Pressure sensors are one of those components that look dead simple on paper and then eat an entire week of your life when the reading drifts, the diaphragm cracks, or the thing reads 3 psi high every time the ambient temperature climbs. The datasheet says ±0.5% accuracy. The datasheet is a sales document. This is the stuff that usually only gets passed around verbally between maintenance techs and controls engineers.
Here’s the honest version of how industrial pressure sensing actually works, what the different technologies are good for, which specs are real, and which ones exist purely to make the spec sheet look better than the sensor.
First, Kill the Terminology Confusion
Half of all “bad sensor” complaints are actually “wrong sensor reference type” complaints. Get this right before anything else.
- Gauge pressure — measured against local atmospheric pressure. The vented-to-atmosphere sensor. Most common in industrial process work.
- Absolute pressure — measured against a sealed vacuum reference. Used when atmospheric changes matter, like barometric or vacuum system work.
- Sealed gauge — measured against a fixed internal reference pressure, not the current atmosphere. Common in high-pressure hydraulics where a vented sensor would be impractical.
- Differential pressure — the difference between two ports. Used for flow, filter clogging, and level measurement in closed tanks.
- Compound / vacuum — spans negative and positive gauge pressure. Used on vacuum pumps and anything that swings both ways.
Order a gauge sensor when you needed absolute and you’ll spend a month chasing a “drift” that’s actually just weather.
The Sensor Technologies You’ll Actually Meet
Piezoresistive strain gauge
The default. A silicon or ceramic diaphragm with strain gauges bonded or diffused onto it. Cheap, accurate, works over a wide temperature range, and available in basically every pressure range you’d want. If you don’t have a specific reason to pick something else, this is what you’re buying. Downside: the diaphragm is a thin physical membrane, so harsh media and water hammer are real threats.
Capacitive (including MEMS versions)
A diaphragm moves and changes the capacitance between it and a fixed plate. Excellent low-pressure sensitivity and great for differential measurements and low-range work where strain gauges get noisy. Very stable, but generally more sensitive to vibration and mounting stress than a good bonded strain gauge. The MEMS variants are the tiny silicon ones that show up in everything from medical gear to HVAC.
Piezoelectric
Only measures changing pressure. Great for fast dynamic events — combustion, shock tubes, blast waves. Useless for steady-state readings because the charge leaks away. People buy these expecting a normal pressure reading and then wonder why the value decays to zero. It’s not broken. It’s physics.
Thin-film / sputtered strain gauge
Metal strain gauge elements sputtered directly onto a stainless diaphragm. Tougher than bonded foil, better for high pressure, high vibration, and nasty environments. Slightly less sensitive at low ranges but it survives things that kill everything else.
Variable inductance / LVDT-based
A core moves inside a coil. Very robust mechanically, immune to a lot of abuse, but bulky and slow. Mostly legacy and heavy-industry territory now.
Optical / fiber-based
Pressure changes how light behaves in a fiber. Completely immune to electromagnetic interference and safe in explosive environments where any electrical signal is a problem. Expensive, niche, and used when nothing else is allowed.
The Spec Sheet Is a Marketing Document
This is the part nobody explains clearly. Sensor accuracy numbers are usually quoted in the most flattering way legally possible.
Accuracy is a trap
“±0.5% accuracy” almost always means best straight line (BFSL) or terminal-point accuracy at a single reference temperature — usually 25°C. What you actually care about is total error band, which bundles non-linearity, hysteresis, non-repeatability, and thermal effects across the whole compensated temperature range. A sensor advertising ±0.25% can easily have a total error band four times that. Ask for the total error band or you’re buying a number, not a performance.
Thermal effects quietly destroy you
Two numbers matter: thermal zero shift (how much the zero point moves per degree) and thermal span shift (how much the full-scale output changes per degree). A sensor with great accuracy but sloppy thermal specs will read fine in the shop and wrong on the roof in July. Also check the compensated temperature range versus the operating temperature range. A sensor rated to 120°C operating but compensated to 70°C will work at 100°C — it just won’t be accurate there.
Long-term stability
Usually quoted as drift per year. This is the number that decides whether you’re recalibrating annually or replacing the thing. Cheap sensors drift. You pay for stability whether you realize it or not.
Overpressure, proof, and burst
Three different numbers people treat as one. Proof pressure is what it survives with no permanent change. Burst pressure is when it becomes shrapnel. Overpressure rating is what it handles without damage but possibly with a spec violation. A sensor with a 100 psi range and a 150 psi proof pressure will die the first time someone closes a valve too fast.
The Specs Nobody Reads Until It’s Too Late
- Wetted materials — diaphragm, body, and seal material versus your actual media. A standard stainless diaphragm does not enjoy certain acids, ammonia, or chlorine. This is the #1 cause of mysterious failures.
- Process connection — thread type, seal style, and whether it’s compatible with the port you’re screwing into. Adapters are where leaks and corrosion start.
- Output signal — 4–20 mA current loop is the industrial default because it’s noise-resistant and self-diagnosing on wire breaks. Voltage outputs are cheaper but degrade over long cable runs. Digital bus outputs give you diagnostics and remote zeroing, but they lock you into a protocol.
- Ingress protection and area classification — washdown, outdoor, and hazardous-area ratings aren’t optional. Getting this wrong is a safety and insurance problem, not a convenience problem.
- Response time — a fast-scanning controller on a slow sensor will happily control a value that moved three seconds ago.
How to Actually Select One
- Write down the real pressure range — minimum, maximum, and worst-case transient. Not the nominal operating point.
- Pick the reference type — gauge, absolute, sealed, or differential.
- Pick the technology — piezoresistive unless you have a reason, thin-film for abuse, capacitive for low range, piezoelectric only for dynamic.
- Add a safety margin — a sensor running at 90% of its range for years is a sensor you’ll be replacing. Aim for the operating point near 50–70% of span.
- Check the total error band at your actual temperature, not the showroom number.
- Match the wetted materials and the connection to the real process, not the ideal one.
- Pick the output that matches your system, and remember that analog loops need a compatible supply and load resistance.
- Confirm the environmental rating for the actual install location — sun, rain, washdown, vibration, solvent.
The Field Tricks That Keep Things Running
- Snubbers and impulse lines save sensors from water hammer and pulsation. Half the “failed” sensors on pumps died from pressure spikes, not wear.
- Mount with a siphon or pigtail on hot lines so steam doesn’t cook the electronics.
- Zero-trim in place when allowed. Many transmitters have a zero adjustment. If your offset is stable and small, trimming is faster than pulling the unit. Just make sure the process is actually at zero pressure when you do it.
- Keep the vent port clear. A plugged atmospheric vent on a gauge sensor makes it behave like a bad absolute sensor.
- Over-specifying is a real cost. Buying a high-accuracy, high-stability unit for an alarm-only application is money burned. Under-specifying buys downtime. Match the sensor to the job, not to your anxiety.
The Bottom Line
Industrial pressure sensors aren’t complicated, they’re just badly explained. Most selection disasters trace back to four things: wrong reference type, a datasheet accuracy number that excludes thermal error, mismatched wetted materials, and a pressure range chosen from the nominal operating point instead of the worst case.
Nail those four and you’ll out-select most people who’ve been specifying sensors for years. The spec sheet isn’t lying to you exactly — it’s just answering a question you didn’t ask.