What Can You Build With the FDC2214? 12 Applications, and Where It’s the Wrong Chip

Almost everything disturbs an electric field. Water rising in a tank, a hand hovering over a panel, a beam flexing under load, frost forming on a coil — to a capacitive sensor, they’re all the same event: a small change in capacitance. That’s why one chip can turn into a dozen different instruments, depending on the electrode you hang off it.

The ProtoCentral FDC2214 Breakout puts Texas Instruments’ FDC2214 on a Qwiic board with the sensing tanks already fitted. This post walks through twelve FDC2214 applications — the well-trodden ones first, then the experimental ones — with the numbers behind each, and the honest caveats. Some of these jobs the FDC2214 does better than anything else in its price range. A couple of them it does worse than its cheaper cousin, and we’d rather tell you that now than after you’ve built the fixture.

How the FDC2214 actually measures

Most capacitance-to-digital converters, including TI’s FDC1004 (we compare the two in detail in FDC1004 vs FDC2214), measure charge. The FDC2214 doesn’t. Each channel drives an LC tank into oscillation and measures its resonant frequency against a reference clock. Your electrode’s capacitance sits in parallel with the tank capacitor, so when that capacitance changes, the frequency moves.

On our breakout, every channel has an 18 µH inductor and a 33 pF capacitor — the same values TI uses in its own characterisation — which resonate at about 6.5 MHz before you attach anything. Add 1 pF at the electrode and the frequency drops by roughly 1.5%, about 100 kHz. That’s the signal you’re working with.

This design gives you a few specific things, all straight from the datasheet:

• A 28-bit output and a 0.3 fF system noise floor at 100 SPS. Sub-femtofarad changes are visible.
• EMI resistance. TI calls it a narrow-band architecture: the chip only looks at a narrow band around the tank’s resonance, so out-of-band noise from motors, switching supplies and fluorescent lights gets rejected instead of showing up in your reading.
• A wide excitation range, 10 kHz to 10 MHz, set by the inductor and capacitor you choose. The high end is what lets it sense conductive liquids like soap, detergent and ink.
• Up to 250 nF of sensor capacitance (at 10 kHz with a 1 mH inductor), so electrodes can sit at the end of a cable rather than on the board.
• Four channels, multiplexed. The FDC2214 samples active channels one after another, not simultaneously. That matters for the rate maths below.

And one trade-off you need to know before anything else: resolution and speed share a budget. Conversion time per channel is set by a reference count, and the full 28-bit result needs the longest one — around 25 ms per channel with a ~40 MHz reference clock, and that’s what our Arduino library configures by default. Scan all four channels and each one updates at roughly 10 Hz. The headline 4.08 kSPS figure is for a single channel with a much shorter conversion and far fewer effective bits. Most of the applications below are slow physical processes, so this rarely hurts — but where it does, we’ll say so.

The breakout itself breaks out IN0–IN3, INTB and the SD (shutdown) pin, runs from 3.3 V or 5 V through an onboard regulator, and talks I²C at 0x2A or 0x2B. You bring the electrode — copper tape is a perfectly good start.

The proven applications

1. Non-contact liquid level — especially conductive liquids

Stick an electrode on the outside of a plastic or glass tank, and the capacitance climbs as liquid replaces air behind it. Nothing touches the fluid: no probe to corrode, no hole to seal.

Here’s the honest part. For plain water in a tank that people walk past and touch, the FDC1004 is often the better chip — and TI’s own datasheet says so: for non-conductive liquids with interference such as human hands, it recommends the FDC1004 because of its integrated active shield drivers. The FDC2214 has no shield drivers.

Where the FDC2214 earns its place is conductive liquids — soapy water, detergent, ink — which its high excitation frequency handles cleanly, and remote electrodes, which its 250 nF input range tolerates. TI’s reference design uses three electrodes on three channels: a level electrode running the height of the tank, a reference electrode that’s always submerged (to cancel the liquid’s own dielectric constant), and a reference electrode that’s always dry (to cancel temperature and humidity). Level falls out of a simple ratio, and the fourth channel is spare. That’s a real, robust gauge, and it’s a weekend build on this board.

2. Proximity detection

A hand approaching an electrode adds capacitance to ground, and the frequency drops. TI characterises the FDC2214 out to 60 mm with a 21 × 14 mm sensor plate — but against a large, grounded aluminium plate as the target. A hand is smaller and less well-grounded, so expect shorter range from the same electrode; a bigger electrode gets some of it back. The EMI resistance is what makes this practical next to motors and switching supplies, where a charge-based sensor tends to false-trigger.

Good for: wake-on-approach on battery devices, automatic dispensers, hands-near-the-blade interlocks on machine guards (as a supplementary signal — never your only safety layer).

3. Gesture recognition

Put electrodes on several channels and compare their readings, and you get where a hand is, not just whether it’s there. Swipe direction comes from the order in which channels respond; hover height comes from overall magnitude. Gesture recognition is one of the three applications TI lists on the front page of the datasheet, and it’s a natural fit for touchless panels where gloves or hygiene rule out buttons. Keep the four-channel scan rate (~10 Hz each at full resolution) in mind — for fast swipes, shorten the conversion time and trade away some bits.

4. Buttons and sliders behind an overlay

The FDC2214 senses through glass, acrylic, or a thin wooden veneer, so you can build buttons and sliders with nothing mechanical to wear out. Signal falls off with overlay thickness, and a 0.3 fF noise floor leaves you more of it to work with on thick overlays. But be honest with yourself: if all you need is four buttons under 2 mm of plastic, a dedicated touch controller is cheaper and simpler. Reach for the FDC2214 when the overlay is thick, the environment is electrically noisy, or you need a slider with fine position resolution.

5. Linear and rotary position

Capacitance varies smoothly with electrode overlap, so interleaved electrode patterns make a contactless linear or rotary encoder — no optical disc, nothing to clog with dust. Two channels in a differential arrangement (one pattern increasing as the other decreases) cancel a lot of common-mode drift. Real-world precision is set by your mechanics and temperature stability long before the 28 bits run out.

6. Material and moisture analysis

Different materials have different dielectric constants, and water’s (~80) dwarfs almost everything else, so the FDC2214 can tell materials apart and estimate water content in powders, paper, or plastics on a line. The catch: capacitance depends on geometry, thickness, packing density and temperature as much as on the material itself. Expect to calibrate per material and per fixture.

A note on humidity: a bare electrode in air is a weak humidity sensor. Going from bone-dry to saturated air at 25 °C moves air’s permittivity by only about 0.03% (from the water-vapour term of the ITU-R P.453 refractivity formula) — a fraction of a femtofarad on a typical electrode, buried under temperature drift. Commercial capacitive humidity sensors get their signal from a hygroscopic polymer between the plates that soaks up water. If you want humidity, either build that (it’s a fun experiment) or use a dedicated humidity sensor and save the FDC2214 for something harder.

The experimental edge

The next six are less commonly built. The physics is sound, but these reward careful electrode design, mechanical rigidity and calibration. Treat them as starting points for R&D, not recipes.

7. Bubble and slug detection in tubing

Electrodes wrapped around a plastic tube see a sharp change when an air bubble or a slug of different liquid passes. Two electrodes a known distance apart give you transit time, and therefore a rough flow rate. This works well for slow flows — drip lines, peristaltic pumps, lab fluidics. For anything fast, the rate budget bites: you’ll need to shorten conversion time and give up resolution, and TI’s FDC2114 (12-bit, up to 13.3 kSPS) is the sibling designed for fast-moving targets.

8. Micron-scale displacement

A fixed electrode facing a grounded target forms a parallel-plate capacitor, and capacitance rises steeply as the gap closes. Run the numbers for a 10 × 10 mm electrode at a 0.5 mm gap: about 1.8 pF, changing by about 3.5 fF per micron of movement. Against a 0.3 fF noise floor, sub-micron steps are within reach on paper.

On the bench, temperature is the enemy: the inductor, the tank capacitor and the fixture all drift. Use a matched reference channel, a COG/NP0 tank capacitor, and a mechanically stiff mount. Done carefully, you can watch a beam flex under load, a spindle’s runout, or a fixture’s thermal expansion — for the price of a breakout.

9. Structural movement and vibration

Electrodes bonded across a joint or a hairline crack turn sub-visible movement into a capacitance change, and continuous sampling picks up low-frequency vibration. EMI resistance is what makes this plausible on a factory floor. With four channels you can instrument several points on one structure — or three points plus a temperature reference.

10. Condensation and frost

A film of liquid condensate is easy to detect — it’s water, with a big dielectric constant. Frost is subtler: at MHz excitation frequencies, ice has a permittivity of about 3.2, not 80 — its big low-frequency response relaxes away in the kHz range — so a thin frost layer gives a much smaller signal than the same thickness of water. That makes “condensation is forming” an easy trigger and “frost has started” a nice characterisation project. Cold-storage shelves and heat-pump coils are good candidates for a defrost-cycle trigger.

11. Soil, grain and seed moisture profiles

Capacitive soil-moisture sensing is well established. Resistive probes pass current through the soil and corrode their own bare electrodes; a capacitive electrode can sit under solder mask or tape and pass no current at all. A 2020 lab study found low-cost capacitive sensors more accurate and precise than very low-cost resistive ones, once calibrated for the soil. The FDC2214’s twist is the four channels: stack electrodes at different depths in a pot, a grain bin or a seed tray and you get a moisture profile, not a single number. Calibrate per soil or grain type — composition and packing density shift the readings.

12. Breathing motion and touchless interfaces

This is where we get excited, because biosignals are what we do. The body is a large, slow-moving conductor, and breathing changes the capacitance of an electrode worn on the chest. A University of Southampton group has been building these “capaciflector” respiration sensors for years — a 2022 study measured respiratory rate within 1 breath per minute of reference devices, including during exercise testing, and a 2024 PhD thesis from the same group built a wireless version around the FDC2214. The same sensitivity makes assistive interfaces possible for people who can’t press a button — a cheek movement, a head tilt. Be clear-eyed about it: this is a platform for prototyping and research, not a medical device, and nothing you build this way should be used for diagnosis or monitoring.

Electrode tips that save a weekend

The chip is half the system. Your electrode and its surroundings decide what you actually measure.

• Single-ended or differential? TI’s guidance: a single-ended electrode (one plate on INxA) gives more range for the same plate area; a differential pair (plates on INxA and INxB) gives better sensitivity at close range. Use single-ended for proximity and level, differential for touch and fine position.
• Spend a channel on a reference. TI’s own advice for temperature drift: a second electrode that sees the environment but not the target, subtracted in firmware. It’s the cure for “my readings wander.”
• Mind the stray capacitance. Each input pin has about 4 pF of its own, and every centimetre of lead adds more — lowering the frequency and adding drift. TI rates the FDC2214 for up to 250 nF of sensor capacitance specifically so it can drive remote sensors, but shorter leads still mean less drift.
• There’s no built-in shield driver. A grounded (passive) shield behind the electrode blocks interference from that side, at the cost of some sensitivity — tune the gap. If you really need an active shield, that’s FDC1004 territory.
• Retuning? Swap the inductor (and capacitor) to move the operating frequency, and keep it under 0.8× the inductor’s self-resonant frequency, per TI.
• It reports frequency, not capacitance. Our Arduino library converts for you: readCapacitancePf(channel, 18.0) with the onboard 18 µH inductor.

Get building

Every idea above runs on the same board, over the same two wires.

• Grab the board: ProtoCentral FDC2214 Breakout — two Qwiic ports plus 0.1″ headers, ₹2,995, with volume pricing from 10 units.
• Arduino: install ProtoCentral_FDC2214_Capacitance from the Library Manager, or browse the library on GitHub (MIT). Start with 03_MultiChannelAutoScan for all four channels. 05_ProximitySensing is a baseline-and-threshold detector for application 2, and 06_LiquidLevel is a two-electrode version of application 1 — a level strip plus a reference above the fill line — that you can extend to TI’s three-electrode method.
• MicroPython: use the FDC2214 MicroPython driver.
• Fork the hardware: the full KiCad design is in the FDC2214 hardware repository under CERN-OHL-P v2. Change the tanks, change the layout, build it into your product — that’s the whole point.
• Not sure it’s the right chip? Read FDC1004 vs FDC2214 first, and Capacitive Sensing Applications for the wider picture.

A strip of copper tape, a Qwiic cable and one example sketch get you a live reading in minutes. Pick one of the twelve — or find a thirteenth — and show us what you build on Hackster.io.