The OZ9902C from O2Micro sits at the heart of countless mid-range LED-backlit televisions built between roughly 2012 and 2020. Repair benches across the world know the part by sight, since boards built around it tend to land on the workshop table with the same handful of symptoms. The chip is reliable when its supporting components stay healthy, but it lives inside a high-voltage, high-current, hot environment that grinds down everything around it over the years. Knowing how the IC actually fails, and how to tell its failure from a fault in a neighboring part, saves hours of guesswork.
What the chip does inside the backlight section
The OZ9902C is a single-channel high-power LED driver in a 16-pin SOP or QFN package. It controls a boost converter that lifts the main 24V or so DC rail up to whatever forward voltage the LED matrix demands, often in the 60 to 150 volt range, then regulates the current through the strings using an external balance transistor. Inside the chip live an oscillator, a pulse-width modulator, gate drivers for the external switching MOSFET, a current-sense comparator, overvoltage protection, overcurrent protection, and the logic that handles the enable signal coming in from the main board.
The sibling part, the plain OZ9902 (without the C suffix), is a dual-channel version in a 24-pin SOP or SSOP package. Service technicians sometimes confuse the two on schematics, since the markings look almost identical under a magnifier and the basic function is the same. The C variant runs a single boost converter; the dual version runs two with 180-degree phase shift. Mid-size and mid-range LCD TVs from brands like TCL, Hisense, Skyworth, BBK, and Polar tend to use the single-channel C variant, while larger or higher-end sets reach for the dual.
The classic symptom: backlight flashes and dies
The most common complaint that lands an OZ9902C-equipped set on the bench reads almost identically every time. The TV powers up, the standby LED behaves correctly, sound comes through, the screen flickers visibly for a fraction of a second, then goes black for good. A flashlight aimed at the panel reveals a sharp menu image, confirming that the T-CON and main board are doing their work. The fault sits squarely in the backlight chain, and nine times out of ten it sits in the loop that includes the OZ9902C.
That fraction-of-a-second flash carries diagnostic weight. It proves that the BL_ON signal arrived at the driver IC, that VCC came up, that the boost stage began switching, and that the LED strings briefly conducted. Something then triggered a protection latch inside the chip, which is a deliberate response to an out-of-spec condition rather than a chip fault on its own. The job becomes finding which protection circuit fired and why.
Reading the board before powering it
Visual inspection earns its keep on these boards. The OZ9902C lives near a small forest of high-voltage parts: a boost inductor in the millihenry range, a TO-220 or D-PAK MOSFET, a fast Schottky or ultrafast silicon diode, a bulk electrolytic capacitor rated around 100 to 250 microfarads at 100 to 160 volts, and several small SMD ceramics. Heat damage shows up first on the electrolytic cap, which tends to bulge slightly on top or vent a brown residue around its base.
The MOSFET package frequently develops a faint discoloration on the drain pad of the PCB, where switching losses have been turning into local heating for years. The catch diode body sometimes shows a chalky or cracked appearance after a single overcurrent event. Around the OZ9902C itself, look for a brown halo on the silkscreen near the VCC pin and a darkened spot under the IC body, both signs that the chip has been running near its thermal limit.
Cracked solder joints on the inductor leads are another classic finding. The inductor is heavy, the board flexes during shipping, and a single hairline crack on one lead creates an intermittent connection that can survive years of normal use before finally breaking the circuit during a cold start.
Probing the key pins with a multimeter
A diagnosis starts with the board powered off and the multimeter set to diode mode. The first measurement targets the supply pin of the IC, usually marked VCC on the datasheet. A reading of zero ohms between VCC and ground indicates a shorted bypass capacitor or a shorted internal regulator, and the chip has to come off the board. A reading of normal capacitance behavior, climbing slowly from zero to a few kilohms as the meter charges the local bypass cap, confirms the supply pin is intact.
The gate-drive pin sits next on the list. Diode mode between this pin and ground should show a normal junction drop in one direction and a high reading in the other, similar to the gate pin of a MOSFET. A short to ground here means either the IC's driver output stage has died or the external gate resistor has failed shorted, dragging the IC's output down with it.
The current-sense pin and the feedback pin both have very small voltages on them during normal operation, typically below 500 millivolts. A short to ground at the feedback pin tricks the chip into thinking the LED string is drawing no current and pushes the boost converter to maximum output, which then trips overvoltage protection. A short on the current-sense input does the same thing through a different path. Both pins should read open or show only the bias network resistance to ground.
Powering up with the protect line lifted
When the static checks pass and the chip seems alive, the next step is to power the board while watching the right nodes. A scope is ideal, but a careful tech can do most of the work with a fast digital meter. The interesting points come up in a specific order:
- Standby comes up first, putting roughly 3.3 or 5 volts on the VCC bypass cap;
- BL_ON rises from zero to its active level, usually 3.3 or 5 volts depending on the platform;
- VCC at the IC pulls up to its operating point, normally between 9 and 18 volts depending on the supporting circuit;
- The gate-drive pin begins switching, which a DC meter sees as a steady reading somewhere between one and four volts;
- The drain of the boost MOSFET shows a switching waveform whose average lands a few volts above the main DC input;
- The LED+ output rises to the working string voltage, and current begins to flow through the balance transistor.
If the sequence stops at step three, the chip is not getting enough VCC, and the bias resistor or bias winding on the boost inductor needs investigation. If it stops at step four, the chip itself has failed or its enable logic is being held off by a stuck protect line. If it stops at step five, the boost stage hardware has a problem rather than the controller. If it stops at step six, the LED matrix or its current-sense path is the culprit.
The three failure patterns the chip itself shows
Out of all the OZ9902C-based boards that come in for repair, the chip itself is genuinely the fault in only a minority of cases. Three patterns account for almost all true chip failures.
The first is a hard internal short between VCC and ground, easily caught by the very first diode-mode check. This usually follows a catastrophic event upstream, such as a failed bulk cap that dumped reverse voltage on the supply, or a transient on the AC line that punched through the regulator inside the IC. The chip pulls the VCC rail down to roughly zero, the standby supply enters foldback, and the whole set refuses to start the backlight at all.
The second pattern is a degraded but not fully broken gate driver. The chip switches, but the gate signal is weak or asymmetric, and the external MOSFET runs hot even at moderate current. The boost rail comes up to the right voltage briefly, then drops as the MOSFET enters thermal foldback. This pattern often goes undiagnosed and gets blamed on the MOSFET; swapping the transistor brings the set back for a week or two, then the new MOSFET dies and the cycle repeats. Replacing the IC at the same time as the MOSFET breaks the loop.
The third pattern is an unstable internal reference, where the chip works but regulates LED current to the wrong value. The screen lights up dimmer than it should, or flickers visibly at low brightness settings, or the protect logic trips intermittently because the feedback never settles at the expected level. This is the hardest fault to confirm without a scope, since static measurements all look reasonable.
Why the IC takes the blame more often than it deserves
Mid-range TV designs run the OZ9902C and its supporting parts close to their ratings. The bulk electrolytic on the boost output dries out slowly, and as its ESR rises the ripple voltage on the boost rail grows, which the chip's overvoltage protection eventually catches. The result looks like a chip fault but is really a capacitor fault. Replacing the cap brings the set back to life. The same logic applies to the small ceramic on the feedback pin, which can develop a hairline crack from board flex and start coupling switching noise into the regulation loop.
LED strips themselves age too. Each individual LED in the matrix sees thousands of hours of forward conduction, and the forward voltage of an aged LED creeps up. The driver pushes its output higher to maintain the rated current, sometimes far enough to trip overvoltage protection. The chip shuts the backlight off to protect itself, which then looks identical to a chip failure. Bench-testing each strip with a regulated current source isolates this case in minutes.
The honest summary is that a dead OZ9902C is rarely the root cause of a no-backlight TV. It is much more often the messenger, doing its protection job correctly in response to a fault somewhere else on the board. Treating every dark screen as a chip swap leads to repeated returns; treating it as a system-level diagnosis, with the chip as one component among many, leads to repairs that actually hold.
Our Telegram Channel
Our YouTube Channel