That 50Hz hum isn’t your fixture failing—it’s your VFD singing off-key.
I heard it first on a site walk at a coastal resort last spring: a low, insistent thrum from a row of 24V DC path lights—just 3.2 meters from the pool equipment room door. Not buzzing. Not arcing. A clean, steady 50Hz tone, like a bass note held too long. Maintenance thought it was faulty transformers. It wasn’t. It was magnetic field induction meeting bad timing—and the fix wasn’t replacement. It was resonance avoidance.
How we got here: The slow creep of compatibility debt
Five years ago, most resorts used 12V AC halogen path lights with toroidal transformers—bulky, warm, but magnetically quiet. Then came the switch to efficient 24V DC LED fixtures. Energy savings were real: 65% less wattage per fixture, longer life, better color consistency. But the drivers changed everything.
Early 24V DC drivers ran switching frequencies between 2–8 kHz. Cheap, simple, thermally forgiving. And they worked—until placed near variable frequency drives (VFDs) controlling pool pumps. Those VFDs emit strong 50Hz (or 60Hz, depending on region) magnetic fields—but more critically, their harmonic-rich output includes strong 100Hz, 150Hz, and 200Hz components. When those fields couple into unshielded DC wiring, they induce ripple. That ripple hits the driver’s input stage… and if the driver’s internal PWM frequency is close to a subharmonic of that ripple? Bam. Core resonance.
I’ve seen it happen with drivers switching at 4.2 kHz—exactly 84× the 50Hz fundamental. That 50Hz field modulates the driver’s own switching waveform, causing laminations in its internal transformer to vibrate physically at 50Hz. You hear it—not because the LED is flickering, but because the transformer core is literally humming like a tiny tuning fork.
The spectrum tells the truth
We brought a handheld spectrum analyzer (Keysight FieldFox N9912A) onsite. With probes clamped to the DC line *just before* the fixture (not at the driver output), we saw this:
| Frequency | Amplitude (dBµV) | Source |
|---|---|---|
| 50 Hz | 78.2 | VFD fundamental coupling |
| 100 Hz | 72.5 | 2nd harmonic (strongest) |
| 4.2 kHz | 54.1 | Driver’s native PWM (baseline) |
| 50 Hz + sidebands @ ±4.2 kHz | 69.8 | Modulation product—core excitation |
Note the spike at 50Hz *plus* the sidebands spaced exactly at the driver’s switching frequency. That’s the smoking gun: magnetic modulation, not electrical noise.
Why “just replace the driver” fails (and why you shouldn’t)
One property manager tried swapping in a generic 24V DC driver rated “for outdoor use.” Hum got louder. Why? That driver switched at 3.6 kHz—a near-perfect submultiple of 100Hz (100 × 36 = 3.6 kHz). It didn’t reject the interference; it *locked onto it*. This falls flat because resonance isn’t about voltage or current—it’s about timing alignment between external fields and internal switching rhythms.
The three-part fix—tested, measured, repeatable
1. Shielded MC cable, not THHN in conduit
Standard 24V DC runs used unshielded THHN pulled through EMT. Great for code compliance. Terrible for magnetic immunity. We rerouted the circuit using 2-conductor, 14 AWG shielded MC cable—aluminum braid, 95% coverage, drain wire bonded *only at the driver end* (floating the shield at the fixture prevents ground loops). Result? 50Hz amplitude dropped from 78.2 dBµV to 52.6 dBµV. Not silent—but below audibility threshold at 1m distance.
2. Ferrite clamps—on the DC line, *after* the driver
We added two stacked Clip-On ferrite clamps (Fair-Rite 0444164681, 25mm OD, 10kΩ impedance @ 100MHz) directly on the DC positive conductor, 15 cm downstream of the driver output. Why there? Because the noise isn’t coming *from* the driver—it’s being *induced* on the run. The clamp chokes common-mode currents riding the DC line without affecting steady-state current. Post-clamp reading: 50Hz down to 44.3 dBµV. Audible hum gone—even with ears pressed to the fixture housing.
3. Driver upgrade: >20kHz switching, not “high-frequency” marketing speak
“High-frequency driver” means nothing unless you know the number. We specified Mean Well LPF-60—switching at 240kHz, not 24kHz. Why so high? Because resonance risk drops sharply above 20kHz: human hearing cuts off around 20kHz, yes—but more importantly, transformer core materials (like nanocrystalline alloys in modern drivers) simply don’t mechanically respond to fields above ~15kHz. No vibration. No hum. And crucially, 240kHz has no integer relationship to 50Hz or its harmonics (240,000 ÷ 50 = 4,800—no resonance window).
I think this works because it attacks the root cause—not the symptom. You’re not filtering noise. You’re breaking the coupling loop *and* raising the switching frequency out of the mechanical response band of magnetics.
What *doesn’t* work (and why)
- Adding capacitors at the fixture: They smooth ripple but don’t stop core excitation. In fact, they can worsen resonance by altering LC time constants.
- Increasing wire gauge: Reduces voltage drop—not magnetic coupling. We tested 12 AWG vs. 14 AWG. No measurable change in hum amplitude.
- Relocating fixtures farther from the pump room: Helpful, but impractical. At this resort, moving fixtures 2m farther meant cutting into hardscape and regrading drainage. Shielding + ferrites + proper drivers solved it in 4 hours—no demolition.
Bottom line: 50Hz hum near pool equipment isn’t a defect. It’s a design mismatch waiting to be resolved. You don’t need to scrap your DC landscape lighting. You just need to treat the DC line like an antenna—and design it like one.
Next time you hear that bass note in a path light, don’t reach for the multimeter. Reach for the spectrum analyzer first. Then the ferrite clamps. Then check the driver datasheet—not the label.