The DevKitC-1 swapped the RGB LED between V1.0 (GPIO 48) and V1.1 (GPIO 38),
and the freed pin is the one PVbrain2 uses for CAN TX. Both pins are therefore
board-revision dependent and cannot stay hardcoded in a single board file.
board_ESP32-S3_DevKitC-1_PVbrain2.yaml becomes the common base :
- tx_pin_4 / rx_pin_4 and the LED strip pin are removed from it
- esp_light now reads ${rgb_led_pin}
- GPLv3 header added, which was missing
Two thin wrappers provide the revision-specific pins and include the base :
- board_ESP32-S3_DevKitC-1_PVbrain2_LED-48.yaml : V1.0, LED 48 / CAN TX 38
- board_ESP32-S3_DevKitC-1_PVbrain2_LED-38.yaml : V1.1, LED 38 / CAN TX 48
YamBMS_LP_PVbrain2.yaml and YamBMS_RP_PVbrain2.yaml include the LED-48 variant,
the LED-38 line being commented right next to it.
Same split applied to the RS485 Pylontech package : the content moves to
yambms_rs485_pylon_core.yaml and yambms_rs485_pylon.yaml keeps only the uart
baud_rate extend, so PVbrain2 — whose board already declares the UART — can
include the core directly.
PCB LED 1 / PCB LED 2 / ESP32 LED are no longer prefixed by ${name}, and the
`name` substitution is added to the PVbrain2 root YAMLs.
The installed total was accumulated once per BMS per boot, latched by
bms<id>_combined_once, so a BMS reconfigured with a different capacity
kept contributing its boot value until the next reboot.
Each BMS now remembers the last capacity it contributed in
bms<id>_installed_capacity and applies only the difference to the total.
- the update runs in the can_be_combined branch, so an offline BMS simply
stops updating its contribution, which stays in the total : the installed
capacity still never shrinks on an uncombine, as before
- a BMS coming back with a different capacity is applied as a delta, with
no double counting
- combined_once had no other reader and is removed
The total is deliberately not turned into a per-round accumulator reset in
the Global vars reset block like the other totals : its sensor is polled
independently of STEP 2 and its state is written straight into CAN frame
0x379 with no has_state() guard, so a transient zero would announce 0 Ah of
installed capacity to the inverter. See Design_decisions.md.
board_options_led_control.yaml called set_brightness() unconditionally, so
every board whose status LED is a `binary` light logged
"'esp_light': brightness not supported" on each state change. Only the RGB
setter was guarded, by the per-board `board_has_rgb` substitution.
Both setters are now guarded by what the light itself reports:
led->get_traits().supports_color_capability(...)
- `board_has_rgb` removed from 21 board files: its only two readers were the
two lines in board_options_led_control.yaml
- the runtime test reproduces the flag exactly on every board - the 15 boards
declaring "true" all drive an addressable strip (esp32_rmt_led_strip, or
spi_led_strip on T-Connect), the 6 declaring "false" all drive a binary
light - so RGB rendering is unchanged and only the warning disappears
- a monochromatic LED now gets brightness without RGB, which a single boolean
could not express
Also migrated 16 `rgb_order: GRB` to `channel_colors: GRB`, deprecated in
ESPHome and removed in 2027.3.0. Same value, no behaviour change.
A CRLF file rewritten by a Windows contributor shows up as a whole-file
diff, which hides the real change and conflicts on every rebase.
.gitattributes was itself listed in .gitignore, so it could never be
committed and protected nothing.
http_request runs on the loop task, which has a 4s budget before the
5s task WDT panics. A dropped connect() blocks for 4.5s (the ESPHome
default timeout) and a filtered port 53 blocks for a fixed 7s, which
timeout does not cover.
- timeout 3s and watchdog_timeout 60s cover the two blocking paths
- a failure counter stops the service after 3 network-level failures,
instead of freezing the loop 14s every 6h forever
- the POST is now skipped when the GET failed
Reproduced and measured on OPNsense.
See Design_decisions.md.
The detector compared var_to_combine_bms/shunt_number with its value on
the previous round, but samples it before the STEP 2 reset - so in a
healthy setup the sampled value is identical every round and was read as
a stall ("Shunt id 2 of 1 declared").
Replaced by an exact completeness test: a chain is complete once its
number is past the number of declared devices. Threshold raised to 5
rounds to absorb interval phase offsets at boot.
* Update EG4 LL-V2 component to v1.1.0
FixesRAR/esphome-eg4-bms#7: some EG4 LL-V2 units intermittently answer
with a CRC-valid function-03 response whose register data area is
entirely zero. The component published those zeroes as a real reading,
collapsing SoC, pack voltage and capacity for one poll cycle.
v1.1.0 rejects implausible blocks and no longer lets them refresh the
online tracker, so a unit stuck returning zeroes now goes offline
instead of reporting a flat-zero battery.
No YamBMS-side changes are required: neither EG4 package uses the
removed designed_capacity sensor, and both already use the
cell_average_voltage key.
Rewrote Auto CVL from template sensor lambdas to interval
lambdas, splitting each into a value sensor and a delta sensor.
Fixed a bug confirmed in production: Auto CVL's Float floor coincided
exactly with the Rebulk V. threshold, letting it trigger its own rebulk.
The floor is now capped relative to the live target instead of a fixed
chemistry constant, the Float/Stop integral term can no longer get stuck
once clamped, and PI state now resets when crossing the Bulk <-> Float/Stop
boundary.
This YAML file implements the Auto CCL Current Taper feature for the YamBMS project, defining various parameters and logic for charge current tapering during battery charging.