Skip to content

0021: Motor Stall Detection via Back-EMF Sensing

Date: 2025-10-15 Phase: 0.4 Status: Accepted Type: Architecture

Summary (Y-Statement)

In the context of motor stall protection for safety-critical EMDR therapy, facing the need for current sensing without dedicated hardware, we decided for back-EMF sensing with resistive voltage summing, and neglected battery voltage drop monitoring as primary method, to achieve power-efficient continuous motor health monitoring, accepting 165µA continuous bias current and 10ms measurement delay.

Problem Statement

ERM motor stall condition (120mA vs 90mA normal) could damage H-bridge MOSFETs and accelerate battery drain. No dedicated current sensing hardware exists in discrete MOSFET design. Battery voltage monitoring for stall detection is power-inefficient (resistor divider draws ~248µA continuously). Back-EMF can swing from -3.3V to +3.3V, but ESP32-C6 ADC only accepts 0V to 3.3V.

Context

Technical Constraints: - Discrete MOSFET H-bridge design (no integrated current sensing) - ESP32-C6 ADC: 0V to 3.3V input range - Motor back-EMF: -3.3V to +3.3V range (bidirectional) - Battery divider: 248µA continuous draw when enabled - Power budget: Target <1mA standby current

Safety Requirements: - Detect stalled motor to prevent MOSFET damage - Continuous monitoring capability during therapy sessions - Graceful degradation to LED stimulation if motor fails

Hardware Resources: - GPIO0 (ADC1_CH0) available for back-EMF sensing - GPIO2 (ADC1_CH2) reserved for battery voltage monitoring

Decision

We will implement software-based motor stall detection using power-efficient back-EMF sensing as the primary method.

Implementation:

  1. Back-EMF Sensing (Primary Method):
  2. GPIO0 (ADC1_CH0) reads OUTA from H-bridge during coast periods
  3. Resistive summing network biases and scales ±3.3V to 0-3.3V ADC range
  4. Low-pass filter (5kΩ + 22nF, fc ≈ 1.45 kHz) removes 25kHz PWM noise
  5. 10ms coast delay allows back-EMF and filter to stabilize

  6. Stall threshold: back-EMF magnitude < 1000mV

  7. Battery Voltage Drop Analysis (Backup Method):

  8. GPIO2 (ADC1_CH2) monitors battery voltage
  9. Compare voltage drop: no-load vs. motor-active

  10. Stall threshold: voltage drop > 300mV suggests excessive current

  11. Signal Conditioning Circuit: 

            R_bias (10kΩ)
3.3V ---/\/\/\---+
                 |
   R_signal (10kΩ)|
OUTA ---/\/\/\---+--- GPIO0 (ADC input) ---> [ESP32-C6 ADC, ~100kΩ-1MΩ input Z]
                 |
              C_filter
               (22nF)
                 |
                GND

  12. Voltage Mapping:

  13. V_GPIO1 = (3.3V + V_OUTA) / 2
  14. V_OUTA = -3.3V → V_GPIO1 = 0V (ADC minimum)
  15. V_OUTA = 0V → V_GPIO1 = 1.65V (ADC center)

  16. V_OUTA = +3.3V → V_GPIO1 = 3.3V (ADC maximum)

  17. Stall Response Protocol:

  18. Immediate coast: Set both H-bridge inputs low
  19. Mechanical settling: 100ms vTaskDelay()
  20. Reduced intensity restart: Retry at 50% intensity
  21. LED fallback: Switch to LED stimulation if stall persists
  22. Error logging: Record stall event in NVS for diagnostics

Consequences

Benefits

  • Power efficiency: 165µA continuous vs. 248µA for battery monitoring (33% more efficient)
  • Continuous monitoring: Can check motor health frequently without battery drain
  • Direct indication: Stalled motor has no back-EMF (direct mechanical failure indicator)
  • Perfect ADC range utilization: 100% of 0-3.3V range maps ±3.3V back-EMF
  • Therapeutic continuity: Graceful degradation to LED stimulation preserves therapy session

Drawbacks

  • 165µA continuous bias current (acceptable for 640mAh battery capacity)
  • 10ms measurement delay required for filter settling (negligible for 500ms cycles)
  • Requires coast period to develop back-EMF (already part of motor control timing)
  • Component count: Adds 2 resistors + 1 capacitor per motor to BOM

Options Considered

Option A: Battery Voltage Drop Monitoring (Rejected)

Pros: - Simple implementation (single ADC read) - No additional analog conditioning hardware

Cons: - 248µA continuous power draw from resistor divider - Less direct indication of motor stall - Requires baseline vs. active comparison (two measurements)

Selected: NO Rationale: 50% higher power consumption and indirect measurement make it inferior to back-EMF sensing

Option B: Back-EMF Sensing with Voltage Summing (CHOSEN)

Pros: - 33% more power efficient than battery method (165µA vs. 248µA) - Direct mechanical failure indication - Continuous monitoring capability - Perfect ADC range utilization (100% of 0-3.3V)

Cons: - Requires analog conditioning circuit - 10ms settling delay needed

Selected: YES Rationale: Superior power efficiency, direct indication, and continuous monitoring capability

Option C: Integrated H-Bridge IC with Current Sensing (Future)

Pros: - Hardware current sensing with dedicated sense resistor - Thermal protection integrated - Shoot-through protection via hardware interlocks - Detailed fault reporting via SPI/I2C

Cons: - Requires PCB redesign - Higher component cost - Not available for current discrete MOSFET design

Selected: NO (future enhancement) Rationale: Excellent long-term solution, but back-EMF sensing is adequate for current hardware

  • AD005: Motor H-Bridge Control - Defines H-bridge GPIO mappings and control signals
  • AD007: Battery Voltage Monitoring - Defines battery ADC channel (GPIO2) as backup stall detection method
  • AD027: Modular Source File Architecture - Battery monitor module calls stall detection

Implementation Notes

Code References

  • Circuit analysis: docs/architecture_decisions.md AD021 (voltage summing equations)
  • Future integration: src/battery_monitor.c (stall detection support functions)
  • Test validation: test/single_device_battery_bemf_test.c lines 1-1500 (hardware validation)

Circuit Analysis

Voltage Summing Principle: 

By Kirchhoff's Current Law (ADC draws negligible current):
I_bias = I_signal
(3.3V - V_GPIO1) / R_bias = (V_GPIO1 - V_OUTA) / R_signal

With R_bias = R_signal = 10kΩ:
3.3V - V_GPIO1 = V_GPIO1 - V_OUTA
3.3V + V_OUTA = 2 × V_GPIO1

V_GPIO1 = (3.3V + V_OUTA) / 2
V_GPIO1 = 1.65V + 0.5 × V_OUTA

Low-Pass Filter Characteristics: 

R_parallel = R_bias || R_signal = 10kΩ || 10kΩ = 5kΩ
f_c = 1 / (2π × R_parallel × C_filter)
f_c = 1 / (2π × 5kΩ × 22nF) ≈ 1.45 kHz

- Filters 25kHz PWM switching noise (17× attenuation, >94% reduction)
- Preserves ~100-200Hz motor back-EMF fundamental frequency
- Settles in ~0.55ms (5τ = 550µs, sufficient for 1ms+ coast measurement window)

Build Environment

  • Environment Name: single_device_battery_bemf_test (hardware validation)
  • Configuration File: sdkconfig.single_device_battery_bemf_test
  • Hardware: Seeed XIAO ESP32-C6 + TB6612FNG H-bridge

Testing & Verification

Hardware Testing Performed: - Validated voltage summing circuit: -3.3V maps to 0V, 0V maps to 1.65V, +3.3V maps to 3.3V - Confirmed 22nF capacitor provides adequate PWM noise filtering - Verified stall detection: Motor with blocked shaft shows <1000mV back-EMF - Normal operation: Back-EMF magnitude 1000-2000mV during coast periods

Known Limitations: - Requires 10ms coast period for accurate measurement - R_load intentionally not populated (would reduce ADC range to 67%) - Production uses 22nF capacitor (prototypes used 12nF, original design 15nF)

JPL Coding Standards Compliance

  • ✅ Rule #1: No dynamic memory allocation - Static circuit, no malloc
  • ✅ Rule #2: Fixed loop bounds - Stall check during fixed coast periods
  • ✅ Rule #3: No recursion - Linear stall detection algorithm
  • ✅ Rule #5: Return value checking - ESP_ERROR_CHECK on ADC reads
  • ✅ Rule #6: No unbounded waits - vTaskDelay() for 10ms settling time
  • ✅ Rule #7: Watchdog compliance - Fast ADC reads don't block task
  • ✅ Rule #8: Defensive logging - ESP_LOGI on stall detection events

Migration Notes

Migrated from docs/architecture_decisions.md on 2025-11-21 Original location: AD021 Git commit: (phase 0.4 implementation)

Template Version: MADR 4.0.0 (Customized for EMDR Pulser Project) Last Updated: 2025-11-21