josh.com

Inductive power on the cheap – Philips Imageo Tea Light Teardown

How can Philips sell a set of 6 inductively rechargeable LED lamps with charger and glass holders for less than $80 retail? It takes some careful and clever value engineering. Let’s crack one open and see if we can figure out what tricks and techniques they used!

The housing has just 3 straight pull injection molded parts:

  1. opaque base plate
  2. opaque top
  3. translucent “flame”

The base plate has ribs for holding everything in place. There are no screws anywhere. The guts can just drop in and then the top is probably spin or ultrasonically welded on. Super fast and simple assembly, and the finished unit is sealed with no way for water to get in.

Inside we find…

  1. Inductive receiving coil. No expensive Litz wire or ferrite field concentrators here. Just enameled wire with some shellac to hold it together after it is removed from the form.
  2. Two through-hole amber LEDs. These are crazy cheap. The LED leads are bent over to position the light up inside the “flame” protrusion without requiring any wires, connectors, or daughter boards.
  3. Two 300mAh, 1.2 volt Nickel–metal hydride batteries. These are cheaper and easier to charge than Lithium Ion. By using two in series they get 2.4 volts nominal output which is perfect for driving the LEDs without a voltage converter.
  4. Circuit board. And what a fine circuit board it is!

The board is single sided FR-4 with punched holes. I love boards like this. They take more work to design, but the effort can be worth it for low cost, high volume products . My guess is that you could make this little gem for less than $0.10 in the USA and even cheaper in China.

On the front of the board, notice that the copper is only etched enough to isolate the traces from each other. The less copper you etch, the longer your etchant will last. It matters when you are etching millions of boards. I wish more design tools supported isolation routing like this rather than typical fixed width traces.

There is a white silk screen pattern around the off-board LED solder connections. This probably makes it easier to inspect for solder bridges. Clever!

There is more functional silk screen under 3 of the pads on the MCU…

I think the screen here serves  as an additional layer of resist and/or to help make it easier to inspect for bridging.

But why not just put unconnected pads under these pins? Routing a single sided board can be hard and sometimes you need every path you can get. There are traces running under these unused pins that probably couldn’t have gone any other way. Why not just run those traces between the unused pads? That would require a smaller minimum clearance, which would require a much more expensive fab process. This is a better solution.

The fact that there are unused pins on the MCU suggests that this is an off-the-shelf component and not a custom chip. It is marked “U-tek P38970”. I’ve never heard of U-Tek, but my guess is that that is a general purpose MCU similar to a AVR or PIC. If you know anything about U-Tek or this chip please let me know in the comments!

These little SOT23’s contain dual switching diodes, which saves 2 parts compared to using individual diodes in this full rectifier.

Finally, we move on to the only part on the back side of the board – a through-hole tilt switch…

This is the only moving part and the only sensor in the unit. The user turns the candle on and off by  turning the candle upside down. This causes a little ball in that tube to roll and either make or break contact between the leads.

A button would have required much tighter tolerances, and would have also made it tougher to keep the case water tight. These are super cheap and probably more reliable than a button and a great choice for this application.

The only disappointment I have in this otherwise lovely board is that it uses 56 ohm resistors to limit current though the LEDs…

These could probably be eliminated by using PWM. This would…

  1. reduce the the parts count by 2, because every resistor matters.
  2. reduce power usage by about 2mA. This could potentially let the candles run a little longer on a charge.
  3. allow smooth dimming of the LEDs for a more natural flicker.

Next time we’ll dig into the circuit design!