Antenna Interference & Servo EMI

RF Interference & Servo Motor EMI

Problem: Battery level stops transmitting after case is closed. GPS still works. Servo is active.
Root cause: Servo motor generates broadband noise (commutator arcing) that couples into I2C bus and antenna circuits.

---

Why Servo Causes EMI

Commutator Arcing

Brushed DC motors generate microarcs (nanosecond pulses) at each brush contact. Each spark = impulse with spectral content up to 1 GHz and beyond.

Frequency	Noise Level	Affected Systems
1 MHz – 100 MHz	-20 dBm typical	PWM harmonics, I2C, analog circuits
100 MHz – 1 GHz	-40 to -60 dBm	GPS L1 (1575 MHz), LTE RX, LNA input

Peak Current Spikes

Servo startup: 300–700 mA for 1–3 seconds. Inadequate filtering → voltage droop → noise injection into analog rail.

Back-EMF Spikes

Motor coil inductance generates 50–100 V spikes when servo stops. Without protection:

• Can exceed MOSFET rating
• Couples RF noise onto power rails (antenna effect)
• Causes I2C communication errors

---

Impact on Cloudlock Systems

GPS Reception (Most Vulnerable)

• Expected signal: -130 dBm
• GPS front-end noise figure: 1–2 dB
• If servo EMI > -80 dBm near antenna → GPS can’t lock

Observed: GPS continues (separate cable, better isolation). Battery level sensor (I2C, on-board) fails.

Battery Fuel Gauge (I2C Vulnerable)

• I2C operates at low voltages (3.3V logic swing)
• Servo noise couples onto I2C SCL/SDA lines
• Device responds with NAK (not acknowledge)
• Fuel gauge reading = 0 or invalid → not sent to Notehub

LTE Modem

• Usually more robust than GPS (designed for noisy environments)
• Higher SNR threshold than GPS
• Still affected if servo near Notecard RF section

---

Solutions (In Priority Order)

1. RC Snubber on Motor Terminals (Critical)

Motor+ ──┬─ [10Ω 1W] ── [10nF/100V] ──┬─ Motor-
         │ (soldered directly on motor terminals)
         │ SMD 1210 size (fast response)

Effect: Back-EMF spike reduced from -100V to -15V (safe). Broadband noise peak reduced by 20 dB.

Part: Vishay CRCW121010R0FKEA (10Ω) + Murata GRM32DR72A103KW01L (10nF)

2. Ferrite Bead on Servo Power Line (Critical)

MOSFET Drain ── [Ferrite: Murata BLM18KG471SN1D] ── Servo V+

Effect: High-frequency noise (>1 MHz) absorbed, low-frequency PWM unaffected. Suppresses antenna-mode RF emission from servo power wire.

3. Separate Power Rail for Servo (Important)

Battery → [MOSFET] → [LC filter: 10µH + 100µF] → Servo
              ↑
         [GPIO control]

Isolate servo spikes from MCU/GPS power. RC time constant ensures slow ramp, no sharp edges.

4. I2C Pull-Up Resistor Value (Careful)

• Notecard has built-in 2.2k pull-ups
• If XIAO also has pull-ups: don’t add external 4.7k (parallel = 1.4k, too strong, bus latency)
• If XIAO has none: add external 4.7k (good)

Check: Measure SDA/SCL voltage with multimeter when I2C idle. Should be 3.2–3.3V, not oscillating.

5. RF Shield Can Over GPS Front-End (Recommended)

Harwin S01 or Würth WE-SHC shield can over GPS LNA section gives 30–40 dB isolation from broadband servo noise.

6. Accelerometer-Based Gating (Elegant)

if (accelerometer_motion_detected() && servo_not_running) {
    enable_gps();
} else {
    disable_gps();
}

Keep GPS off during servo operation → no interference.

---

PCB Layout Recommendations

Floorplan: Separate Zones

┌─────────────────────────────────────────┐
│  ZONE A: RF (GPS, LTE front-end)        │
│  [Keep servo area >100mm away]          │
├─────────────────────────────────────────┤
│  ZONE B: Digital (MCU, I2C)             │
│  [Via fence between A and C]            │
├─────────────────────────────────────────┤
│  ZONE C: Servo + Power                  │
│  [LC filter, ferrite beads here]        │
└─────────────────────────────────────────┘

Servo Traces

• Wide (≥2 mm) to reduce inductance
• Short, kept in ZONE C only
• 100Ω series resistor + ferrite on PWM signal line
• GND plane stitching (via spacing ≤10 mm)

I2C Traces

• Keep away from servo power lines
• 100nF bypass capacitor at each I2C device
• If >50 mm long: ferrite bead on SCL or common-mode choke on pair

---

Testing Procedure

Test 1: GPS C/N0 (Signal-to-Noise)

1. Servo OFF, measure GPS C/N0 for satellites → baseline
2. Servo ON (positioning), re-measure C/N0
3. Acceptable: <3 dB degradation
4. Unacceptable: >5 dB drop (needs filtering)

Test 2: I2C Communication

1. Loop: read fuel gauge every 100 ms while servo OFF
2. Repeat while servo running
3. Count communication failures (CRC errors)
4. Acceptable: 0 failures
5. Unacceptable: >1% error rate (needs RC snubber)

Test 3: LTE RSRP

1. Read Notecard RSRP (via card.wireless) with servo OFF
2. Activate servo, re-read RSRP
3. Acceptable: <5 dB difference

---

Frequency Isolation Strategy

If using multiple RF systems (LTE, GPS, Skylo satellite):

Pair	Freq Gap	Isolation Needed	Method
GPS L1 ↔ LTE	225 MHz	30 dB	Distance (100 mm) + shield can
GPS L1 ↔ Iridium	41 MHz	40 dB	Notch filter + firmware muting
LTE ↔ Skylo	450+ MHz	20 dB	Ferrite bead + distance

Critical: GPS L1 ↔ Iridium overlap is dangerous. If satellite TX active, disable GPS RX (firmware lock).

---

Summary Table: EMI Sources → Solutions

EMI Source	Symptoms	Quick Fix	Proper Fix
Back-EMF spike	I2C timeouts, reset	RC snubber	RC snubber + soft-start PWM
Servo power noise	GPS C/N0 drops	Ferrite bead	Separate servo rail + LC filter
PWM harmonics	I2C glitches	Ferrite on PWM	Common-mode choke on I2C pair
Broadband arcing	Battery level fails	Shield can	Proper grounding + via stitching
Antenna coupling	GPS lock loss	100 mm spacing	Separate antennas + isolation

---

Specific Recommendations for Cloudlock

✅ Must do:

1. Add RC snubber (10Ω + 10nF) directly on servo terminals
2. Ferrite bead on servo power line (after MOSFET)
3. Verify I2C pull-ups are correct (don’t double-up)

✅ Should do: 4. RF shield can over GPS/LTE front-end 5. Separate servo power rail with LC filter 6. Via stitching between RF and power zones

✅ Nice to have: 7. Common-mode choke on I2C if traces >50 mm 8. Firmware muting of GPS during servo operation 9. Accelerometer-gated GPS (only on while moving)