LED Nightdark Capacitor Hernia

We have had some color changing LED night lights for maybe 15 years. Last night one nightlight became a nightdark. I opened it up to see whether there might be an easy fix. I was not necessarily expecting an obvious failure mechanism. I laughed when I was surprised by dramatic evidence of failure of (what is probably a metallized film or film-foil capacitor, probably used to limit the AC current.). The photo shows what seems to be a 6 mm long blob of aluminum which found its way out of a crack in the capacitor epoxy case. I don’t think I will try to repair it, largely because I lost a few small plastic parts when I opened it up. I probably have some replacement capacitors which could function correctly but they would not fit inside the case.

Nevermind the hernia, magic smoke escaped, he’s a gonner.

I just make my own color changing LED night lights using KiCad. To avoid having to deal with 120VAC, I put a USB A plug on the back, so it can plug directly into a USB charger.


Using a capacitor as a mains voltage dropper, without the heat, used to be a common trick and is prone to this effect

Mine’s bigger than yours :slight_smile:

1 Like

I am not sure what this is, but in context it is pretty funny…

Beautifully executed! I love this idea.

I am wondering what sort of chip controls the LEDs. My sample has another blob (this appears to be black epoxy intentionally placed on the pcb) which may conceal the “brains” of the night light. When this one was working, it cycled through (probably 7 = 2^3 -1) colors and with varying intensity; a pushbutton allowed the user to freeze the light at one color-intensity setting.

A small microcotroller with an ADC as I see an LDR light sensor

This is common on many inexpensive, mass produced products. The bare silicon chip is glued to the PCB and wire bonded directly to the PCB. Then a blob of expoy is used to seal it up. The technique is called COB, which stands for Chip-On-Board. Sparkfun did a write up on this a while ago, here:


1 Like

I “had to” design electronics with capacitor power supply for a while and this is typical for cheap input capacitors.

You will need a high quality capacitor made for this purpose - in Germany it has to carry the “X2” specification.
These capacitors are typically in a rectangular plastic housing and yes, they are bigger and more expensive.
As far as I remember, you should calculate with aprox. 50mA per 1uF (230V !).
And replace the ELKO also. check the Zener-Diode that limits the voltage of the capacitor supply. The charge resistor should be a fusible type, but they are not very common.

With high quality components this type of power supply can perform for many years though it is sensitive to pulses on the AC line.

Thanks for that. I am familiar with photocells but not the term “LDR”. That is OK; I looked it up…

Thank you. Checked out the SparkFun link. It is particularly endarkening. :grinning: In the Sparkfun photo, it looks like the board can accommodate a leaded chip. But on the nightlight I guess that there is no room for that.

I am encountering a couple of shifts in mindset. One of them involves taking seriously a company with the name “Sparkfun.” That is because I have previously enjoyed (and risked) spark fun.

I am in the USA with 120 VAC 60 Hz. My calculations indicate that 100 nF will produce a capacitive reactance of 26 K ohms and permit a current of ~ 4 mA. (rounding down to allow for voltage across the rectifier which would be in series with the capacitor.) My guess is that the capacitor is smaller than a proper X2 type and probably less robust.

Incidentally; in the night of the failure we were having storms in our area; subtle undulations in our lighting strongly suggested some power line transients.

1 Like

Sparkfun is very much like Adafruit.
Don’t ask me where they find those names.

I probably am less familiar with Adafruit than with SparkFun. I have barely heard of Adafruit and wonder whether it is related to Adaboy. :grinning: But anyway SparkFun reminds me of when my colleague and I connected a TO92 SCR across a 5600 uF (I do not remember exactly) “computer grade” electrolytic capacitor using short fat wires. The cap was charged up to 200V. We triggered the SCR from maybe 10 feet away with long wires. There was a “bang” (not surprising) after which the TO92 case had vanished.

Now that’s a monostable, a.k.a one-shot circuit. :rofl: