Fix GPS on ThinkNode M3: power rail and reset polarity#3009
Conversation
The M3 has a serial CASIC AT6558R GNSS receiver on Serial1, but it never emitted a single NMEA sentence, so gps_detected stayed false and the app never offered the location setting. Two independent problems kept it mute: The module is fed through a load switch on PIN_GPS_POWER (pin 14, active HIGH). The pin was declared in variant.h but nothing ever drove it, so the receiver was permanently unpowered. Drive it through a RefCountedDigitalPin owned by the board and hand it to MicroNMEALocationProvider, matching what the heltec variants already do. Its reset line is asserted HIGH. variant.h defined GPS_RESET_ACTIVE, a name nothing reads; MicroNMEALocationProvider looks at PIN_GPS_RESET_ACTIVE and fell back to its LOW default, so the state it drove to "release" reset was in fact the asserted level and the module sat in reset forever. Rename the macro and set it to HIGH. Detection also replaces the fixed 1 s delay with a poll, since a receiver may take several seconds to say anything after power-on. This board answers in ~200 ms, but the fixed delay is a trap for slower modules. Verified on hardware: the receiver now identifies itself over NMEA (IC=AT6558R, SW=URANUS5) and tracks satellites with a valid UTC timestamp. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
setSettingValue("gps", ...) called start_gps()/stop_gps() unconditionally.
Those paths claim()/release() the GPS power rail through a ref-counted pin,
so a repeated "gps=1" pushed the claim count above 1 and a following single
"gps=0" then failed to actually power the module down. Only act on a real
state transition.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
thanks for your PR #2887 addresses the same issue and I was just waiting some input to merge the PR ... Think I'll do it anyways now event if I don't have my m3 (it stayed in France since january :() Don't know if I'd merge your PR, no offence there, but I would not let a Claude authored commit enter the repo |
|
Sorry, I missed the prior open PR. For the Claude authored commits: you mean I should remove the co-author in future contributions? Or do you prefer no LLM contribution? |
|
About LLM assistance, Not having co-authored PR is a minimum. We can't forbid you to get some help, but LLM have tendancy to rewrite lot of things, so try to restrict them to the problem you solve, and do a review work of everything they did. Ideal would be, take some inspiration from their solution and write your own code ... |
2 participants
Fix GPS on ThinkNode M3
The ThinkNode M3 ships with a serial CASIC AT6558R GNSS receiver (UART, 9600 baud, GPS+BeiDou), but GPS never worked on this variant because of two bugs. The receiver stayed completely silent, so
gps_detectedwas always false and the app never exposed the location section.Root cause
PIN_GPS_POWER(pin 14) is a load switch feeding the module, active HIGH, but it was declared and never driven — so the module was permanently unpowered → 0 bytes onSerial1.GPS_RESET_ACTIVE LOW, whileMicroNMEALocationProviderreadsPIN_GPS_RESET_ACTIVE(default LOW) — so the module was held in reset forever.Verified in isolation: with only fix (1) →
0 bytes on Serial1; with (1)+(2) →GPS detect: waited 199 ms, 7 bytes on Serial1+GPS detected.Changes
variants/thinknode_m3/variant.h—PIN_GPS_RESET_ACTIVE HIGH(replacing the unreadGPS_RESET_ACTIVE LOW).variants/thinknode_m3/ThinkNodeM3Board.{h,cpp}— drive the pin-14 load switch via aRefCountedDigitalPin periph_power(PIN_GPS_POWER, HIGH),begin()-ed at board init (same pattern as the Heltec variants).variants/thinknode_m3/target.cpp— pass&board.periph_poweras the location provider's power pin so enabling/disabling GPS actually cuts the rail.src/helpers/sensors/EnvironmentSensorManager.cpp— GPS detection now pollsSerial1up toGPS_DETECT_TIMEOUT_MS(default 10 s) instead of a single fixeddelay(1000), so modules that take a moment to emit their first sentence are still detected (the M3 needs ~200 ms in practice). Also guards thegpssetting toggle against repeated values:start_gps()/stop_gps()claim/release the ref-counted power pin, so a repeatedgps=1would push the claim count above 1 and a following singlegps=0would then fail to power the module down — now only a real state transition acts.Testing
Flashed on hardware (
ThinkNode_M3_companion_radio_ble). Boot log showsGPS detectedand the AT6558R banner, followed by a continuous, well-formed NMEA stream ($GNGGA,$GNRMC,$GPGSV, …), up to 15 satellites in view (10 GPS + 5 BeiDou). Confirmed a real position fix outdoors during a full day of field use. Toggling location off in the app powers the receiver down via the load switch (cold restart on re-enable), so battery is preserved when GPS is disabled.Refs #1864.