Troubleshooting
Why a Capacitive Touch Sensor Trips From Centimetres Away
Short answer
The room is the defect: the frequency converter output run at the 8 ft ceiling turns the operator into a receiving antenna, and the 60 Hz fundamental plus kHz-range common-mode switching noise capacitively injects into the touch electrode until it trips from 3-4 cm of air. Fix the coupling path — bonded conduit, a grounded metal canopy over the bench, or a reroute — and leave the firmware alone.
On this page
An overhead converter output run at an 8 ft ceiling is the source, the operator standing under it is the receiving antenna, and the touch electrode is the victim. The machine converts 240 VAC 50 Hz mains to 125 VAC 60 Hz, and that output run carries the mains fundamental plus kHz-range common-mode switching noise. Standing roughly 2 ft below it, the operator's body couples that field into the sensor through centimetres of air, and the trip distance grows from finger contact to 3-4 cm. The unit is fine: the room is the defect. Fix the coupling path — bonded conduit on the run, a grounded metal canopy over the bench, or a reroute — and leave the firmware alone.
Voltage coupling, not current coupling #
Self-capacitance sensing charges the electrode against earth and resolves sub-picofarad changes; a finger registers because the human body is a large, earth-referenced plate. That same body is also an antenna. Under an energized overhead cable it sits at a raised AC potential relative to earth, and that potential couples capacitively through the plastic handle from centimetres away — charge injection the controller cannot distinguish from an approaching finger.
This is electric-field coupling driven by cable voltage. Magnetic coupling is driven by cable current, enters current loops, and answers to twisted pairs and loop-area control — the wrong toolbox here. Voltage-driven coupling answers to shielding and separation.
Two properties make the converter run worse than ordinary building wiring. Coupling between conductors rises with frequency, so the kHz switching component couples harder than the mains fundamental. And converter data sheets specify output noise while saying almost nothing about common-mode noise, so the kHz term riding the output is invisible on paper. The same common-mode energy also corrupts data links; for that failure mode, read Diagnosing intermittent HMI to PLC Ethernet comms hangs.
Confirm the room before touching the unit #
- Carry the failing unit into a high-ceiling bay with no overhead converter run; the trip distance should return to finger contact.
- Carry a known-good unit from that bay into the low-ceiling room with the overhead run; it should fail identically.
Both checks hold, and the fault follows the room rather than the serial number. Log it that way so nobody burns spares chasing a building defect.
Measure the field with the scope and your finger #
- Leave the probe's ground lead unconnected — you are measuring a coupled voltage, not a circuit node.
- Pinch the probe tip between your fingers and stand exactly where the operator stands when pressing the handle.
- Read the amplitude and the frequency off the screen.
In any powered building you will see something; that is normal and low. Under the overhead converter run the amplitude climbs hard, and two components show: the mains fundamental of the converter output, and a kHz-range component from the switching stage riding the same cable as common-mode. The kHz term matters most, because touch controllers sample the electrode at rates that alias kHz-band interference directly into the touch delta.
Make the survey repeatable #
A finger is not a calibrated transducer. Clamp the probe vertical, tip up, where the handle sits during test, and fix about 30 cm of small-gauge wire pointing straight up from the tip. Trim the wire in steps until the fixture reads what your finger read in the same spot, then freeze the geometry — height, wire length, probe and bench position all change the number. Two readings taken minutes apart should agree before you trust it.
Walk the fixture through every bay with the converter running and log the amplitude, the ceiling height, whether the run passes overhead, and the lateral distance from the bench. An 8 ft room that reads low is telling you what is different about it — conduit instead of open cable, a bonded metal ceiling grid — and that room is your free reference design. Cross-plot the fixture amplitude against the measured trip distance; the correlation is the evidence you take to the factory. Read each bay once more with the product unplugged: a large change when it is plugged in means part of the coupling arrives on the product's own supply cord as common-mode.
Fixes that look right but fail #
- Retune the touch threshold. The unit then under-triggers in every clean room, and the coupled noise still walks the baseline, so margin is unstable in both directions. You ship hardware tuned to one contaminated bay.
- Swap the unit or the touch controller. The coupled field is common to every unit in that room; replacements fail identically and you lose inventory and a shift.
- Add filter capacitance at the sense electrode. Loading the electrode cuts the touch delta along with the noise, and can push a charge-transfer front end out of its usable range.
- Switch the converter off. Valid as a diagnostic, useless as a fix — its 125 VAC output is what powers the device under test.
Coupled current under the converter run at an 8 ft ceiling #
Inputs:
- Converter output: 125 VAC at the mains fundamental, from the machine converting 240 VAC 50 Hz mains.
- Switching component: kHz range, riding the output as common-mode; the arithmetic uses 3,000 Hz as the low end of that band, and scales linearly for any other frequency.
- Coupling capacitance: 1 pF (1×10⁻¹² F), the same order the controller resolves, since the touch delta itself is sub-picofarad.
- Geometry: run at the 8 ft ceiling, about 2 ft above the operator's head, hand at the bench.
Rule in words: the displacement current a capacitance passes is proportional to frequency, capacitance and voltage together — I = 2πfCV.
Arithmetic, at the fundamental:
I = 2π × 60 Hz × 1×10⁻¹² F × 125 V = 4.7×10⁻⁸ A, about 47 nA
Arithmetic, at the low end of the switching band:
I = 2π × 3,000 Hz × 1×10⁻¹² F × 125 V = 2.4×10⁻⁶ A, about 2.4 µA
Ratio: 3,000 Hz ÷ 60 Hz = 50.
Result: for the same volts and the same pF, the kHz band injects fifty times the current of the mains term, and every further doubling of frequency doubles it again.
Decision: current falls in direct proportion to capacitance, so the one lever that cuts both terms at once is C — grounded metal between run and bench, bonded conduit around the run, or separation. No threshold setting can tell a fifty-times-larger injection from a finger. Pass line: with the converter running, trip distance back to finger contact; any activation from air, 1 cm of gap or 3, means the coupling path is still open and the bay keeps its fix.
The fixes that work, in order #
- Reroute the run so it no longer passes over the bench. A cable kept high and clear of the operator stops being an antenna suspended over the victim.
- Put the run in metal conduit, bonded to the building's equipment grounding system. The bond is what makes the enclosure a shield; without it you have added weight, not shielding.
- Hang a grounded sheet-metal canopy over the bench when the run can neither move nor be enclosed.
Then have a qualified electrician check the converter output itself: a frequency converter output is frequently a floating secondary with no neutral-to-earth bond on the 125 VAC side, and bonding a separately derived system is code work for an electrician following the converter documentation, not a jumper a technician adds. Verify every fix against the frozen fixture, not by feel.
Questions asked
Will lowering the touch sensitivity threshold fix it?
No. Retuning the threshold so the sensor behaves in the contaminated room makes it under-trigger in every clean room, and the coupled noise keeps walking the baseline, so the margin is unstable in both directions. A unit tuned down in the noisy bay can even pass there and fail in the field. Fix the coupling path first, and only retune a unit that still misbehaves in a room the fixture survey has cleared.
Why does the kHz switching noise matter more than the 60 Hz?
Displacement current through a capacitance scales with frequency, capacitance and voltage together. At 3,000 Hz the same 1 pF and 125 V couples fifty times the current of the mains fundamental, and touch controllers sample the electrode at rates that alias kHz-band interference straight into the touch delta, so the switching term dominates the baseline shift.
Is this magnetic coupling from the cable current?
No, it is electric-field coupling from the cable voltage. Magnetic coupling comes from current, enters current loops, and is treated with twisted pairs and loop-area control. Here the effective measures are shielding and separation against voltage, so treating it as a magnetic problem sends you down the wrong path entirely.
How do I check a room quickly without special instruments?
Pinch an oscilloscope probe tip between your fingers with the ground lead connected to nothing, and stand exactly where the operator stands. Every powered building shows a low 60 Hz level; a much larger reading with kHz content riding on it marks the overhead run as the source. Confirm by repeating the same check in a room where the sensor behaves normally.
Does putting the cable in conduit actually work?
Yes, when the conduit is metal and properly bonded; plastic conduit does nothing, and unbonded metal does little, because the bond is what stops the run acting like a free-space antenna. Moving the run so it no longer passes over the bench, or hanging a grounded sheet-metal canopy above the bench, achieves the same cut in coupled capacitance.