Pressure Sensor Interface Circuit Design

A pressure sensor turns a mechanical deflection into an electrical signal, and the circuit that follows has to remove a set of errors that are larger than the signal itself. Offset, span, temperature drift and nonlinearity are all specified separately on the datasheet, and a channel that ignores any of them will be limited by the one that was overlooked.

What a Pressure Sensor Outputs

The most common sensing element is a silicon diaphragm with four piezoresistors diffused into its surface and connected as a bridge. Pressure deflects the diaphragm, the resistors change by a small amount, and the bridge produces a differential output proportional to the pressure. The signal is small, typically tens of millivolts at full scale, and it sits on a common mode level near half the excitation.

Some parts add a signal conditioning circuit inside the package and provide a ratiometric analogue output, a digital interface or a current loop output. These are easier to use but their specifications are the same set of errors, expressed at the output rather than at the bridge. Reading the error budget the same way is still the right approach.

The choice between a raw bridge and a conditioned part depends on the volume and on the required accuracy. A raw bridge with an external amplifier and converter gives the best performance for a given cost when the board is already designed for it, while a conditioned sensor shortens development and moves the calibration burden to the manufacturer.

Bridge Offset and Span Error

bridge offset is the output the sensor produces at zero pressure. It is produced by mismatch in the bridge resistors and by mechanical stress in the die attach, and it can be several percent of the full scale output. Offset is easy to remove by calibration, provided the front end has enough range to accommodate it without clipping.

span error is the difference between the actual sensitivity and the nominal value, and it is removed by a gain correction. Together, offset and span define a straight line through the two calibration points, and calibrating at the extremes of the range is what makes the correction valid across the whole span rather than only near the calibration point.

Nonlinearity is the error that calibration cannot remove. The diaphragm response is slightly nonlinear, and the bridge output also contains a second order term that grows with pressure. A part specified at 0.1 percent of full scale nonlinearity cannot be made more accurate by better calibration, so the specification has to be checked against the requirement before the part is chosen.

Temperature Compensation

The bridge sensitivity changes with temperature, typically by a fraction of a percent per degree, and the offset moves as well. temperature compensation is therefore part of the channel rather than an optional refinement, and it is normally implemented by measuring the sensor temperature and applying a correction that depends on both temperature and pressure.

Where the sensor package includes an internal temperature sensor, the correction can be derived from that reading. Where it does not, a separate sensor mounted close to the package gives a usable approximation, provided the two are thermally coupled. A thermistor on the far side of the board measures the board temperature, not the sensor temperature, and the difference can be several degrees.

A first order correction removes most of the error. The offset is corrected with a linear term in temperature and the sensitivity with a second linear term, which brings a typical part from several percent of full scale over the temperature range down to a fraction of a percent. Higher order terms are added only when the residual error demands it.

Excitation and Ratiometric Conversion

The bridge output is proportional to the excitation, so a ratiometric arrangement removes the excitation from the transfer function. Driving the converter reference from the same source as the bridge makes the reading independent of the exact excitation voltage, which is particularly useful when the system runs from a battery whose voltage changes with charge.

Excitation current should be kept modest to limit self heating in the bridge. A few milliamperes is typical, and the resulting temperature rise is small compared with the errors already discussed. Where the bridge has to be powered continuously, the excitation can be pulsed and the measurement taken before the bridge warms, which reduces the self heating to almost nothing.

The common mode voltage at the amplifier input is close to half the excitation, so the amplifier input range has to include that level with margin for the offset. A part with an input range that just covers the nominal common mode will clip when the offset pushes the signal to one side, and the clipping appears as a nonlinearity at low pressure that is easy to mistake for a sensor fault.

Media, Sealing and Overpressure

How the pressure reaches the diaphragm decides what the sensor can be used for. A gauge sensor references one side to atmosphere and needs a vent path that will not be blocked by potting compound or conformal coating, while a sealed gauge or absolute part references a vacuum and has no vent. Blocking or opening the wrong path gives a reading that is plausible and completely wrong.

The overpressure rating is the pressure the sensor survives without a permanent change in calibration, and it is normally several times the rated range. A system that can be over pressured in normal operation, for example by a blocked line or a water hammer, should be specified with that event in mind rather than only for the normal working range. The burst pressure is a different and much higher number.

Media compatibility concerns the wetted materials rather than the electronics, but it affects the board as well. A sensor with a gel filled cavity tolerates humidity but not aggressive chemicals, while a part with a stainless steel diaphragm and an oil filled cavity handles a wider range of fluids. The port, the seal and the thread all have to be compatible with the fluid as well as the diaphragm.

Layout of the Sensor Front End

The sensor front end is an analogue circuit and should be laid out as one. Keep the bridge, the amplifier and the converter in a compact area over a solid ground plane, route the excitation pair together, and keep digital traces away from the bridge connections. This is the same partition used in any mixed signal design, as described in mixed signal board design.

Length matching of the bridge connections is worth the effort. A difference in resistance between the two sides of the bridge appears as an offset that changes with temperature, because copper has a large temperature coefficient compared with the bridge resistors. Keeping the traces equal, short and away from heat sources makes that error small and stable.

Where the sensor is a separate assembly connected by a cable, treat the excitation and sense conductors as a group and provide a separate return. If the excitation current returns through the analogue ground on a path the amplifier does not follow, the resulting voltage appears as a span error that varies with temperature, and the symptom looks exactly like a device specification problem.

Calibration and Verification

Calibrate at the two ends of the range with a reference gauge that is at least four times more accurate than the requirement. Record the raw converter counts and the reference reading at each point, and compute the offset and gain coefficients from those pairs. Calibrating at the middle of the range as well gives a quick check on nonlinearity without additional equipment.

Verify the channel with a pressure that was not used in the calibration, then repeat the measurement at the temperature extremes. A channel that is linear but drifts with temperature has a compensation problem, while one that is stable but inaccurate at intermediate pressures has a nonlinearity problem. Distinguishing the two before changing components saves a great deal of time.

For boards that will be exposed to moisture, conformal coating protects the high impedance nodes from leakage that would otherwise appear as drift. The coating must not block the vent path or cover the sensor port, and the assembly drawing should show the keep out area clearly. Standard fabrication notes for that kind of feature are collected in our guide to manufacturable PCB design.

A final production check is to record the offset and the span coefficient for every unit. Comparing those values across a batch shows whether the front end is consistent and whether the sensor supplier is holding its specification. A drifting batch distribution is visible long before the first field failure, and it costs nothing more than a line in the test record.

FAQ

How accurate can a low cost pressure sensor be? With two point calibration and a first order temperature correction, a bridge sensor specified at one percent of full scale can typically be brought to a fraction of a percent over a moderate temperature range.

Do I need an instrumentation amplifier? For a bridge with a few tens of millivolts of output, yes, unless the converter has a differential input with enough gain and common mode rejection of its own.

Why does my reading change when I touch the board? Either the sensor is being heated by the finger or the high impedance nodes are picking up leakage current. Mount the sensor away from heat sources and consider coating the front end.