power-testing

MagTag Power Testing

Testing the MagTag's power usage to better estimate battery life in various types of use cases.

This type of testing will also be useful for comparing different versions of CircuitPython, different hardware configurations, and different code as time moves on.

Tools Required

• Adafruit MagTag
  • Adafruit: https://www.adafruit.com/product/4800
  • Digi-Key: https://www.digikey.com/en/products/detail/adafruit-industries-llc/4800/13616787
• Nordic PPK2 Power Profiler Kit
  • Adafruit: https://www.adafruit.com/product/5048
  • Digi-Key: https://www.digikey.com/en/products/detail/nordic-semiconductor-asa/NRF-PPK2/13557476
• Nordic Power Profiler software
  • Included with the freely available nRF Connect for Desktop software package
• JST PH 2-Pin to female socket cable for power
  • Can't find one of these online, but can be custom crimped pretty easily
• JST PH 3-Pin (STEMMA) to female socket cable for digital IO
  • Adafruit: https://www.adafruit.com/product/3894
  • Can also be custom crimped pretty easily

Setup

1. MagTag battery input 2-pin JST PH connector VBAT, GND wired to PPK2 VOUT, GND in SMU mode
2. MagTag D10 3-Pin JST PH connector V+, GND, Signal wired to PPK2 Logic Port VCC, GND, and D7
3. Nordic PPK2 USB/Data connected to an Ubuntu Linux system running Nordic's Power Profiler v3.2.0
4. Nordic PPK2 USB Power connected to a 5V 1A USB-A charger
5. Nordic Power Profiler software configuration:

• Mode: Source Meter
• Samples Per Second: 100,000
• Sampling Duration: 432 seconds
• Timestamps: On
• Digital Channels: On, D7 enabled, all others disabled
• No changes to default "advanced settings" (gain/smoothing)

Test Methodology

Start with PPK2 and MagTag devices disconnected with power switches in the off position.

1. Attach MagTag to PPK2 SMU and Digital Input headers
2. Plug in PPK2
3. Start Power Profiler software
4. Start Data Logger
5. Enable Power Output
6. Turn on MagTag
7. Wait for test to complete
8. Stop Data Logger
9. Save/Export Logged Data

Outputs

Each test run should result in the creation of two files:

• A binary ppk for analysis within Nordic Power Profiler
• A plain-text csv for analysis using any other tools/methods

Loading New Code

Because the PPK2's SMU can be disabled in software, we can connect the MagTag to a computer via USB-C for code updating.

Its probably a good idea to make sure the SMU is disabled when the MagTag is attached to the host system though.

Once code is updated/debugged, power off the MagTag and detach the USB-C cable before turning its power back on.

Test Concepts and Methodology

The general idea is to import modules, execute user code, change states, use peripherals, and perform other actions with periods of time.sleep() between them. Performing only one one (or very few) actions before entering 1 or 10 second sleep states assists greatly in identifying the specific code being executed while analyzing things.

PPK2 Digital IO

The PPK2's digital IO inputs can be useful for helping to pinpoint specifically where we are in the code while digging into things.

In the first test code revision this is flipped to high on a few occasions to make it easier to identify the execution a few specific chunks of code.

• While SPEAKER_ENABLE is set to True
• While blinking the board LED
• While testing the NeoPixels
• While connecting to the WiFi network

A second output pin is configured on A1 but didn't end up being used in this testing. This is flipped to True when the display is refreshing.

Enabling one or more of these, regardless of whether they're connected or floating, appears to increase power consumption by around 6-7 mA while idle so this should be taken into account while doing any analysis.

Code

magtag-power-test-v1.py

The first tests were conducted using magtag-power-test-v1.py. This was written in just a few minutes without a lot of debugging just to see if this would work conceptually.

There are 7 main phases of this code with 10 second sleeps between them:

1. Startup and import of time, board, digitalio
2. Init of peripherals

• indicator_pin
• indicator2_pin,
• board LED,
• SPEAKER_ENABLE, enabled for 1 sec, disable for 1 sec
• NEOPIXEL_POWER_INVERTED, enabled for 2 sec, disabled for 1 sec
• Init of NeoPixels

3. LED/Neopixel Tests

• Flash of board D13 LED for 1 sec
• Single NeoPixel max/half-bright for 1 sec each
• All NeoPixels max/half-bright for 1 sec each
• NeoPixels deinit(), sleep 1 sec
• NEOPIXEL_ENABLE pin to False, sleep 1 sec

4. Refresh of display (will contain console output from print statements)
5. Preparing WiFi connect with import of secrets and wifi and definition of a helper method
6. Connect to WiFi network
7. Disable WiFi
8. Loop:

• If radio enabled and WiFi not connected, reconnect.
• Refresh display, toggle wifi.radio.enabled

Test Results

magtag-power-test-v1, first run

Device: MagTag w/ ESP32-S2-WROOM module Bootloader: 0.5.2 CircuitPython: 7.0.0 (release)

Data from first test run, 3.7 V:

• Initial boot: 34.91 mA avg (870 mA peak) over 1.329 s for 46.41 mC of charge
• CircuitPython boot: 44.03 mA avg (700 mA peak) over 1.054 s for 46.43 mC of charge
  • Each NeoPixel blink (~125 ms) averages 45.23 mA (96.80 mA peak) for 5.70 mC of charge
  • Idle between blinks: 44.22 mA
  • Idle after blinks, before first imports: 27.09 mA (about 1.5 s)
• Phase 0: 1x print statement, import time, sleep(10): 46.30 mA avg (peak 65.31 mA) over 488.1 mS for 23.63 mC of charge
  • Idle after print, before import: 27.1 mA avg (31.30mA peak) over 35.12 mS
• Phase 0 End, sleep(10): 29.28 mA avg (34.53 mA peak) over 9.999 s for 292.76 mC of charge
• Phase 1: print statement, import board, import digitalio, print statement
  • 52.02 mA avg (57.48 mA peak) over 0.596 mS for 0.031 mC of charge
• Phase 1 End, sleep(10): 29.30 mA (34.23 mA peak) over 9.999 s for 292.98 mC of charge
• Phase 2: Digital I/O and NeoPixel setup: 30.57 mA avg (720 mA peak) over 6.035 s for 184.52 mC of charge
  • Pin setup: 49.54 mA (56.11 mA peak) over 0.25 mS for 0.012 mC of charge
  • Speaker pin enable, sleep(1), disable: Idle increases to 36.28 mA (delta of +6.98 mA)
  • NeoPixel import and setup: 52.21 mA avg (720 mA peak) over 32.49 mS for 1.70 mC of charge
• Phase 2 End, sleep(10): 30.54 mA avg (35.98 mA peak) over 9.999 s for 305.38 mC of charge
• Phase 3: LED and Neopixel tests: 51.55 mA avg (207.53 mA peak) over 16.01 s for 830 mC of charge
  • Indicator pin 1 on, on for 1 sec: 32.90 mA avg (59.50 mA peak) over 1 s for 32.91 mC of charge
  • I1 and LED on for 1 sec: 33.07 mA avg (56.69 mA peak) over 1 s for 33.07 mC of charge
  • pixel[0] = (255,255,255) @ 1.0: 68.68 mA avg (95.26 mA peak) over 1 s for 68.74 mC of charge
  • pixel[0] = (255,255,255) @ 0.5: 49.99 mA avg (96.80 mA peak) over 1 s for 50.05 mC of charge
  • pixel[0] = (255,255,255) @ 0.1: 35.48 mA avg (94.49 mA peak) over 1 s for 35.52 mC of charge
  • pixel[0] = (255,255,255) @ 0.05: 33.35 mA avg (83.73 mA peak) over 1 s for 33.38 mC of charge
  • pixel[0] = (255,0,0)@ 1.0: 45.35 mA avg (77.58 mA peak) over 1 s for 45.39 mC of charge
  • pixel[0] = (0,255,0)@ 1.0: 45.48 mA avg (76.81 mA peak) over 1 s for 45.48 mC of charge
  • pixel[0] = (0,0,255)@ 1.0: 43.90 mA avg (76.05 mA peak) over 1 s for 43.94 mC of charge
  • pixels.fill(0,0,0): 32.94 mA avg (75.28 mA peak) over 1 s for 32.97 mC of charge
  • pixels.fill(255,255,255) @ 1.0: 163.29 mA avg (207.53 mA peak) over 1 s for 163.44 mC of charge
  • pixels.fill(255,255,255) @ 0.5: 99.30 mA avg (191.96 mA peak) over 1 s for 99.40 mC of charge
  • pixels.fill(255,255,255) @ 0.1: 43.84 mA avg (181.07 mA peak) over 1 s for 43.86 mC of charge
  • pixels.fill(255,255,255) @ 0.05: 35.01 mA avg (131.01 mA peak) over 1 s for 35.07 mC of charge
  • pixels.deinit(): 31.68 mA avg (99.87 mA peak) over 1 s for 31.71 mC of charge
• Phase 3 End, sleep(10): 29.28 mA avg (34.58 mA peak) over 9.999 sec for 292.83 mC
• Phase 4: Display Init and Refresh: 32.8 mA avg (66.84 mA peak) over 5.371 s for 176.15 mC of charge
  • Core display refresh process: 35.04 mA avg (66.84 mA peak) over 2.371 s for 83.11 mC of charge
  • Note: indicator2 pin is pulled up, probably adding 6-7 mA on top of base current
• Phase 4 End, sleep(10): 29.30 mA avg (34.63 mA peak) over 9.999 s for 293.03 mC of charge
• Phase 5: from secrets import secrets, import wifi: 109.45 mA avg (289.75 mA peak) over 154.5 ms for 16.91 mC of charge
• Phase 5 End, sleep(10): 79.52 mA avg (107.57 mA peak) over 9.999 s for 800 mC of charge
  • Average current draw delta of +50.22 mA, more than doubling current, after importing wifi and before any WiFi activity
• Phase 6: wifi.radio.connect(): 92.40 mA avg (670 mA peak) over 3.362 s for 310.72 mC of charge
• Phase 6 End, sleep(10): 31.32 mA avg (307.08 mA peak) over 9.999 s for 313.21 mC of charge
  • Average current draw is back down a delta of only +2.02 mA over previous base power draw but there are large 300 mA spikes every 315 ms or so lasting about 12-13 ms before dropping back to a new, slightly lower baseline of about 27.02 mA.
• Phase 7: wifi.radio.enabled = False: 154.51 mA avg (328.38 mA peak) over 29.26 ms for 4.52 mC of charge
• Phase 7 end, sleep(10): 27.04 mA avg (51.2 mA peak) over 9.999 s for 270.42 mC of charge

Totals: v1, first run

Run Time: 115.3 s Current (avg): 39.28 mA Current (peak): 870 mA Charge: 4.53 C

EOF