Capacitive Sensing on Battery Power

I'm helping out with a project for the 2015 Alameda County Fair in Pleasanton, CA. The project is not finalized but includes touch sensors. The project lead has chosen MaKeyMaKey as an input device.

MaKeyMaKey works by sensing current flow through an object, such as a banana, through your body, toward ground.

But it requires the user to touch a ground wire or contact as well as the object you want to sense.

I was interested in getting rid of the ground wire so I looked into capacitive sensing.


There are a lot of articles about capacitive sensing, including the documentation of the Arduino CapSense library. Here's a few:

CapSense library documentation
Turn a pencil drawing into a capacitive sensor
3D position sensor using three capacitive sensing panels

You can find many more by searching for "capacitive sensor arduino", for example. Most of these use the CapSense library. The library tests how long it takes to charge a receiving pin to +5V using a circuit like this. (There are minor variations in the circuit in the various projects on the web.)

In this circuit, a human touching or coming near a contact point forms a capacitor which affects the time it takes to increase the voltage at the receiving pin. The resistor R is a large value, say 1MOhm to 10MOhm, which causes the voltage to increase to +5V on a negative exponential curve.

Experience with CapSense and Similar Approaches

The CapSense library certainly works, but with some caveats:

• The circuit does not seems to discharge completely, making the "no touch" values increase from their power-on state. However, they seem to stabilize.
• The wires in the circuit are also sensitive to proximity, especially as the resistance R increases. Several projects suggest using shielded or twisted-pair wires to reduce this.
• I've had difficulty getting any of the projects to work on battery power. It appears that the discharge process is much slower while powered by batteries, possibly because the ground fluctuates.

I've also written my own code to time the charging, with good success. But I was also unable to get the project to work on battery power.

Measuring Discharge Timing

Another approach was suggested in a project to create a touch-sensitive table lamp. Instead of measuring the time to charge the circuit, the approach is to completely charge, then cut off the input supply and measure the time to discharge. The author also reported problems with the time-to-charge approach and found this technique by searching the web. I've made some minor modifications to his circuit:

Here a 1N4148 diode is used, as recommended by the author, to avoid having to change the output pin mode to high impedence after charging. I added a resistor to limit the charging current through the diode to keep the pin output well within safe limits. The charge current will be 5mA initially and decay exponentially. The total charge time is under 500ns.

The discharge resistor has a high value, here 10MOhm. The total capacitance of the human plus the object will be in the range of tens to hundreds of pico-Farads. That resistor value makes the discharge time at least 200us when no touch occurs. (With no touch object connected the time is about 225us on my setup.) The discharge current starts at .5uA and decays exponentially. The time to discharge goes up by at least 100us when the object is touched. This circuit has good sensitivity. I've tried it with coins, foil plates, bananas, and plant fronds. All require only a light touch to sense. Another plus with this approach is that there is no need for shielded wires. The downside to this approach is that it appears to be much less sensitive to non-contact sensing, albeit with greater reliability than the time-to-charge approach. The variance of the timing measurements is quite small.

Making it Work on Batteries

I still had a problem: it wouldn't work on battery power alone. (In my case I'm using 6xAA = 9V as input to the Arduino.) However, at one point I unplugged the USB cable at the computer end instead of the Arduino end, turned on the battery, and it worked! I assumed the USB cable formed a ground stabilizer of some sort, so I connected some 28 gauge wire to the ground pin and strung out about 3 feet. It worked intermittently. Rolling out another few feet and it was reliable. The whole spool was still attached, though rolled up, so I don't know exactly what length of wire would give a good ground. Replacing the cable and wire with a couple of feet of 1/2" copper pipe (about $3 at OSH) makes a reliable circuit. I'm going to try further experiments to see if a shorter piece of pipe or, perhaps, aluminum foil will also work.