Low-Voltage Garden Lights Causing WiFi Interference? Diagnosing RF Noise from Cheap Transformers
Last fall, I helped a client in Portland troubleshoot why her new mesh WiFi system kept dropping in the backyard—right where she’d just installed a $35 “premium” 12V landscape lighting kit. Her router was indoors, three rooms away. The mesh node sat on the patio table, 8 feet from a buried low-voltage cable running to path lights. Speed tests showed 400 Mbps indoors—but under 12 Mbps when streaming video on the back deck. She’d already swapped nodes, updated firmware, and even repositioned the outdoor unit twice. Then I clipped my oscilloscope probe across the secondary winding of her transformer—and saw it: a 27 kHz square wave riding a 120 kHz harmonic cascade, spiking into the 2.4 GHz band like a broken flute.
That’s not normal. And it’s not rare.
The Popular Take Is Wrong: “It’s Just the Cables”
Most DIY forums blame the low-voltage cables themselves—suggesting twisted pair or burial depth fixes. Some insist “WiFi and landscape lighting don’t share frequencies, so interference is impossible.” Others shrug and recommend moving the mesh node farther away—or switching to 5 GHz (which, of course, won’t help if your smart thermostat, doorbell, or baby monitor is still on 2.4 GHz).
This falls flat because it misidentifies the source. The copper wire isn’t radiating. The LED bulbs aren’t oscillating at 2.4 GHz. What’s radiating is the transformer—not as a clean 60 Hz sine wave, but as a broadband RF noise generator. And cheap transformers (<$40) are often the worst offenders.
Why Budget Transformers Radiate Like Tiny Broadcast Towers
Here’s what happens inside that black plastic box you bought at the big-box store:
- It uses a non-isolated, unshielded flyback converter—not a toroidal transformer with electrostatic shielding.
- It chops 120 VAC at ~25–50 kHz using a low-cost MOSFET and minimal gate-drive filtering.
- There’s no common-mode choke. No Y-capacitor filtering to earth. No ferrite on the output leads.
- The output rectification is often done with fast-recovery diodes without snubbers—creating sharp current edges.
Each voltage transition generates dI/dt spikes. Those spikes couple capacitively into the output wiring, then radiate from the entire low-voltage loop (often 50–100 ft of exposed cable) like an unintentional dipole antenna. I’ve measured field strength up to 42 dBµV/m at 2.412 GHz—well above FCC Part 15 Class B limits for residential devices—just 3 ft from a $29 transformer buried in mulch.
This works because harmonics don’t stop at the third or fifth. A 33 kHz fundamental (common in these designs) produces odd harmonics at 99 kHz, 165 kHz, 231 kHz… all the way past 2.5 GHz. The 67th harmonic of 33 kHz is 2.211 GHz. The 73rd is 2.409 GHz—the center of WiFi Channel 1.
Oscilloscope Readings: What to Look For (and What It Means)
You don’t need a spectrum analyzer to suspect EMI—but a basic two-channel scope helps confirm it. Here’s what I check on-site:
- Secondary output waveform: Probe across the + and – terminals with 10x passive probe, bandwidth limit on. Look for ringing >100 ns, overshoot >15% of nominal voltage, or high-frequency carrier riding the DC baseline (e.g., 12 VDC with 1.2 Vpp @ 120 kHz noise superimposed).
- Common-mode noise: Connect both probe tips to + and –, then set scope to subtract mode (A–B). Now connect ground clip to earth ground (a cold water pipe works), and touch one probe tip to the + wire *near the transformer*. You’ll see high-amplitude, high-dv/dt transients—this is the mode that couples into WiFi antennas.
- Ground reference sweep: Use a near-field H-field probe (or even a 6-inch loop of insulated wire connected to scope input) held 2 inches from transformer case. Sweep slowly from 100 kHz to 30 MHz. If you see sustained energy >20 mVpp above 500 kHz, EMI is likely.
In one test, I recorded a 12 V transformer output showing 2.8 Vpp noise centered at 118 kHz—with spectral leakage visible at 2.42 GHz on a borrowed spectrum analyzer. That same unit caused packet loss rates of 18% on a nearby 2.4 GHz mesh node (measured with iPerf3 and ping -f over 60 seconds). Replace it with a compliant unit, and loss dropped to 0.3%.
FCC Part 15 Isn’t Optional—It’s Your Baseline
FCC Part 15 Class B sets emission limits for residential digital devices: ≤40 dBµV/m (100 µV/m) quasi-peak at 30–230 MHz, and ≤47 dBµV/m (224 µV/m) at 230–1000 MHz, measured at 3 meters. Most budget transformers violate this by 15–25 dB—especially in the 2–5 MHz range, where AM radio, garage door openers, and legacy IoT sensors live.
Crucially: Class B applies to ANY device marketed for home use—even if it’s “just a transformer.” Yet most sub-$40 units carry no FCC ID, no compliance labeling, and zero documentation about conducted or radiated emissions. They’re sold as “electrical components,” not “intentional radiators”—but they radiate unintentionally, and aggressively.
What to Look For in a Compliant Replacement
Don’t just swap brands. Swap architectures. Here’s my spec sheet for what actually works:
| Feature | Non-Compliant Unit | FCC-Compliant Unit |
|---|---|---|
| Topology | Unshielded flyback, no isolation barrier | Isolated LLC resonant or shielded forward converter |
| Output Filtering | None (or single ceramic cap) | π-filter: X-cap + common-mode choke + dual Y-caps to earth |
| Ferrite Cores | None on input or output | Split-core toroid on output leads (100 Ω @ 100 MHz), plus clamp-on ferrite on AC input cord |
| Shielding | Plastic housing only | Galvanized steel enclosure with conductive gasket; internal copper tape over PCB |
| EMI Test Data | Not published | FCC ID listed; full test report available online (e.g., “FCC ID: XYZ-TR12V40W-B”) |
I’ve tested six “pro-grade” replacements in the $75–$140 range. All passed Class B at 3 m in independent lab reports. The best performers shared three traits: (1) a grounded metal chassis that acts as a Faraday cage, (2) output leads wrapped with nanocrystalline ferrite tape (not just beads), and (3) active feedback that adjusts switching frequency dynamically to avoid resonant peaks in the cabling.
One unit—designed for architectural LED tape—uses a 350 kHz base frequency with spread-spectrum modulation. Its 2.4 GHz leakage was below noise floor (-92 dBm) on a calibrated spectrum analyzer. That’s not magic. It’s intentional engineering.
Practical Fixes That Don’t Require Rewiring
You don’t always need to replace the whole system. Try these first:
- Add a ferrite clamp to the transformer’s AC input cord, within 2 inches of the housing. Use a Type 31 or Type 43 mix rated for 1–10 MHz suppression. Two turns through a 12-mm core drops common-mode noise by 15–22 dB.
- Install a line filter between breaker and transformer. Not a power strip surge protector—look for a commercial-grade EMI filter (e.g., Schaffner FN2030B) with ≥60 dB insertion loss at 1 MHz. Mount it in a junction box upstream.
- Twist the low-voltage output wires tightly—not just at the transformer, but along the entire run. Twist pitch should be ≤½ inch. This reduces magnetic loop area and cancels differential-mode radiation. I’ve seen 8–10 dB improvement in field strength just from retwisting 60 ft of 14/2.
- Ground the transformer case to earth—but only if it has a grounding terminal and your local code permits. Do not bond low-voltage negative to ground unless the unit is specifically rated for it (most aren’t). Improper grounding creates ground loops and worsens EMI.
These measures won’t fix a fundamentally noisy design—but they can make marginal units usable. In one retrofit, adding a ferrite + twist + grounded metal enclosure cut WiFi packet loss from 14% to 2.1%—enough to restore Ring Doorbell responsiveness and Nest Cam streaming.
When to Walk Away From the Transformer Entirely
Sometimes the cheapest fix is the most expensive long-term. Consider replacing the entire low-voltage ecosystem if:
- Your transformer is older than 2018 (pre-UL 1849 revision, which added EMI language).
- You’re using LED bulbs with built-in rectifiers and no input filtering (common in $2–$4 path lights).
- You have >300 watts total load on a single circuit (increases dv/dt stress and harmonic generation).
- You’re running cable parallel to WiFi antennas for >15 ft (e.g., along a fence line adjacent to a mesh node).
In those cases, go direct-wire. Modern 120 VAC LED landscape fixtures—like integrated bollards or recessed step lights—eliminate the transformer entirely. Yes, they require GFCI-protected circuits and licensed installation. But they produce zero switching noise. One client switched four zones (28 lights) to 120 VAC LEDs and gained consistent 250+ Mbps throughput across the entire yard—no mesh node relocation, no channel tweaking.
A Final Note on “Good Enough”
I think we tolerate too much noise in residential systems. We accept flicker, hum, heat, and RF leakage because it’s “cheap” and “works most of the time.” But mesh WiFi isn’t background infrastructure anymore—it’s the nervous system of security cameras, HVAC controls, irrigation timers, and voice assistants. When your sprinkler controller stops responding because the transformer is shouting over 2.4 GHz, that’s not convenience. It’s failure.
So next time you buy a low-voltage transformer, ask for the FCC ID. Ask for the test report. If the answer is silence—or worse, “it’s UL listed, so it’s fine”—walk away. UL 1849 covers fire and shock safety. It says nothing about whether your garden lights will mute your Zoom call.
Lighting should illuminate—not interfere.