Skip to content

power.md

Latest commit

 

History

History
397 lines (332 loc) · 14.5 KB

power.md

File metadata and controls

397 lines (332 loc) · 14.5 KB

Power Report

This document contains power profiling results for embr, captured with the Nordic Power Profiler Kit II (PPK2). Values are intended to provide a repeatable baseline and enable apples-to-apples comparison as features are added (e.g., DMIC bring-up, wake-on-sound (WOS) + DSP/inference bursts, Thread/Matter activity).

Measurement Setup

Thingy:53 connected to the PPK2

Thingy:53 connected to the PPK2 via the current measurement and debug board.

Tool: Nordic Power Profiler Kit II (PPK2)
Measurement mode: Ammeter mode (target powered through PPK2)
Supply voltage: 3.12 V (measured at Thingy:53 VIN with DMM; PPK2 VOUT → VIN)
Target hardware: Nordic Thingy:53
Latest firmware state: WOS capture + DSP/inference + Thread-connected CoAP command transmission
Logging: RTT

Notes

  • Capture windows are taken after initialization settles unless stated otherwise.

Interpretation

The measurements below currently establish:

  • A general baseline for:

    • Verifying the system reaches low-power idle between specific events (e.g LED blink, WOS + DSP/inference events)
    • Comparing incremental power cost as features are added
    • Identifying unexpected background activity early

  • DMIC audio capture baseline:

    • 16 kHz sample rate
    • Continuous capture, no sleep/idle

  • DMIC WOS event baseline:

    • 1 s bursts, 16 k samples
    • Idle preceding/following burst activity

  • WOS bursts with subsequent DSP/inference baseline:

    • 1 s capture events, 16 k samples
    • ~175 ms avg. DSP + inference time
    • Idle preceding/following burst activity

  • Thread networking and CoAP command baseline:

    • OpenThread MTD/SED command-source role
    • Network-unavailable attach retry behavior (Thread stack up, no network reachable)
    • Thread-connected idle with WOS + DSP/inference armed
    • Voice-triggered WOS capture, DSP/inference, and multicast CoAP command transmission in the same event path

Future profiles planned:

  • Longer-duration Thread/SED connected idle
  • Thread commissioning / join transient
  • Isolated radio TX/RX bursts
  • End-to-end voice-triggered lighting control path through the Matter boundary

Results Summary

Notes

  • Results are presented in reverse chronological order. Measurements for the latest firmware/feature set will appear first.
  • Trial count and selection window are listed per section.
  • Steady-state captures use longer post-init windows.
  • Event measurements use selected regions around the active behavior being profiled.
  • Primary measurements:
    • Average current and average peak current
    • Measurements from DMIC bring-up and later also include: min/max, range, standard deviation (std dev), and coefficient of variation (CV)

~~~ Thread Networking + CoAP — Attach Retry, Connected Idle, and WOS Command Event ~~~

Behavior

  • Thread stack enabled as an OpenThread MTD/SED command source
  • When no Thread network is available, the device repeatedly attempts to attach instead of reaching the connected idle floor
  • When the Thread network is available, the device attaches, idles with WOS + DSP/inference armed, and sends voice-triggered multicast CoAP light commands
  • WOS command event path: WOS audio capture → DSP/inference → CoAP command transmission → return to Thread-connected idle

Notes

  • No-network attach retry trials: 3; 24.80-25.59 s per trial
  • Thread-connected idle trials: 3; 20.01-26.63 s per trial
  • WOS command event samples: 3 events from one longer Thread-connected run
  • Event sub-selections:
    • WOS audio capture: ~1.003 s average selection window
    • DSP/inference + CoAP transmission: ~217.8 ms average selection window
    • Full WOS-to-CoAP command window: ~1.218 s average selection window
  • Full command-window current and charge are taken from the direct PPK2 selection spanning WOS capture through CoAP transmission.
  • The no-network captures are not an idle power floor; they represent repeated Thread attach attempts when no network is reachable.
  • Figures below show one representative screenshot per profiled behavior/sub-selection; all trial screenshots are used for the calculations but are not embedded.
  • PPK2 screenshot readouts are rounded to the displayed precision; derived statistics inherit that precision.

Thread network unavailable (attach retry)

  • Average current: 3.97 mA

    • Min/Max: 3.91 mA / 4.07 mA
    • Range: 0.16 mA
    • Std dev (sample): 0.09 mA
    • CV: 2.26%

  • Average peak current (attach-retry bursts): 25.50 mA

    • Min/Max: 21.86 mA / 28.18 mA
    • Range: 6.32 mA
    • Std dev (sample): 3.27 mA
    • CV: 12.81%

  • Average charge per attach-retry window: 100.40 mC

    • Min/Max: 100.09 mC / 100.93 mC
    • Range: 0.84 mC
    • Std dev (sample): 0.46 mC
    • CV: 0.46%

Thread-connected idle (WOS + DSP/inference armed)

  • Average current: 113.98 µA

    • Min/Max: 113.84 µA / 114.09 µA
    • Range: 0.25 µA
    • Std dev (sample): 0.13 µA
    • CV: 0.11%

  • Average peak current (background spikes): 10.22 mA

    • Min/Max: 8.56 mA / 11.13 mA
    • Range: 2.57 mA
    • Std dev (sample): 1.44 mA
    • CV: 14.11%

WOS audio capture event (Thread connected)

  • Average current: 713.67 µA

    • Min/Max: 708.22 µA / 719.07 µA
    • Range: 10.85 µA
    • Std dev (sample): 5.43 µA
    • CV: 0.76%

  • Average peak current: 14.68 mA

    • Min/Max: 13.98 mA / 15.06 mA
    • Range: 1.08 mA
    • Std dev (sample): 0.61 mA
    • CV: 4.13%

  • Average charge per WOS capture: 715.95 µC

    • Min/Max: 710.84 µC / 720.10 µC
    • Range: 9.26 µC
    • Std dev (sample): 4.70 µC
    • CV: 0.66%

DSP/inference + CoAP transmission event

  • Average current: 3.34 mA

    • Min/Max: 3.28 mA / 3.40 mA
    • Range: 0.12 mA
    • Std dev (sample): 0.06 mA
    • CV: 1.80%

  • Average peak current: 13.34 mA

    • Min/Max: 12.87 mA / 13.91 mA
    • Range: 1.04 mA
    • Std dev (sample): 0.53 mA
    • CV: 3.95%

  • Average charge per DSP/inference + CoAP transmission: 726.97 µC

    • Min/Max: 708.57 µC / 742.19 µC
    • Range: 33.62 µC
    • Std dev (sample): 17.04 µC
    • CV: 2.34%

Full WOS-to-CoAP command event

  • Average current: 1.187 mA

    • Min/Max: 1.18 mA / 1.19 mA
    • Range: 10.00 µA
    • Std dev (sample): 5.77 µA
    • CV: 0.49%

  • Average peak current: 14.68 mA

    • Min/Max: 13.98 mA / 15.06 mA
    • Range: 1.08 mA
    • Std dev (sample): 0.61 mA
    • CV: 4.13%

  • Average charge per WOS-to-CoAP command: 1.443 mC

    • Min/Max: 1.43 mC / 1.45 mC
    • Range: 20.00 µC
    • Std dev (sample): 11.55 µC
    • CV: 0.80%

Derived comparisons against previous WOS + DSP/inference profile

  • Thread-connected idle current is 17.74 µA above the previous WOS + DSP/inference armed idle baseline (113.98 µA vs 96.24 µA).
  • Network-unavailable attach retry averages 3.85 mA above Thread-connected idle (~34.8x connected idle current).
  • Full WOS-to-CoAP command charge is ~86.8 µC above the previous WOS + DSP/inference event charge (1.443 mC vs 1.3565 mC), over a direct selection window ~43 ms longer (1.218 s vs ~1.175 s).
  • These deltas include the full Thread-enabled runtime path and are not isolated radio-only TX/RX costs.

Representative attach-retry capture (Thread network unavailable)

Thread stack enabled with no Thread network available, attach retry current

Representative attach-retry capture with no Thread network available; repeated attach attempts dominate the window.

Representative steady-state capture (Thread connected idle)

Thread connected idle with WOS and DSP/inference armed

Representative Thread-connected idle capture with WOS + DSP/inference armed.

Representative event selections (one screenshot per sub-selection)

Thread-connected WOS audio capture selection showing approximately 1 second active capture

Representative Thread-connected WOS audio capture selection showing an approximately 1 s active capture window.

Thread-connected DSP/inference and CoAP command transmission selection

Representative Thread-connected DSP/inference + CoAP command transmission selection.

Full Thread-connected WOS-to-CoAP command selection

Representative full Thread-connected WOS-to-CoAP command selection spanning capture through command transmission.

~~~ DSP/Inference after DMIC Wake-on-Sound (WOS) — Idle + Wake/Inference Event ~~~

Behavior

  • Wake-on-sound (WOS) armed; device remains in low-power idle until a wake event occurs
  • Wake event produces an ~1 s active capture window, then performs DSP and runs inference
  • Returns to idle state

Notes

  • Idle trials: 3; 26.85 s per trial
  • Wake + DSP/inference event samples: 3 events, ~1.175 s average combined selection window per event
  • Combined event average current is calculated from selected charge and time:
    • I_avg = (Q_capture + Q_DSP/inference) / (t_capture + t_DSP/inference)
  • Combined event peak current is the larger peak observed across the capture and DSP/inference selections for that event

Idle current (WOS + DSP/Inference armed)

  • Average current: 96.24 µA

    • Min/Max: 95.33 µA / 96.79 µA
    • Range: 1.46 µA
    • Std dev (sample): 0.80 µA
    • CV: 0.83%

  • Average peak current (background spikes): 12.83 mA

    • Min/Max: 11.40 mA / 14.34 mA
    • Range: 2.94 mA
    • Std dev (sample): 1.47 mA
    • CV: 11.48%

Wake + DSP/Inference event

  • Average current: 1.154 mA

    • Min/Max: 1.149 mA / 1.159 mA
    • Range: 9.75 µA
    • Std dev (sample): 4.89 µA
    • CV: 0.42%

  • Average peak current: 16.34 mA

    • Min/Max: 15.73 mA / 17.30 mA
    • Range: 1.57 mA
    • Std dev (sample): 0.84 mA
    • CV: 5.16%

  • Average charge per wake + DSP/inference event: 1356.53 µC

    • Min/Max: 1351.54 µC / 1362.08 µC
    • Range: 10.54 µC
    • Std dev (sample): 5.29 µC
    • CV: 0.39%

Representative steady-state capture (idle)

WOS + DSP/inference armed, idle current (no wake events)

WOS + DSP/inference armed, idle current (no wake events).

Representative event selections (wake + capture/DSP/inference events)

WOS + capture event (selected region) showing ~1 s active capture window

WOS + capture event (selected region) showing ~1 s active capture window.

DSP/Inference event (selected region) showing ~176 ms active inference window

DSP/Inference event (selected region) showing ~176 ms active DSP/inference window.

~~~ DMIC Wake-on-Sound (WOS) — Idle + Wake Event ~~~

Behavior

  • Wake-on-sound (WOS) armed; device remains in low-power idle until a wake event occurs
  • Wake event produces an ~1 s active window, then returns to idle

Notes

  • Idle trials: 3; 26.85 s per trial
  • Wake event samples: 6 events, ~1.0 s selection window per event

Idle current (WOS armed)

  • Average current: 101.44 µA

    • Min/Max: 98.08 µA / 105.75 µA
    • Range: 7.67 µA
    • Std dev (sample): 3.92 µA
    • CV: 3.87%

  • Average peak current (background spikes): 11.33 mA

    • Min/Max: 10.82 mA / 11.63 mA
    • Range: 0.81 mA
    • Std dev (sample): 0.44 mA
    • CV: 3.92%

Wake event (selection window ~1.006 s)

  • Average current: 673.25 µA

    • Min/Max: 666.52 µA / 677.54 µA
    • Range: 11.02 µA
    • Std dev (sample): 4.63 µA
    • CV: 0.69%

  • Average peak current (wake burst): 14.83 mA

    • Min/Max: 12.45 mA / 16.19 mA
    • Range: 3.74 mA
    • Std dev (sample): 1.36 mA
    • CV: 9.14%

  • Average charge per wake (selection): 677.66 µC

    • Min/Max: 672.72 µC / 681.65 µC
    • Range: 8.93 µC
    • Std dev (sample): 3.93 µC
    • CV: 0.58%

Representative steady-state capture (idle)

WOS armed, idle current (no wake events)

WOS armed idle current (no wake events).

Representative steady-state capture (wake events)

WOS events showing ~1 s active window per wake

WOS events showing ~1 s active window per wake (selection window used for per-event stats).

~~~ DMIC (PDM) Bring-up - Continuous Audio Capture ~~~

Behavior

  • Audio sample rate: 16kHz
  • Mode: Mono, single channel
  • PDM buffer size: 4000 samples
  • Buffer release rate: 4 releases per second
  • Peripherals enabled: DMIC

Steady-state current (post init)

  • Average current: 646.08 µA
    • Min/Max: 644.98 µA / 647.29 µA
    • Range: 2.31 µA
    • Std dev (sample): 1.16 µA
    • CV: 0.18%
  • Average peak current: 16.70 mA
    • Min/Max: 16.37 mA / 17.31 mA
    • Range: 0.94 mA
    • Std dev (sample): 0.53 mA
    • CV: 3.18%

Representative steady-state capture

DMIC running continuous capture at 16 kHz

DMIC running continuous audio capture at 16 kHz

~~~ Baseline — Blinky ~~~

Behavior

  • Blink period: 500 ms
  • Duty cycle: 50%
  • Peripherals enabled: LED, Timer

Steady-state current (post init)

  • Average current: 10.46 µA
  • Average peak current (blink event): 13.60 mA

Representative steady-state capture

Blinky running with 500 ms period and 50% duty cycle

Blinky running with 500 ms period and 50% duty cycle

Informational: Boot Transient

Boot current is recorded separately as a transient event.

  • Average boot peak current: 53.20 mA
  • Observations: Generally stable across boots, logging enabled

Representative boot capture

Boot + init power sequence

Boot + init power sequence