MAKE Blog How-to Tuesday posted a challenge: Optimize the sketch code for a pressure sensor bar graph. Here is my second attempt:

int gate[] = {10, 20, 50, 100, 200, 32767};
void setup() {DDRB = 0b00111110;}
void loop() {
  int dval = 0;
  int rval = analogRead(2);
  for (int x = 0; rval > gate[x]; x++) {
    dval = dval * 2 + 1;
  }
  PORTB=dval*2;
}

I'm running my latest handheld computer on solar power. It starts up when the sun rises and runs all day. I started wondering how I could measure the voltage coming in from the solar panel. I came up with a way to use a diode and a resistor and the ADC, but I wondered if there was a way to skip the ADC input.

I could time charging a capacitor, but since I don't know the supply voltage I don't think that would work. More voltage means more current, so the capacitor charges faster. On the other hand, when the capacitor discharges the time is based on the actual voltage. As long as the discharge resistance doesn't change I should be able to time it and get a reasonably accurate voltage.

I rigged up a test circuit using the Arduino. It charges the capacitor (33 uF) to 5 volts when I turn on pin 4. I let the capacitor charge for 1 second, then discharge it and measure the time until pin 4 goes low. The main discharge capacitor is 100K ohms.

I wanted to see what lower voltages would do, so I added a second 100K resistor on pin 3. This formed a voltage divider with the discharge resistor, so the capacitor would only charge to 2.5 volts. The software to drive this is cool: I set the charge pin (3 or 4) to output and drive it high. After 1 second of charging I turn the pin back to input mode. That makes it high impedance (but not high enough relative to the 100K discharge resistor...). Then I set pin 4 low and time how long it takes to read back as low. I'm getting about 172 milliseconds for 5 volts and 70 seconds for 2.5 volts.

I'm not sure about the input impedance of the Mega8 chip. I added another wire to pin 5 so I could charge the capacitor to 5 volts without switching the sense pin (pin 4) back and forth from input to output. When I did this the time went down to 160 milliseconds. I figure that the leakage back into the chip was the cause. All the resistances are in parellel. This means each time you add a pin to the circuit the overall resistance goes down. Perhaps keeping the charge pins high will change things.

I've already built a six pin ISP adapter, but my Olmex cards use the ten pin ISP. I put a header on a scrap of perf board, added an LED for the heartbeat, and then connected the ten pin cable. Actually it is a fifteen pin cable, off a joystick adapter, but it works on one of my Olmex boards. I'll have to cut it down to fit the socket on the 28 pin Olmex board.

The 8 pin board has a 10 MHz crystal on it. I'm hoping that will give me a stable platform for my digital radio project (APRS to Morse Code).