I got for free a laptop with broken screen. The backlight in the screen is a T0.8 CCFL lamp (single pin on each side) along the bottom, powered on an inverter (dimmable HF ballast)

The initial fault of the laptop was just the broken screen. I asked friends etc and one happened to have a used but good LCD panel of the same type

I connected the new LCD and there is display but no backlight whatsoever. I then tried to connect again the CCFL of the original (broken) panel, but it won't work anymore

The inverter have a trnasformer, where the secondary coil is partitioned into "partitions" with plastic dividers (probably to not let turns with high voltage btween them to be closwe together). The entire coil (including the primary, 1 partition, no much turns with thick wire) I noticed that the 2 last "partitions" of the high voltage coil on the high potential end are kinds "smoked" under the tape

How can a CCFL make such failure (if its open the inverter can detect this and stop, if its shorted why would it smoke exactly those 2 partitions)

What was going on there and how can i test the "new" CCFL without blowing a replacement inverter too ?

The older (simple Royer oscillator style; you recognize such concept by common secondary and series capacitor connected in series with each lamp branch) inverters usually blow up when the lamp fail, because a lot of power get transferred, instead of to the lamp, to places not designed for that (dielectric losses of the bobbin material between the top HV section and the core, switching transistors, primary winding - depend on the particular design and a "bad luck").
But usually this take at least few seconds before it reaches the dangerous temperature, so you may try the lamp with new inverter - if it start up immediately normally and there is no excessive consumption or heat in the inverter (mainly the transistors), it should work well.

Newer designs with dedicated IC's usually contain open load protection, so they would survive the open load.

But even the simple Royer oscillator based inverter board does contain some IC's, those control the dimming (turn the power ON and OFF by a PWM of few 100's Hz into the main Royer inverter), so the fact the board contain IC's does not mean it is the more intelligent ballast.

The inverter is a new design with all EOL protections in place. It is all 1 chip (with the driver mosfets inside) MPS MP1010B, there are protections from open and shorted output

What i dont understand is what exactly happened

If the new lamp is open, the EOL protections should have protected the circuit. I seen many of those with blown lamps (driven to vacuum loss), and they work perfectly as soon as the lamp is replaced

If the fault is a short (to earth, the design is such that it is not possible to have a short just acros the lamp as the wires are going in opposite directions), and in case the protection failed (or it was not designed to handle a short to earth), i'd expect either the driver to blow, or the shorted transformer to overheat uniformly (so label over all partitions warped, not 1 smoked partition and everything else fine)

Then the single smoked partition could mean the transformer was faulty there - bad insulation between turns, so short some of them out. With this fault initially the inverter may appear to work correctly, but the shorted turn heat up, locally overheating the turns around. In this initial stage the related total losses are no so large, so the controller's protection do not trip yet.
As this elevated temperature gradually degrade the wire coat, more and more turns get involved in the short, so the dissipated power in this short rise, till it become high enough to teat up the section. Other sections remain cold, as the separation insulate them thermally as well, so the heat does not transfer to them.
And when this failure chain progress beyond certain extend, the losses become so great, the power stage stop to operate in the designed mode and so the controller respond and shut it down, so the ballast stop working.

The label was not warped there. How can it be hot enough to damage enamel isolated wire n the coil, but not hot enough to warp the plastic label (which is in contact with the outer wire layer in the coil) ?

The insulation is not true enamel (= glassy substance, quite robust against high temperatures, but mechanically quite fragile - tend to crack with sharper bends,...; used for temperatures up to 200degC and only for devices requiring to operate at high temperatures, as it is more difficult to handle), but a lacquer, plastic with not as high maximum temperature (usually 125degC), mainly because it does not have to be scrubbed out for soldering (it melt easily), so the transformer assembly is faster, simpler and so cheaper. The bobbin material then could easily endure higher temperatures (ABS could easily withstand more than 200degC; the tape used for wraps endure even more - 300degC solder tip does not melt it, only the unregulated solder gun capable to reach 500degC) than the wire insulation lacquer (frequently melt around 180degC).

Nope the label is kinda scotch, it melts very easily, sure easier than the wire isolation

Update

I tried to scratch away the label, the white stuff is not smoke but a filling material like resn, perhaps its supposed to be there

Anyway, the fuse is blown. I tried to bypass it with a smal resstor (touch across it with the 2 pins of a 8W T5), as soon as i do this the laptop reboots so i think the switchong transistors are blown. As they are in the controller chip (so i cant replace them) i'll get a new inverter board

Looking on Ebay for an inverter, i see that some sellers sell this exact one in "manufacturer refurbished" condition, and some a compatible one from another manufacturer in "new" condition, Does this mean that there is inherent problem with all of them from this manufacturer so better get the other one ?

I don't think the output transistors are on the controller IC chip, that would be too expensive solution.
But power transistors tent to be in rather small packages: SOT23-6 could house two 3A/35V FET chips (for 8..28V CCFL inverter you need four power transistors), so could be actually in a way smaller package than the controller IC.

Power transistors require large silicon area, while suffice with rather cheap, few step and rather rough lito process. On the other hand the controller require finer and rather complex process in order to integrate all the controller functionality and reduce the silicon area (to maintain low cost). Fabricating all on one silicon would mean large silicon area for power stage on an expensive, high density complex process (and moreover integrated power transistors are way less silicon area efficient than their discrete counterparts, so power devices are integrated only for low power or when it is technically the only feasible way - e.g. required good matching, protection, high speed drive,...)

They are in the chip : MPS MP1010B

*I see, definitely not the cheapest solution for that task, maybe for really low power it could be working.

It seems to be the resonant start concept (I've seen with Maxim's controller with four external MOSFET's).
It operate at fixed frequency (resonance of the transformer leakage and C9+C10 combination), while the 1'st harmonic amplitude is controlled by the pulse width on OUTx (they are most of the time low and alternatively pulsing high - in order to keep C7 and C12 charged).
This topology cause quite high current load of the output stage, mainly on the bottom side transistors. On higher junction temperatures it mean their body diode may start to conduct, what mean on the common substrate parasitic bipolars turn ON and cause excessive dissipation (drain of the bottom side transistors become bipolar emitter, the substrate become base and the drain of the top side switches become a collector and dissipate a lot of extra heat there). On top of that when some PNPN structure (you can not avoid that in any CMOS based process, you may only influence it's sensitivity; the SOI help only partly - the dielectric insulation is unusable for dense control part due to required chip area) appear to be somewhere close by, it may trigger and short out the supply (with immediate deadly consequence for the chip)

UPDATE

The replacemant from Ebay arrived, and.... It is the wrong model, the connector does not match the cable in the laptop

I mailed the seller, he said he has no inverters of the correct model in stock so he knowingly sent me this one instead And he wants the inverter back to give refund. Shipping the inverter back will cost allmost as much as buying another inverter from another seller, so i better just do that and keep the wrong model inverter

Before i buy a new inverter from another seller, i tried to hack the new inverter to work in 2 ways :



The chip in the old and new inverters is identical

The old inverter have 7 pin connector with +v, earth, and control signals to pins 2,4,5 in the chip. The connections are ordered like this :

+v
+v
4
5
2
earth
earth

The new inverter is similar except it is 6 pin, and control signal to 2 is skipped :

+v
+v
4
5
earth
earth

So i changed the conenctor (left the wire from the lapptop to 2 disconnected), no light




Then i tried to return to the old inverter, and repair it with parts from the new one. I soldered back the correct connector the way it was originally, and replaced the transformer, chip and fuse with the ones from the new inverter (the chip is the only "silicon" component in the circuit)

Still nothing - low voltage in ok, high voltage out absent entirely (not even a single flash at startup)




Is there anything else to try

I think better would have been to connect the notebook to the new inverter. You said, it uses the same chip, so the main control should be the same.
Trying to desolder components is asking for their damage. MAinly SMD capacitors are sensitive to thermal soldering cycles...
The connection to pin 2 is strange, it is the sense pin to sense the real lamp current (so to allow to regulate it to desired value), so it should not be connected anywhere else.
What came to my mind: The supply connector is used to connect the lamp current return, for the communication with the NB are the pins 4, 5 and maybe 1, but for sure not the pin 2.

Thats what i tried 1st (before trying to desolder the chip) - i moved the conenctior from the old to the new inverter, so that +v, earth, and control pins 4 and 5 connected the same, pin 2 open

It did not work

Then is when i tried the opposite - return the conenctor to the old inverter, and replace the transformer and the chip in it with those from the new inverter

It did not work either

In both cases chip getting 12V, stays cold, and no HV output at all, not even briefly at startup

The question is, if the chip does not contain some configuration, what may differ (e.g. the communication with the notebook,...). I wasn't able to find any decent datasheet, so it is hard to say.