I connected a computer to display notepad with a full size blank white screen and find it is lit evenly all over, can't fault the look of it. I'm inclined to think the best solution would be to change the sensitivity of pin10 somehow.
Just a thought, maybe one of those orange polyester type caps across the leds has a changed esr value.
Still got more things to check, had another busy day here.
The zero ohm resistor is just a link and the smd version of a jumper to allow a track to go underneath, they can function as fuses too
The current sensing resistor wouldn't necessarily be close to the IC, it should be positioned to minimise the length of track in the path of the full led current (an extra 0.1ohm in that circuit can throw things out and would need to be compensated for)
To increase the voltage swings on pin 10 you'd only need to increase the value of that resistor a little but I'd be looking for the reason the led current is so low first (if that's what's triggering the FAULT state).
Last edited by Skepticist; 23-09-17 at 11:10 AM.
loopyloo (23-09-17)
Don't know how I missed it. 8 x 4r7 resistors sharing the current. Should have been obvious.
The resistance of the batch is around 0.6 ohms with ESR meter. Those diodes are in there too.
I checked the current draw on each string.
1/ 85mA...2/ 85mA...3/ 92mA...4/ 100mA
That's strange I thought, so I went back and checked again.
1/ 100mA...2/ 100mA...3/ 100mA...4/ 100mA.....So they all started around 85mA and climbed to 100mA after a few minutes, and stayed there.
With that total current (360-400mA), the drop across the 8 parallel resistors should always be >200mV so it's strange that only half that was measured earlier (~100mV)?
Removing 1 of the parallel resistors will increase the voltage by more than 10%, a safe way to prove whether the problem is the led current spiking intermittently a little too low and tripping the FAULT counter.
Ref to earlier post:
It's hard to read the text on the board (looks like R30 or R38) but that 62k? resistor you mentioned is the one near the heatsink on the primary side of the board between the transformer and the large 450V cap? The gold band is the tolerance 5% so a 1Mohm reading in circuit is definitely strange and probably means the resistor is open circuit or gone extremely high in value but it doesn't appear overheated. Looks to be part of the snubbing circuit on the transformer primary.
Last edited by Skepticist; 23-09-17 at 08:52 PM.
loopyloo (23-09-17)
Yeah I checked that resistor today out of circuit and it is 62k.
I think that 100mV reading you're referring to is where I measured from pin10 to gnd with the oscilloscope. I'll try a measurement across the resistor bank in the morning. Might be interesting.
The ISNS input pin 10 should be a very high impedance so the voltage drop across the 2k4 resistor will be negligible. It also means the loading of your CRO probe 'could' be producing some drop across that resistor so taking the reading on the other end of that resistor would be more accurate than pin 10 of the IC.
The specs for the IC indicate a range of 85mV to 115mV for the 'low' threshold so you need to get the voltage up to the point where the -ve spikes are above 115mV to rule out low current causing the Fault trigger
The low spikes might be caused by a problem with the caps across the strings too
loopyloo (24-09-17)
Yeah those caps are the orange polyester type I mentioned in an earlier post.
The probes I used with my oscilloscope aren't (intentionally) loaded, it's just a couple of alligator clips on one end and BNC the other end, so they are subject to picking up noise. Tomorrow I will use a proper probe and measure from the other side of that 2.4ohm to gnd, that will be effectively measuring across the parallel resistors.
Last edited by loopyloo; 24-09-17 at 10:10 AM.
The way that backlight supply is set up makes the lower voltage on the HVDC rail a very credible reason for the fault triggering
Check pin 8 UVS/SYN and pin 9 VSNS with the CRO to see if those are close to the threshold for the fault counter as well
Pin 8 needs to be > 0.4V, pin 9 needs to be < 2V or they'll increment the counter
The counter doesn't discriminate between which inputs are causing triggering
If it's just the ISNS input that's on the edge, removing 1 of the 8 parallel 4R7 resistors almost perfectly compensates for the lower HV rail voltage with the PFC disabled
There are voltage dividers for the other 2 inputs that could be easily fiddled to compensate as well if necessary.
Last edited by Skepticist; 24-09-17 at 11:22 AM.
loopyloo (24-09-17)
Pin 8 is 1v.
Pin9 is 1.2v.
Across the resistor batch is 102mV. That should be pretty close because as I see now the four strings are in pairs in series. 1 and 2 in series, 3 and 4 in series.
Sounds like just remove a resistor is the way to go.
Skepticist (24-09-17)
Posting Permissions
Bookmarks