Skip to content

HARDWARE_TESTPLAN.md

Latest commit

 

History

History
217 lines (196 loc) · 12.3 KB

HARDWARE_TESTPLAN.md

File metadata and controls

217 lines (196 loc) · 12.3 KB
Raw

Hardware Test Plan: PCB V3 First, PCB V4 Second

Successful compiler builds validate interfaces and memory limits, but they do not replace tests on real hardware. Record each result with the date, firmware commit, board revision, power source and measurement equipment.

1. Station only

  1. Upload the station target. Confirm that the upload erases the old flash/NVS state and that the W-Charger-XXXXXX setup network appears.
  2. Connect with W-Charger-Setup, open http://192.168.4.1/ and complete the unconfigured portal without a website login. Confirm that its ThingSpeak step only asks for the optional User API Key; channel IDs and Read/Write API Keys must remain in the later dashboard settings.
  3. Enter an invalid home Wi-Fi password before the first successful setup and wait through at least two reconnect cycles. The setup access point must remain available so the credentials can be corrected.
  4. Save valid home Wi-Fi credentials. Confirm that the setup AP switches off after connection and that http://w-charger.local/ or the LAN IP reaches the station.
  5. Set a website password of at least eight characters. Verify that dashboard, configuration, history, Wi-Fi scan and ThingSpeak APIs require login. Check that CSRF or cross-origin state-changing requests are rejected and that the dashboard, API and Wi-Fi-scan responses include the configured no-cache, frame, MIME-sniffing and content-security headers.
  6. Log in from two clients. Five failed attempts from one client must trigger a 30-second block without blocking the other client. Signing out one client must invalidate only its server-side session; the other session must remain valid. Confirm that an idle session expires after 30 minutes.
  7. Restart the station and confirm that settings and local history persist, while all previous session cookies no longer authenticate.

2. Sensor PCB V3

  1. Upload sensor_pcb_v3. An unprovisioned sensor must find the station without a compiled-in station MAC address or radio channel. Verify this once during the fast pairing window and once after it has expired; every discovery wake must cover all 13 channels.
  2. Confirm that the sensor advertises every ten seconds during the initial ten-minute fast-pairing window, but refreshes its sensor/battery snapshot only every five minutes. Both the captive setup page and the dashboard's pending-sensor list must show Setup mode.
  3. Compare dashboard temperature and humidity with reference instruments.
  4. Measure battery voltage simultaneously with a multimeter. The current V3 firmware intentionally uses the field-validated base scaling factor 1.67, which takes precedence over a nominal schematic ratio. Record the residual error before applying any per-sensor calibration.
  5. Set the interval to one minute in the web interface. Confirm that appliedRevision matches the station revision no later than the second subsequent contact and that the dashboard changes to Energy-saving mode only after the sensor reports that mode.
  6. Switch the station off during a transmission. The sensor must still power down safely, enter deep sleep and retry only at the next configured report interval.
  7. Add a sensor while the station is still using its captive-portal setup AP, then finish Wi-Fi setup on a different channel. At the next due report, the sensor must try its saved setup channel, scan all remaining channels in the same wake, receive the station response and persist the new channel. Change the router channel again after this transition; steady-state recovery must then use at most three rotating channels per scheduled exchange. During a received recovery packet, the dashboard must show Channel recovery and return to Energy-saving mode after the next normal report.
  8. Queue a sensor factory reset, interrupt the station during its response and verify that the command remains pending until the sensor acknowledges the matching revision.
  9. Delete a configured sensor from the station without resetting its power. At its next scheduled known-channel check-in, the station must return provisioned=false. The sensor must clear the station identity, wake again after one second and reappear as an unprovisioned sensor using a complete 13-channel discovery scan.
  10. Upload a new sensor firmware image without erasing NVS manually. The ELF SHA change must clear the old station assignment once; another restart with the same image must not clear it again.

3. BME680 and IAQ on PCB V3

  1. Connect the BME680 to the switched I2C supply, restart the sensor and verify that the add-sensor wizard detects BME680. Then select BME680 explicitly and initially use a five-minute interval.
  2. Check the serial log and timing: BSEC must use only ULP/300 s from the first start. There must be no LP/three-second cycle and no later sample-rate switch. Permanent direct raw measurements indicate a fault.
  3. The dashboard must show temperature, humidity, pressure, gas resistance and, after BSEC learning, Static IAQ. While accuracy is 0, the dashboard must show the real elapsed learning time without a fixed-duration progress bar, and ThingSpeak must not receive IAQ. When accuracy first reaches 1, the large learning notice must disappear permanently and the compact Accuracy badge must remain visible during normal operation.
  4. Operate the sensor in a normally used room for at least 24 hours and observe whether accuracy progresses through 1, 2 and eventually 3. Ventilation should improve IAQ with a delay, while ordinary indoor pollution should worsen it. Do not expose the sensor directly to liquids or concentrated vapours.
  5. After reaching accuracy 2 or 3, disconnect the battery briefly. A saved BSEC state must load after restart; restarting the complete long-term learning process indicates a persistence fault.
  6. After thermal equilibrium, compare temperature and humidity with a known reference instrument. Change the temperature offset only in small steps and evaluate at least three further cycles.
  7. Set the reporting interval to 30 minutes. The BME680 must still measure every five minutes, while ESP-NOW and ThingSpeak operate only every 30 minutes. Include several seconds of awake processing in the test and verify that the logical report clock does not postpone transmission by an extra five-minute cycle.
  8. Alternate between BME680 and BME280 and update the station selection. An incorrect explicit type must appear as a read failure/type mismatch; Automatic must recognize both chip IDs at 0x76 and 0x77.
  9. Measure the sensor rail directly during deep sleep on both PCB V3 and V4; it must be 0 V. After at least ten fully power-cycled sensor measurements, IAQ learning must continue instead of starting with a fresh algorithm state on every wake-up.
  10. Reset the IAQ learning state in the wizard. After the next contact, accuracy must intentionally restart at 0 and then begin increasing again.
  11. Force one BSEC cycle to provide only a direct Bosch raw fallback. The last genuine IAQ may remain visible as stale, but it must not create a new history point or ThingSpeak upload.

4. ThingSpeak

  1. Store only the User API Key during initial setup. In the dashboard, load the account channels and create a private channel for a sensor.
  2. Confirm that channel ID, channel profile and write key remain available after a station restart.
  3. Map every supported quantity to a unique field and verify its unit and destination. Battery voltage is uploaded in volts, gas resistance in ohms and pressure in hPa. Fields configured as Do not upload must remain empty.
  4. Confirm that an HTTP 200 response without a positive entry ID is shown as a failure rather than Connected.
  5. Disconnect the internet for several measurements and restore it. Observe droppedJobs; the current RAM queue is not persistent across restarts and does not guarantee delivery of offline samples.
  6. Verify that unprovisioned discovery telemetry and stale BSEC fallback IAQ values are never uploaded.
  7. After a successful upload, confirm that the overview status remains Connected while the redundant success message is hidden. A later upload or station-connection error must make its warning visible again.

5. Local station history

  1. Set one BME280 sensor to a 10-minute measurement interval. Confirm that the dashboard and /api/history report 10-minute steps while retaining at most the rolling last 24 hours.
  2. Change the same sensor to 20 minutes. Existing recent history should remain, be rebucketed to 20-minute steps and survive a station restart.
  3. After collecting several 10-minute points, change the interval to two minutes. The existing line must remain plotted while new two-minute points are appended; cursor inspection alone is not sufficient.
  4. Repeat with a second sensor at a different interval and confirm that each chart keeps its own step size. Missing measurements must create a visible line gap instead of being connected as continuous data.
  5. With website protection enabled, confirm that the protected-access banner disappears automatically after its initial display. The unprotected warning must remain visible when no password is set.

6. PCB V3 energy profile

Before measuring current, verify the configured-mode radio policy:

  1. Add a BME280 sensor, wait for the station to acknowledge provisioning and select a ten-minute interval. Across three successful cycles, verify that it wakes at the configured interval, sends unicast on the saved channel and emits neither discovery beacons nor scans on other channels.
  2. Power-cycle that sensor without changing its firmware. The persisted assignment must keep it in the same energy-saving mode; it must not return to ten-second discovery.
  3. Repeat with a reporting interval longer than five minutes on a BME680. The node may wake every five minutes for BSEC ULP, but ESP-NOW must remain off on intermediate wakes and transmit only when the configured report is due.
  4. Delete the sensor or complete its factory reset and verify the opposite: ten-second discovery and complete 13-channel scans must resume.

Use a power analyser to measure at least 20 cycles of each case:

  • wake-up to sensor power-on
  • BME280 measurement duration
  • BME680 ULP measurement including heater phase
  • ESP-NOW transmission on a known channel
  • BME680 intermediate cycle without ESP-NOW at a longer reporting interval
  • stale-channel recovery scan as the radio worst case
  • deep-sleep current
  • charge consumed by a normal measurement cycle
  • discovery mode during both the ten-second and five-minute phases

Derive runtime only from measured charge per cycle, the battery's measured usable capacity and self-discharge. Runtime estimates in documentation are not acceptance values.

7. Battery-voltage calibration

  1. With USB disconnected, measure the battery simultaneously using a calibrated multimeter and the dashboard.
  2. Enter the multimeter value in the sensor wizard and save it. The new factor must arrive at the next sensor wake-up.
  3. Compare again at at least three voltages, for example 4.15 V, 3.70 V and 3.25 V. One-point calibration may correct gain but must not conceal strong non-linearity or an incorrectly populated resistor network.
  4. Confirm that the calculated correction factor is rejected outside the permitted range of 0.7 to 1.3.
  5. Reset calibration to its default and verify that correction factor 1.0000 is active again on top of the board-specific base scaling.

8. Moving to PCB V4

  1. Before assembly, continuity-check GPIO6/ADC_EN and GPIO10/SENSOR_PWR_EN against both schematic and PCB.
  2. During sleep, GPIO6 must be low and the battery divider must draw no measurable continuous current.
  3. GPIO10 is active low on V4: low enables the PMOS sensor rail and high disables it.
  4. Compare battery voltage near 4.2 V, 3.7 V and 3.2 V. V4 uses 100 kOhm over 150 kOhm and a base scaling factor of 1.667.
  5. Repeat BME280, BME680, discovery, stale-channel recovery and factory-reset tests with the sensor_pcb_v4 target.
  6. Perform brownout, USB connect/disconnect and charger/battery switchover tests.
  7. Use the V4 environment for longer field tests only after pin, voltage, polarity and current tests have passed.