June 2005, Issue 179

Accurate Capacitance Meter
Cypress PSoC High Integration Challenge 2004 Contest Winner

COUNTER & ADCINT12

The Timer32 counter is enabled while the voltage across C_X is in the range defined by the CompLow and CompHigh thresholds. Timer32’s 32-bit counter is clocked by SysClk to 24 MHz. This enables it to count for almost 3 min. without overflowing.

The time measured by Timer32 is equal to CNT/SysClk. Substituting everything we have so far in Equation 1, we arrived at Figure 4d, which we’ve rearranged and presented as Figure 4e. The latter is the cornerstone of this application. Let’s analyze it further.

As you can see in Figure 4e, we grouped the parameters that affect the measured C_X value in two sets. The first set consists of numerical values we can configure (DAC, Ref_HIGH, and Ref_LOW) or measure (CNT). The second set contains the system’s physical parameters. It’s clear that measurement precision depends on the accuracy of these parameters.

The SysClk is stabilized with an external quartz, so you can assume it’s sufficiently accurate. R_SET is an easily obtainable, accurate reference resistor. In addition, you can measure R_SET and V_DD with a common digital multimeter so long as it’s sufficiently accurate. The only physical parameter that you can’t measure directly is V_REF. For this reason, a calibration procedure will measure it internally with the ADCINT12.

We use only the ADCINT12 to measure the V_REF value. Because the ADCINT12 measures the output of BufAmp relative to V_REF, we can calculate V_REF from the ADCINT12 reading. Figure 4f shows the logic behind it. This measurement should be taken when a capacitor isn’t connected. It’s possible to do so because the DiffAmp switches off during the process. The BufAmp normally operates at a unity gain and outputs AGND. In order to have a meaningful reading on ADCINT12, we set BufAmp to a higher gain of 1.33 for the V_REF measurement procedure.