In some applications of capacitive sensors, the instrument’s front end must be small enough to fit into a narrow space. Figure 1 shows a precision capacitive-sensor interface for such use. The square-wave output from a low-voltage 555 timer, IC₁, constantly triggers the precision one-shot, IC₂, to produce quasistable outputs for time periods T₁ and T₂, which are proportional to external timing capacitance: T₁=KR₀(C_S+C₀), and T₂=KR₀C_S, where K is the multiplier factor. K is nearly independent of the external timing capacitance when that capacitance is more than 100 pF (Reference 1). So, a 150-pF capacitor, C₀, in shunt with the capacitive sensor, C_S, supplies an offset so that operation of the one-shot remains within a linear range even if the value of C_S is less than 100 pF.

To achieve good measurement accuracy, connect a reference channel with a fixed 150-pF capacitor. This method cancels the effects of both stray capacitance and transition time. A single 3.3V supply powers this interface circuit. The circuit’s compact design permits flexibility, and you can easily integrate the circuit into a miniature sensor head near the measuring point. IC₃ converts the outputs to LVDS (low-voltage-differential-signaling) levels and then transmits these outputs using a standard Category 5e cable to the terminal, which may be some distance away. As long as the cable is shorter than 10m, the transmission bandwidth is adequate for ensuring acceptable measurement accuracy within several picofarads to hundreds of picofarads (Reference 2). In Figure 2, the terminal at IC₄ converts the signals it receives from the interface to LVTTL (low-voltage-transistor-to-transistor-logic) levels and then feeds them to a set of passive filters. Each dc output is proportional to the signal’s duty cycle:



and

where V_H is the high-level output voltage of IC₄ and T_P is IC₁’s oscillation period. By digitizing the two outputs, you can obtain a reading proportional to the sensor’s capacitance, V_O1–V_O2. Be sure that T₁<T_P,—that is, C_S<T_P/(K×R₀)–C₀; otherwise, the final output will be erroneous. For the sake of a wide measurement range, keep T_P as long as the target application permits.