Conditioning techniques for real-world sensors

Modern sensors detect a multitude of real-world analog attributes—temperature, force, pressure, humidity, flow, and power, just for starters. In turn, they typically output some level of voltage, current, charge, or resistive analog signal, or a purely digital signal, in proportion to their respective environmental stimuli. Some sensors operate autonomously; others need power supplied, typically in the form of a voltage or current source. Many times, unique signal conditioning is needed or incorporated to provide a useful electrical output. Here, we look at some state-of-the-art techniques for sensor signal conditioning used in modern analog electronics.

As the need for highly precise operational amplifiers continues to grow, the self-correcting architectures—designs that continuously correct for offset error—have become increasingly popular. Many leading amplifier manufacturers use “zero drift” to refer to any continuously self-correcting architecture, whether it is an auto-zero or a chopper-stabilized topology, observes Kevin Tretter, principal product marketing engineer at Microchip Technology Inc. Typically, chopper amplifiers are better suited for dc or low-frequency applications, whereas auto-zero amplifiers are suitable for wider-band applications.

Tretter notes that auto-zero architectures used for zero-drift signal conditioning contain a main amplifier, which is always connected to the input, and secondary amps that continuously correct their own offset and apply the offset correction to the main amplifier. Microchip Technology has implemented this type of architecture on the MCP6V01, in which the offset error of the main amplifier is corrected 10,000 times/sec, resulting in what Microchip says are extremely low offset and offset drift.

A chopper-stabilized architecture also uses a high-bandwidth main amplifier that is always connected to the input, as well as an “auxiliary” amplifier that uses switches to chop the input signal and provide offset correction to the main amplifier. In Microchip’s MCP6V11 low-power amplifier, for example, chopping action minimizes offset and offset-related errors.

Although their internal operation differs, auto-zero and chopper-stabilized amplifiers share the same goal: to minimize offset and offset-related errors. This results in not only low initial offset but also low offset drift over time and temperature, superior common-mode and power-supply rejection, and elimination of 1/f (frequency-dependent) noise.

The offset voltage, offset-voltage drift, and 1/f noise of nonprecision amplifiers used for signal conditioning of these sensors cause errors that require hardware or software calibration. Moghimi offers examples of high-precision signal conditioning in which zero-drift amplifiers—designed to achieve ultralow offset voltage and drift, high open-loop gain, high power-supply rejection, high common-mode rejection, and no 1/f noise—benefit designers by eliminating the need for calibration.

The circuit in Figure 1 uses the AD7791, a low-power buffered 24-bit sigma-delta ADC, along with external ADA4528-x zero-drift amplifiers, in a single-supply precision weigh-scale application. The circuit, built and tested by ADI and described in Reference 1, yields 15.3-bit noise-free code resolution for a load cell with a full-scale output of 10 mV and maintains good performance over the full output data range, from 9.5 Hz to 120 Hz.

Figure 1 This low-power, zero-drift sensor-amplifier-converter circuit uses the AD7791 24-bit sigma-delta ADC and external ADA4528-x zero-drift amplifiers (courtesy Analog Devices).

The differential amplifier in the circuit comprises two low-noise, zero-drift ADA4528 amplifiers with 5.6 nV/ of voltage noise density at 1 kHz, 0.3-μV offset voltage; 0.002 μV/°C offset-voltage drift; and 158 dB and 150 dB of common-mode and power-supply rejection, respectively. Circuit gain is equal to 1+2R₁/R_G, and the lowpass filters implemented by placing capacitors C₁ and C₂ in parallel with resistors R₁ and R₂ limit the noise bandwidth to 4.3 Hz, restricting the amount of noise entering the sigma-delta ADC. C₅, R₃, and R₄ form a differential filter with a cutoff frequency of 8 Hz to limit noise further. C₃ and C₄ in conjunction with R₃ and R₄ form common-mode filters with a cutoff frequency of 159 Hz.

Another example of high-precision, low-power signal conditioning is the electrocardiogram circuit shown in Figure 2 and described in Reference 2. The ECG circuit must operate with a differential dc offset because of the half-cell potential of the electrodes. The tolerance for this overvoltage is typically ±300 mV, but in some situations it can be 1V or more. The downward trend of supply voltages in ECG circuits and the presence of this larger half-cell potential limit the gain that can be applied in the first stage of signal conditioning.

Figure 2 The AD8607 dual micropower instrumentation amplifier is used for integration, buffering,
and level shifting in this zero-drift signal-conditioning circuit for
an ECG application (courtesy Analog Devices).

The AD8237 architecture solves this problem by connecting a low-frequency inverting integrator from the output to the REF pin that only has to swing as far as the dc offset, instead of the dc offset multiplied by the gain. Because the amplifier applies gain to the integrator output, large gains can be applied at the amplification stage, and the precision requirements of the rest of the system can be reduced. Noise and offset error from devices after this amplification in the signal path contribute less to the overall accuracy. The AD8607 dual micropower instrumentation amp, with 115 μA of supply current, is used for integration, buffering, and level shifting. Proper decoupling is not shown.

The zero-drift, rail-to-rail input and output instrumentation amplifier can operate with a minimum supply voltage of 1.8V, gain drift of 0.5 ppm/°C, and offset drift of 0.2 μV/°C. Two external resistors set gain range from 1 to 1000. The AD8607 can fully amplify signals with common-mode voltage at or up to 300 mV beyond its supplies.