Piezoelectric Sensor Vibration Interface

A piezoelectric accelerometer produces a charge proportional to acceleration, and the interface circuit decides what part of that charge survives the trip from the sensor to the electronics. Cable capacitance, insulation resistance and input noise are the three numbers that set the usable range of the measurement.

What the Sensor Produces

A piezoelectric element generates a charge when it is stressed, and in an accelerometer the stress comes from a seismic mass pressing on the crystal. The charge is proportional to the acceleration, and for a typical sensor it is a few picocoulombs per unit of acceleration.

The charge leaks away through the insulation resistance of the crystal, the connector and the cable, and that leakage sets the low frequency limit of the measurement. A sensor with a time constant of a second cannot measure a slow vibration, while one built for a low frequency has a much higher insulation resistance.

The high frequency limit is set by the mechanical resonance of the sensor, which is usually in the tens of kilohertz. A measurement well above the resonance reports the resonance rather than the machine, so the useful band always sits below it.

Charge Amplifier Versus Voltage Mode

A charge amplifier holds the input at virtual ground so the cable capacitance carries almost no signal, and the charge is transferred to a feedback capacitor. The output is then independent of the cable, which is the reason this circuit is used with long cables.

A voltage mode connection with a high impedance buffer at the sensor, often built into the sensor as an IEPE type, converts the charge to a voltage at the source. The cable then carries a low impedance signal, and cable capacitance only affects the high frequency limit.

The choice follows from the environment. A charge amplifier at the instrument end suits a sensor in a hot or inaccessible place, while an IEPE sensor with a constant current supply is simpler and is the default for permanent monitoring.

Cable Capacitance and Cable Noise

cable capacitance is the reason a voltage mode connection with a passive sensor is impractical. A hundred picofarads per metre over twenty metres is two nanofarads, and against a sensor capacitance of a few hundred picofarads it forms a divider that attenuates most of the signal.

cable noise is the other classic problem. A piezoelectric sensor and its cable generate a charge when they are bent or moved, because the insulation and the connector materials are themselves piezoelectric to a small degree. Tying the cable down firmly and avoiding a loop that can flap removes most of it.

The connector is a frequent failure point. A coaxial connector with a poor shield connection couples the local electric field into the high impedance input, and the resulting hum is easily mistaken for a sensor fault.

Designing the charge amplifier

The feedback capacitor sets the sensitivity of the channel: a hundred picocoulombs of charge through a thousand picofarad capacitor gives a tenth of a volt, so the gain is set by choosing that value. The feedback resistor defines the low frequency corner together with the capacitor.

The operational amplifier has to have a very low bias current, because that current flows into the feedback network and appears as a drift. A part with a bias current of a few femtoamps is the right choice, and a general purpose amplifier is not.

Guarding is standard practice at the input. The guard ring is driven at the same potential as the input, and it intercepts leakage currents from the board surface so that they do not reach the summing node. Contamination and humidity then matter much less.

Low Frequency and High Frequency Limits

The low corner is set by the feedback time constant, and it also depends on the insulation resistance of the sensor, the cable and the connector. A dirty connector can reduce the resistance enough to raise the low corner by a factor of ten, which shows up as a loss of the slow components of a vibration.

The high corner is set by the feedback capacitor together with the feedback resistor and the open loop bandwidth of the amplifier. Where the required band is wide, the gain is split between two stages so that the second stage can define the high corner.

The usable dynamic range is then the ratio of the largest signal the output can produce to the noise floor, and in a vibration measurement that ratio is often sixty to eighty decibels. The noise budget has to be checked at the input, because every later stage amplifies the input noise with the signal.

IEPE Supply and Signal Extraction

An IEPE sensor contains its own amplifier and draws a constant current of a few milliamps from the instrument. The bias voltage of a few volts appears on the same conductor as the signal, and a capacitor at the instrument separates the two.

The constant current source has to be quiet, because its noise appears directly at the sensor. A resistor from a clean supply is often enough, and a current source built from a reference and an amplifier is used where the supply is shared with digital circuits.

The coupling capacitor sets the low corner of the channel together with the input resistance of the following stage. Its own leakage must be low, because a leakage current through it develops a voltage across the input resistance that appears as a slow drift at the output.

Layout, Shielding and Grounding

Keep the input node small and guarded, and place the amplifier within a few millimetres of the connector. Every extra millimetre of trace adds capacitance and picks up more of the local field, and at the input the signal is a charge rather than a voltage.

Earth the cable screen at the instrument end and insulate it from the machine frame unless the sensor is designed for a grounded installation. A ground loop through the screen carries the machine current, and that current appears as a voltage in series with the signal. The general approach is described in our guide to mixed signal board design.

Separate the analogue ground from the digital ground and join them at one point under the converter. A vibration channel is often on the same board as a processor and a communication interface, and the return currents of those circuits are large compared with the charge signal.

Calibration and Verification

Verify the channel with a calibrated shaker or a back to back reference accelerometer. A comparison over the band shows both the sensitivity error and the shape of the frequency response, and the shape is what reveals a low corner that is too high.

Check the noise floor with the sensor connected and the machine stopped. A noise floor that rises when the sensor cable is moved indicates a triboelectric effect rather than an electronic fault, and the fix is mechanical.

Record the sensitivity and the serial number of each sensor with the channel it is connected to. Sensors are calibrated individually, and swapping two without updating the record introduces an error of several percent that is otherwise invisible. The release and inspection practices that keep these channels consistent are collected in our PCB design release checklist and judging PCB quality.

FAQ

Why does my vibration reading drift slowly? Insulation resistance at the input is the usual cause. Clean the connector, check the cable and guard the input node on the board.

Can I use a long cable with a piezoelectric sensor? Yes with a charge amplifier at the instrument end, or with an IEPE sensor whose own amplifier drives the cable.

Why does the reading change when the cable moves? That is the triboelectric effect in the cable. Clamp the cable so it cannot flex, and replace a cable with a stiff or damaged screen.