Nuts and Volts - February 2018

chip SRAM and jumps to the application code entry point.

There are only a few minor modifications needed for LaunchPad:

1. Add a resistor divider across the battery terminals and connect to ADC (analog-to-digital converter) pin 58.

2. Open J2 and J3 so that the yellow and green LEDs are not ON in Hibernate mode.

3. Open J13 to supply the board from the J20 battery connector.

4. Depopulate R3 to disable the D1 LED.

5. Depopulate R20 to disable the D4 LED.

6. Connect the reed switch to wake-up input GPIO (general-purpose input/output) #13.

The CC3200 ADC is 12-bit/eight-channel, out of which four channels are available for user applications. The sampling rate is fixed at 16 µs per channel, and the channels are sampled in a round-robin fashion. The pins tolerate a maximum of 1.8V, but full scale is 1.467V.

To keep it simple, I chose to implement a divide-by-two resistive divider. Since this will always be connected across the battery terminals, I went with one meg resistors to keep the current consumption low. The drawback to this approach is that the sampling capacitor current will be severely limited, thereby affecting ADC readings. To fix this problem, I placed a 100 nF capacitor across the lower leg of the resistive divider.

The CC3200 internal sampling cap is 12 pF and is switched on for 400 ns. So, a much larger 100 nF external cap will supply enough current for the sampling cap to reach the final voltage value of the divider in time to get correctly sampled. The simulation in Figure 5 shows the voltage on the sampling cap with and without the 100 nF external cap.

Also, I knowingly compromised the voltage range measurement and bandwidth. This means the ADC input will get full scale when the battery voltage is 2.9V. Also, the voltage change on the pin will be slower. In this application, that didn’t matter since I was only interested in a low point in the battery voltage range. If the battery voltage dips below 2.4V, a text will go out as a warning to replace the batteries.

Assuming two AA batteries (not at full capacity) and an average load current about six times higher than expected, the batteries should last well over a year: 2,000 mAh/0.1 mA = 20,000h 833 days.

Whenever the reed switch input is sampled, it’s also debounced. The pin is pooled every 10 ms, 100 consecutive times.

Only if it reads “1” every single time will it return “success.” This will prevent a false alarm.

To preserve the batteries, the device is kept in Hibernate mode. In this state, most of the SoC is powered down except the RTC (real time clock) and 2x32-bit OCR registers.