Repairing faulty IKEA Norrfly LED light

[pertinax]

So, after around 3 years, one of our wardrobe lights failed. The 96 cm IKEA Norrfly LED light, with dual sensors to sense when any one of the two doors is opened.
Initially the light failed to turn on when opening the doors, this could be fixed by cycling power. After some time, even this didn't work and the only option was to manually turn the light on by setting the switch to "ON" instead of "AUTO".

After some hassle removing the light I was able to disassemble and troubleshoot the board. See attached images. Looks like the switch either grounds the low side of the LED string (ON mode), or the ground connection to the sensor electronics (AUTO mode). I probed around and found only static signals from the 8 pin controller. Also, both IR LEDs were turned off. After some pondering, I concluded that the circuit with a diode, resistor and capacitor at the controller was a reset circuit. I tried to short the capacitor using a tweezer. This revived the unit and I saw pulsing on the IR LEDs as well as corresponding sense signals from the IR sensors. My guess is that the reset circuit has a too short time constant, perhaps due to small initial margin and subsequent aging of the capacitor or controller. The original capacitance was measured to 100 nF and I added an additional 1 µF on top of it. This seems to have fixed the problem.

It would be interesting to hear if anyone had a similar problem, and any thoughts about the reset fix. Might be a Microchip microcontroller, they seem to have this kind of pinout.

[Ubuntix]

Hello, thanks for the info. I had a similar issue with mine (6-LED only), the LEDs would not completely extinguish. This was due to a leaky mosfet controlling the LEDs, Q5 in the schematic provided. This schematic is not complete, I concentrated on the main functionality. I your case a faulty capacitor seems weird, these 100nF are usually pretty good. Anyway, if the problem is solved... 

[amyk]

It's not uncommon for MLCC to crack under repeated thermal cycling, especially if they were originally stressed.

[pertinax]

Hello again,

Thanks for the input. Nice schematic! Strange fault with the leaking MOSFET. How did you identify the PIC microcontroller? On my board it is marked "48A". I read up on the reset circuit and also the appnote https://ww1.microchip.com/downloads/en/AppNotes/00000607C.pdf about troubleshooting power-up problems for the MCU. The same reset circuit is given in figure 15 of the appnote.

I had a second Norrfly fail last week, as power was cycled due to a new distribution panel installation. I did some more measurements this time. 

Below are attached waveforms before and after the additional 1 µF capacitor. As seen, MCLR (yellow trace) follows the 5 V supply (green trace) almost exactly. With the added  1 µF the MCLR is delayed and the unit functions as normal. The orange trace is +24 V supply from the IKEA TRADFRI (ICPSH24-10EU-IL-1)

Regarding the MLCC possible cracking, I did measure the capacitor to be 100 nF after having desoldered it. Perhaps the fault is due to a bad design choice for the reset circuit, or the power supply has degraded with time leading to a increased startup time which the reset circuit cannot handle.

[ShocWave]

Hello, thanks for the info. I had a similar issue with mine (6-LED only), the LEDs would not completely extinguish. This was due to a leaky mosfet controlling the LEDs, Q5 in the schematic provided. This schematic is not complete, I concentrated on the main functionality. I your case a faulty capacitor seems weird, these 100nF are usually pretty good. Anyway, if the problem is solved...

I also have an issue with the LED not shutting off in auto. It dims quite a bit, but not all the way.
Is there an easy way to check the component, and can it be easily purchased?
I'm not too good at electronics, but I do have a soldering iron.

[rteodor]

Nice schematic! Strange fault with the leaking MOSFET.

I don't know what to say about that schematic: Q5 is FDN340P that is a PMOS with max VDS of -20V with marking '340M'. In a 24V circuit! That could not be right.
In your pictures Q5 is marked 'J6' (did I get it right ?). J6 seems to be NPN S9014 50V/100mA. This NPN looks more appropriate.

I did measure the capacitor to be 100 nF after having desoldered it.

MLCC loose capacity both with time and voltage but hey can be healed thermally. By unsoldering that cap you might have done just that.

There is at least one thread on this forum discussing MLCC ageing but on a quick search I found only this link:
https://passive-components.eu/high-cv-mlcc-dc-bias-and-ageing-capacitance-loss-explained-2/

[rteodor]

Hello, thanks for the info. I had a similar issue with mine (6-LED only), the LEDs would not completely extinguish. This was due to a leaky mosfet controlling the LEDs, Q5 in the schematic provided. This schematic is not complete, I concentrated on the main functionality. I your case a faulty capacitor seems weird, these 100nF are usually pretty good. Anyway, if the problem is solved...

I also have an issue with the LED not shutting off in auto. It dims quite a bit, but not all the way.
Is there an easy way to check the component, and can it be easily purchased?
I'm not too good at electronics, but I do have a soldering iron.

Try to simulate an un-soldering: just melt the solder at one or both ends (C4 in the OP picture) as you would when unsoldering it then check if that recovered the normal functionality.
It would be informative to everyone if you could measure the capacity before and after that simulated unsoldering but this is only to satisfy our curiosity.

This simulated unsoldering is, as mentioned in the reply above, and attempt to heal/rejuvenate that MLCC.