MINIMUM I/O

Now that 8-pin microcontrollers are no longer the laughing stock of the industry, Microchip has reduced the pin count again. Are three I/Os and one input useful? Isn't this getting a bit whacky?

I designed the next circuit using a PIC10F20x series microcontroller that comes in an SOT-6 package (see Photo 1). With two output pins defined as SCK and SDA to control a digital potentiometer or DAC, there was little left for monitoring the JS1100AQ switches. I connected JS1100AQ's independent switch to the microcontroller's input. I then needed to monitor the other four directional switches with the single pin that was left. Not an easy job for a digital input.

I decided to monitor the capacitor of an RC circuit with the last I/O pin. The direction switches (JS1100AQ) connect different resistor values between 5 V and the capacitor, which has its other lead grounded. By using resistor values that are half the previous value, I could get eight different RC time constants charging the capacitor by using one switch or combinations of two switches. As an input the microcontroller could count how long the digital input took to go from a logic 0 to a logic 1. This was dependent on the VIH of the port input, but it was consistent and dependent on VCC. By configuring the same port pin as an output (outputting a logic low), the capacitor could be discharged.

The main loop of execution program monitors the JS1100AQ's independent switch for closure. This reconfigures the RC monitor pin to an input and starts a software loop (12 µs) monitoring the input pin and incrementing a 16-bit counter variable. Depending on the tilt direction of the JS1100AQ, one or more of the corner switches will be closed. Each combination of switch closures creates a different current flowing into the capacitor. Thus, the capacitor will charge at eight different rates. Execution breaks out of the counter loop after the monitor input is read as a logic 1. The monitor pin is reconfigured to an output (discharging the capacitor). The timer's count is compared to a table of values. The closest selection from the table returns the position of the JS1100AQ (N, NE, E, SE, S, SW, W, NW). Branching to appropriate routines can adjust the XVal and YVal variables used to update the SPI device.

Just when I thought I had it licked, I remembered a little note that explained how SPI values would be internally transferred after the raising of the device's CS input. Whoa! I wasn't using the CS and had it tied low (enable). Did I not use all of the I/Os? No, I used all four. What to do?

I was able to tie the monitor pin to the SPI device's CS input (see Figure 6). Timing the capacitor and sending SPI commands were independent operations, so the monitoring pin could do triple duty by monitoring the voltage input, shorting the capacitor to ground, and providing a CS signal.

I chose an RC combination that would be approximately 100 ms. (The value limits the maximum step speed.) The three other values became one-half, one-quarter, and one-eighth of the first value. I needed to find values for the LUT for each combination of switch positions. By altering the program to output a pulse on the SCK pin based on the monitor input, I could measure the pulse width in microseconds based on what the circuit actually saw at the monitor input. I placed the measured values in the LUT to let the program determine which switches the JS1100AQ was closing. If I'd had 1% resistor values available, I might have been able to choose table values based on a single measurement of the highest resistor or perhaps a self-calibration routine during power-up.