FAQ
TL;DR: If you have a Della mini split with a WBR3 module, you can flash OpenBeken locally, back up the original 2 MB flash first, and then control cooling, heat, fan, swing, buzzer, display, and even GEN mode. As one developer put it, "It worked!" This FAQ is for owners who want cloud-free control without losing core AC functions. [#21551100]
Why it matters: This thread shows that some Della WiFi modules are not standard TuyaMCU devices, so using the correct Realtek tools and the TCL driver is the difference between a working local AC and a dead-end flash attempt.
| Option |
Chip family |
Fit/use in thread |
Main caveat |
| WBR3 |
Realtek RTL8720CF/AmebaZ2 |
Native Della module; fully tested with OBK TCL driver |
Fragile pads and traces during wiring |
| WBR1 on TCLWBR |
Realtek RTL8720CF |
Also worked after correct VIN/VCC power wiring |
Needed both 5 V and 3.3 V path active |
| WB3S |
Beken |
Considered compatible replacement for TuyaMCU-style UART layout |
Must verify 3.3 V, GND, reset, and UART pin match |
| CB3S |
Beken |
Considered compatible replacement with OBK support |
Same pinout checks as WB3S |
Key insight: The Della TCWBRCU1/WBR3 setup in this thread was ultimately treated as a custom TCL serial protocol, not a normal TuyaMCU datapoint bridge. That is why flashing succeeded only after using Realtek-specific tools and why control required the OpenBeken TCL driver, not GPIO templates. [#21549462]
Quick Facts
- The confirmed stock backup sizes were 384 KB ROM and 2 MB flash, and the 2 MB dump was verified to boot before flashing OpenBeken. [#21544743]
- The working UART settings for the TCL protocol were 9600 baud, 8 data bits, even parity, 1 stop bit (9600 8E1); parity was a real failure point until fixed. [#21550462]
- The Della unit discussed was a 9,000 BTU heat/cool mini split on 115 V, 60 Hz power, using a removable USB WiFi/Bluetooth module with a WBR3 soldered to a TCWBRCU1 carrier. [#21536655]
- Internal target temperature in the TCL protocol is Celsius only, with a usable command range of 16–31 °C; entering 75 or 77 as if they were Fahrenheit produced wrong results such as 81 °F and 84 °F on the indoor display. [#21553236]
- Reverse-engineered GEN mode testing showed approximate winter heating limits of 1.2 kW at L1, 1.8 kW at L2, 2.3 kW at L3, and 2.9 kW with no GEN limit, based on clamp monitoring. [#21781377]
How do I flash a Della mini split WBR3 WiFi module on the TCWBRCU1 board with OpenBeken step by step?
Use Realtek tools, not BK7231 tools. 1. Back up the stock WBR3 first with the Realtek workflow, because the verified flash dump was 2 MB and recoverable. 2. Flash an OpenRTL87X0C build that includes the TCL driver or update later by OTA. 3. Start the driver with
startDriver TCL, then test commands such as
ACMode 1,
FANMode 3, and
TargetTemperature 24. The thread showed this sequence working on the Della 048-TP-9K2V-23S-IN after successful backup and flash.
[#21547316]
What is TuyaMCU, and how is it different from the custom TCL serial protocol used by some Della mini split WiFi modules?
"TuyaMCU is a host-MCU protocol layer that lets a WiFi module exchange standardized datapoints with a separate appliance controller, usually over UART." In this thread, that turned out not to be the whole story for the Della mini split. The WBR3 could run OpenBeken, but the AC control path matched a custom TCL serial protocol instead of normal TuyaMCU DPID traffic, which explained why TuyaMCU commands did nothing while the TCL driver worked.
[#21549462]
Why does a Della WBR3 module get stuck on "getting bus" in BK7231 tools, and which flashing tools should I use for Realtek RTL8720CF or AmebaZ2 instead?
It gets stuck because BK7231 tools do not support Realtek AmebaZ2 chips like the WBR3. One helper stated plainly that Easy Flasher does not back up or write Realtek chips. The working toolchain in the thread was
ltchiptool for backup and
AmebaZ2 PG Tool for writing on RTL8720CF/AmebaZ2 hardware, which immediately solved the stalled "getting bus" problem.
[#21544726]
What is a DPID in the Tuya ecosystem, and how were the Della mini split datapoints like power, mode, windspeed, and swing identified in this thread?
"DPID is a Tuya datapoint identifier that maps a cloud-visible property to a device function, type, and value range." The thread identified the Della AC datapoints from the Tuya IoT Platform and later from the product data model. Examples included
DPID 1 power,
4 mode,
5 windspeed,
113 vertical swing,
114 horizontal swing, and
120 GEN mode, each with enum or value ranges shown by the Tuya schema.
[#21536711]
How can I back up the stock firmware from a WBR3 or WBR1 module before flashing OpenBeken?
Back it up with the Realtek toolchain before you flash anything. The proven path was to read both the ROM and full flash, then verify the sizes: the ROM should be
384 KB and the flash dump
2 MB on the tested WBR3. That backup later booted successfully, which gave the thread a recovery path when hardware pads were lifted during rework.
[#21544743]
Which OpenBeken commands were used to control the Della mini split after flashing, including ACMode, FANMode, SwingH, SwingV, TargetTemperature, Buzzer, Display, and Gen?
The working commands were
startDriver TCL,
ACMode,
FANMode,
SwingH,
SwingV,
TargetTemperature,
Buzzer,
Display, and later
Gen. Confirmed examples included
ACMode 1 for cool,
ACMode 3 for fan-only,
TargetTemperature 24,
Buzzer 0,
Display 0, and the test PR command
Gen 0 through
Gen 3. Those commands were validated progressively in the thread and then exposed to Home Assistant through MQTT discovery.
[#21761114]
Why did TargetTemperature seem wrong when entering Fahrenheit values in OpenBeken, and how does the Della TCL protocol actually handle temperature internally?
It seemed wrong because OpenBeken expected Celsius while the indoor display was showing Fahrenheit. The TCL protocol stores target temperature as a
4-bit value for 16–31 °C, so entering
TargetTemperature 75 was interpreted as a Celsius-side command, wrapped internally, and produced values like
29 °C, which appeared as about
84 °F on the unit. Once the user switched to
TargetTemperature 24 and
25, the display correctly showed
75 °F and
77 °F.
[#21553392]
What was the correct UART configuration for the Della TCL mini split protocol on Realtek, including baud rate and parity, and why did parity matter?
The correct serial format was
9600 8E1: 9600 baud, 8 data bits, even parity, 1 stop bit. Parity mattered because the first TCL driver builds used the wrong parity setting, so commands were sent but not understood correctly by the AC controller. Once even parity was fixed, the unit started beeping on valid commands and mode and fan tests began to work immediately.
[#21551100]
How do Home Assistant MQTT discovery and the OpenBeken TCL driver work together for a Della mini split, and what commands or settings are needed to make the climate entity appear?
Start the TCL driver first, then trigger Home Assistant discovery. The thread used
startDriver TCL and then either Home Assistant discovery from the UI or the command
scheduleHADiscovery 1. After that, Home Assistant created a climate entity and additional controls, and later builds added mode, fan, horizontal swing, and vertical swing support to the discovered entity.
[#21561364]
What troubleshooting steps helped when Home Assistant showed Della mini split state updates but power-on or mode changes did not work correctly?
The main fixes were driver-side, not MQTT-side. Early on, Home Assistant could show state changes from OBK, but power-on and mode commands were incomplete because the TCL mode parse and mode publishing were not fully implemented. Once the driver parsed AC mode correctly, Home Assistant could both read OBK state and write changes back, including on/off and mode switching.
[#21563089]
WB3S vs CB3S vs WBR3 for a Della mini split replacement module — which options were considered compatible with OpenBeken and what pinout checks matter most?
WBR3 was the native tested module, but WB3S and CB3S were both considered workable OpenBeken replacements. The key check was not brand alone; it was whether
GND, 3.3 V, reset, and UART matched the original footprint and host board expectations. The thread specifically noted that other GPIO positions may differ, but the basic UART-based control path could still work if those core pins lined up.
[#21548807]
How can I capture UART traffic between the WiFi module and the indoor unit MCU on a Della mini split without a full logic analyzer?
You can capture each UART lane separately with a USB-UART adapter if you match the serial settings. One developer explained that you do not need a full logic analyzer to start; you can connect RX to the line you want to observe, set the correct baud, and record module-to-MCU and MCU-to-module traffic independently. For this TCL setup, correct serial settings mattered, especially parity.
[#21550113]
What caused the WBR1 TCLWBR board to crash when starting UART-based OpenBeken drivers, and how was the missing VCC or VIN power connection eventually identified?
The crash came from powering the WBR1 incorrectly during bench testing. The user had supplied only
3.3 V for flashing-style setup, but the TCLWBR adapter also needed the
VIN/VCC path active from 5 V so the board’s power scheme matched normal operation. Once they powered the adapter board correctly instead of only the WBR1 core rail, the TCL driver started normally and the module worked.
[#21579198]
Where can I find replacement Della-compatible WiFi modules like WBR3, WBR1, or Pioneer USB adapter boards if I damage the original during flashing?
One confirmed source was Pioneer’s replacement USB WiFi/BLE dongle for WYT/WT systems, which a later poster said used the same style of module and compatible firmware family. Earlier in the thread, users also sourced WBR3 replacements from AliExpress after damaging traces and pads during desoldering. The practical lesson was to buy a spare before risky rework, because these module pads are fragile.
[#21766471]
How was the GEN or generator mode on supported Della mini splits reverse engineered, and what do the L1, L2, and L3 power limits correspond to in practice?
GEN mode was inferred from the Tuya schema, then tested in a custom OpenBeken PR with
Gen 0 to
Gen 3, and finally validated by watching real power draw. Practical results on one unit were about
1.2 kW for L1,
1.8 kW for L2,
2.3 kW for L3, and
2.9 kW with no GEN limit during cold-weather heating. The user mapped L1, L2, and L3 to roughly
30%,
50%, and
80% compressor power limits.
[#21762524]
Generated by the language model.
