Centrifugal activated power supply
I mentioned that I will be making a new board to drive my matrix display and I have. This here is a DC-DC power supply for the new matrix driver. What makes this interesting is the method I used to turn on the power. I have plans to encase everything (RGB matrix + driver board + power supply board) up once it’s completed. There will be no external switches whatsoever. So, how does one switch on the power without any accessible switch? My answer is to use two regular rolling-ball tilt switches.
Before I explain how the tilt switches work, let me just say that the DC-DC boost converter is a TPS61030 by Texas Instruments. Like a lot of other boost converters, the chip has an input pin that is used to enable/disable the converter. However, unlike most other converters, disabling this chip also disconnects the load from the output. The output of other converters on the other hand usually floats at a diode drop below the battery voltage which means there will still be current drain. It’s a feature that’s really good to have, but can easily be added with a few extra components.
Anyways, there are two tactile switches on the board to turn the chip on and off. These are for backup and debugging purposes. Also, the off switch doesn’t work right now (more on this later). One might be wondering how tactile switches, which are momentary, keep the chip enabled? What I did was route the output voltage back to the EN pin via a p-type mosfet. The mosfet is pulled to ground via a resistor so it normally behaves like a short. When the on switch is depressed, the enable pin is shorted to the batteries, thus turning on the chip. Once the chip starts to pump out voltage at the output (which happens relatively fast), the chip can maintain a logic high on its own enable pin even after the switch is released. The mosfet is there to break this feedback loop. This is done by simply applying 5V to the gate of the mosfet which can either come from the off switch or a connected microcontroller. This works surprisingly well but I should have placed a pull down resistor at the EN pin.
Connected parallel to the on switch is a pair of tilt switches, which are wired in series. The tilt switches have two leads which are closed or shorted when the ball inside rolls to the end closest to the leads. The switches are mounted horizontally on the board, facing opposite directions, and are colinear. The leaded end of each tilt switch is closest to the respective board edge, and the other end points toward the center of the board. Since the tilt switches are wired in series, both have to be closed at the same time to short the enable pin to the batteries. In order for that to happen, the balls will have to move outwards from the center of the board. This cannot happen under normal circumstances due to the way the tilt switches are mounted on the board, no mater how the board is tilted or shaken.
This is where centrifugal force comes in. I’m too lazy to explain it here, so go look it up on wiki. What happens is when the board is spun about an axis normal to the board and passing through the center, the balls get flung outwards. Both tilt switches are closed in this case and the chip turns on. Note that a diode is in series with the switches. This prevents the output voltage of the chip from flowing back to the battery in the event that the tactile on switch or tilt switches is triggered while the chip is already on. This also makes it possible to use the very same switches as inputs to a microcontroller by connecting the node of the diode anode to an input pin.
Check out the video below to see it in action.
An LED, in series with a current limiting resistor, connected from VOUT to GND is initially unlit. It remains unlit (the chip remains disabled) even after some vigorous shaking. However, after a simple twist motion the chip is enabled and the LED turns on. Sorry for the constant annoying clicking noise in the video, it was the camera trying to focus.
Notice I used a wire to turn off the board instead of depressing the off switch. I actually routed the switch incorrectly. When the switch is depressed, it’s supposed to short the gate of the mosfet to VOUT. Instead, I made a routing error and the switch actually shorts the gate to battery voltage, which is insufficient to turn off the mosfet. Luckily, a connected microcontroller could easily supply the needed voltage.
Aside from the routing mistake, I did a couple of soldering mistakes. I initially soldered the tantalum capacitor in reverse. This capacitor is the black component closest to the tactile switches. Usually, tantalum capacitors are yellow, and a stripe indicates the anode end. Since this particular tantalum was black with a white stripe, I momentarily thought it was a diode. This made me soldered it in reverse because a stripe on a diode indicates the cathode end.
I also soldered a resistor of the wrong value. This is the resistor used to pull the mosfet gate to ground. I did a lousy job at desoldering it and destroyed some of the solder mask around the pads. This is clearly visible in the 2nd picture of this post.