Repository files navigation

Documentation about Huawei 48V power supplies

• Review with lots of information (pinout, CAN, voltage range, etc.):https://www.beyondlogic.org/review-huawei-r4850g2-power-supply-53-5vdc-3kw/
• List of other related power supplies:https://www.ycict.net/de/products/huawei-r48100g1-rectifier-module/
• PHZ-F020304-BWxxx connector datasheets:https://www.medlonchina.com/FCI-connector-DC-120A-AC-40A.html

Should also fit R4830G2, R4850G2 and R4850N2 as these are all 40.8x105x281mm.

Is modeled with a bit of clearance so it can be used as a negative directly and fit when 3D printed.

Might also be compatible with other connector manufacturers if they have the same dimensions.

Is modeledwithout clearance so it should be expanded a bit to use as a negative.

Fits the PSU card-edge when positioned at the bottom back of the PSU extrusion/profile (translate([0,-frontDepth - profileLength,-profileHeight/2])).

Parts to mount a Huawei R4830S1 PSU and PHZ-F020304-BW001 connector to a plate/board.

Mounts to the board using 4mm countersink screws, the connector is mounted to the part with M3 screws.

The PSU can slide in from the front and is held in place using the original metal handle/clip on the front of the PSU.

Also onPrintables.

(Photo is of an old version with less space for the screwdriver and with steeper angles which was more fragile)

There seem to be a handful of connectors that are compatible with these power supplies.I have not yet done an exact comparison between the different connectors / PSU card edges, but likely some of these connectors and card edges are cross-compatible with each other as they look at least similar.

Fits at least R4850G2

My R4830S1 came with one, so it does definitely fit that.

looking at the connector at the back of the PSU, the connections (left to right are):

• DC negative (not earth referenced)
• (DC Pre-charge)
  • connected to DC negative internally
  • likely used to pre-charge internal capacitors via a resistor when plugging in module with battery voltage present
• DC positive (not earth referenced)
• Slot detect (top/bottom)
  • can both be pulled to DC negative to enable PSU
  • resistor signaling used for addressing PSU by its slot
• CAN (top/bottom)
  • top/white = CAN_L
  • bottom/black = CAN_H
• PE
• N
• L

The PSU output capacitors arelarge.If you connect the unpowered PSU up to a battery,therewill be sparks.To avoid this, the PSU should either be on already or a pre-charge resistor (possibly the internal one?) should be used.

The PSU's internal fan control method uses theinput/ambient temperature to determine the fan speed.This means, even when under heavy load, the fan speed increases only marginally because theinternal/PCB temperature doesn't really affect the input/ambient temperature sensor.
This lets the PSU runreally hot (>80°C at full load).

There are CAN commands to set the fan mode (auto / full speed) and to set the fan duty cycle (0-100%).
If running at anywhere near full power (>50% maybe) the fan speed should either be set to 100% or temperaturecontrolled via the duty cycle (if noise matters).

If you just want to control the PSU without implementing all the commands yourself, you can use myfork of an existing ESPHome integration (either directly or as a reference implementation):patagonaa/esphome-huawei-r4850

The CAN protocol is almost the same between all PSUs of the R48xx-series and apparently (with other protocol IDs)is also used in other PSUs (like the R100030G with 1000V 30A, used for EV charging, for which there are docs in/docs)and possibly even different power electronics products (there are hints pointing at the C28005G1 48V to 280V 5ADC-DC converter and MPPT solar chargers).

This documentation is a summary of own tests, documentation of other products and a bunch of other peoples'implementations and documentation (seeSources).
Thus, this is incomplete and possibly not 100% correct. I did, however test and verify almost all of the thingsdocumented here, with both an R4830S1 and R4850G6, so I'm reasonably sure itis correct.

The CAN interface uses 125kbps rate with extended 29-bit identifiers.

The CAN message IDs do not specify a single value/register.Instead, the message ID is used to encode the prococol, PSU address,command, message direction, etc.

Example:1081407F

Interpretation:
Bits:000a aaaa abbb bbbb cccc cccc deee eefg

• 0 (bit 31-29): always zero (CAN ID is 29-bit)
• a (bit 28-23): protocol ID (always21)
• b (bit 22-16): address (0 = broadcast, 1 = first, ...)
• c (bit 15-8): command id
• d (bit 7): message source (0 = from PSU, 1 = to PSU)
• e (bit 6-2): group mask (always1F)
• f (bit 1): hardware / software address (0 = hw, 1 = sw, always 1)
• g (bit 0): finished marker (0 = finished, 1 = more data coming)

Apparently the PSU can have a "hardware address" and "software address"but even the original "SMU02B" controller seems to use only the software address.

The "software address" seems to be negotiated automatically if multiple PSUs are on one CAN bus (starting at 1).

The "hardware address" might be fixed per slot, because the original PSU rack has a network of resistorsand dip switches on the two "slot detect" pins of the connector. Possibly, this allows the PSU to knowwhich slot it is in, which might set the PSU "hardware address", but I haven't tested this.

Requests a data response composed of multiple status messages (including one register each).

Example (to PSU):108140FE: 00 00 00 00 00 00 00 00

Example (from PSU):

1081407F: 01 0E 00 00 00 00 2A 011081407F: 01 70 00 00 00 1C DC 311081407F: 01 71 00 00 00 00 C7 F51081407F: 01 72 00 00 00 00 20 911081407F: 01 73 00 00 00 1C 20 9A1081407F: 01 74 00 00 00 00 03 E61081407F: 01 75 00 00 00 00 CD B11081407F: 01 76 00 00 00 00 04 001081407F: 01 78 00 00 00 03 8B 801081407F: 01 7F 00 00 00 00 84 001081407F: 01 80 00 00 00 00 6C 001081407F: 01 81 00 00 00 00 8C 111081407F: 01 82 00 00 00 00 8C 071081407E: 01 83 00 00 10 00 00 00

Data bytes:

• Byte 0-1: register id
• Bytes 2-3: ? (always 0)
• Bytes 4-7: int32 value

Register values:

* Not necessarily the set limit, rather what the PSU is currentlycapable of outputting, seeOutput current limit

Requests an info response composed of multiple messages (including one register each).

Example (to PSU):108150FE: 00 00 00 00 00 00 00 00

Example (from PSU):

1081507F: 00 01 00 00 40 46 12 671081507F: 00 02 64 46 85 50 02 AF1081507F: 00 03 32 31 30 32 33 311081507F: 00 04 31 54 52 52 4C 551081507F: 00 05 05 00 01 0D 01 0D1081507E: 00 06 01 01 00 00 00 00

Data bytes:

• Byte 0-1: register id
• Bytes 2-7: data

Registers:

Requests an "E-Label" response composed of multiple status messages (including one part of the ASCII response each).

Example (to PSU):1081D2FE: 00 00 00 00 00 00 00 00

Example (from PSU):

1081D27F: 00 01 2F 24 5B 41 72 631081D27F: 00 02 68 69 76 65 73 491081D27F: 00 03 6E 66 6F 20 56 651081D27F: 00 04 72 73 69 6F 6E 5D[...]1081D27F: 00 33 3D 30 30 0D 0A 431081D27F: 00 34 4C 45 49 43 6F 641081D27F: 00 35 65 3D 0D 0A 42 4F1081D27E: 00 36 4D 3D 0D 0A 00 00

• Byte 0-1: Message part number
• Bytes 2-7: ASCII Data (null-terminated)

Decodes to:

/$[ArchivesInfo Version]/$ArchivesInfoVersion=3.0[Board Properties]BoardType=EN1MRC3S1A1BarCode=2102311TRRLUL4000687Item=02311TRRDescription=Function Module,R4830S1,EN1MRC3S1A1,1U2000W Super High Efficiency Rectifier,DS SpecialManufactured=2020-05-06VendorName=HuaweiIssueNumber=00CLEICode=BOM=

Sets the value of a register.

Example (to PSU):108180FE: 01 34 00 01 00 00 00 00

Data bytes:

• Byte 0-1: register id
• Byte 2-7: data

Registers:

* SeeOutput current limit
** Can only be set to 0 or above the min. duty cycle (seeRegister Get Response). (other values return an error and reset the internal value to 0 (auto)).
On the R4850G6, the min. duty depends on the temperature
On the R4830S1, the min. duty is the current setpoint (which is probably a bug), so the duty cycle can only be increased or set to 0.

Response to setting a register value.Has an error status field in case the value is out of range or the parameter doesn't exist.

Example (from PSU):1081807E: 01 34 00 01 00 00 00 00

Data bytes:

• Byte 0 (low nibble): status (0 = ok,2 = parameter error)
• Byte 0 (high nibble) - 7: copied from request

Gets the value from a register.
Seems to work only for status registers (01 70, ...) not for config registers (01 00, ...).

Example (to PSU):108182FE: 01 34 00 00 00 00 00 00

Data bytes:

• Byte 0-1: register id
• Byte 2-7: zero

Response to getting a register value.Has an error status field in case the parameter doesn't exist or can't be read.

Example (to PSU):1081827E: 01 34 00 01 00 00 00 00

Data bytes:

• Byte 0 (low nibble): status (0 = ok,2 = parameter error)
• Byte 0 (high nibble) - 1: register id
• Byte 2-7: register value

Registers (in addition to ones from40 data response);

* On the R4850G6, the min. duty cycle is based on the current temperature (seeFan control).
On the R4830S1, the min. duty cycle is always equal to the duty cycle target.
** On both PSUs, the duty cycle target is the greater of both the temperature-based duty cycle and the duty cycle set via CAN.

Sent from the PSU unsolicited (without requesting anything).
Due to no documentation being available on any of this, the following is mostly speculation / own findings.

Example (from PSU):

1001117E: 00 01 00 00 00 00 04 97100011FE: 00 02 00 00 00 00 03 FE108111FE: 00 03 00 00 00 01 00 00

Messages:

Due to the different message directions and addresses, and due to the load/current being included,I think these are used for both address negotiation and automatic load sharing between multiple PSUsconnected to the same CAN bus.

Due to the messages including the current and being in regular intervals, they might (or at least could)also be used as a coulomb counter.

Data bytes:

• Byte 0-2: register id (?)
• Bytes 2-7: register values (?)

* The value in00 01 seems to update faster than the one in00 03. For calculatung current from this, seeOutput current limit
** ready: PSU is ready to output voltage/power (AC input available, not faulted, ...)
*** active: device is outputting voltage/power (not booting, not standby, ...)

The CAN communication is powered by the DC side.This means CAN communication is always possible when there is a battery connected, even if there is no AC input present.

For some values (mostly voltage and current) there is a "default" value (which is saved in non-volatile memory)in addition to the "running" value.
This means, the voltage/current will be reset to these default settings after power lossor CAN communication loss (after ~60s timeout, shown by flashing yellow light on front panel)and has to be set again once CAN communication is (re-)established.

Most CAN values have a multiplier of 1024 (50V * 1024 = 51200 = 0xC800).
When setting parameters, the PSU reports an error when the value exceeds the valid range.

Valid ranges (tested with R4830S1 and R4850G6)

• Voltage: 41.0V (A4 00) to 58.6V (EA 67)
• Default Voltage: 48.0V (C0 00) to 58.4V (E9 9A)
• Current: 0% (00 00) to 100% (04 E2)
  • Weirdly enough, this value goes up to 1250 instead of 1024, but 1250 coincidespretty well with the maximum current from the graph in the datasheet (seeOutput current limit).
• Fan duty cycle:
  • R4830S1: can be set to auto (00 00) or between 30% (1E 00) and 100% (64 00)
  • R4850G6: can be set to auto (00 00) or betweenmin. duty cycle (temperature dependent) and 100% (64 00)

The output current limit is set as a ratio of the maximum possible output current.
For the R4830S1, this means 1250 ≈ 42.6A, for the R4850G6 1250 ≈ 63.3A(both compared to the PSU's internal current measurment).

This means, to set a current limit of 10A for the R4830S1,the register should be set to: 10A / 42.6A * 1250 ≈ 293 =00 00 01 25.

The max output current readout (register01 76) uses the same scale but does not necessarily equal the limitthat has beenset, but rather specifies what the PSU is currently set toand capable of outputting.
This includes the output current limit (of course), but also the AC current limitand possibly also the AC voltage, DC voltage, temperature derating, etc.

Examples (R4850G6):

This information was put together from several sources, such as:

• own CAN dumping/testing
• https://github.com/craigpeacock/Huawei_R4850G2_CAN
• https://endless-sphere.com/sphere/threads/rectifier-huawei-r4850g2-48v-42-58v-3000w.86038/post-1809287
• https://endless-sphere.com/sphere/threads/rectifier-huawei-r4850g2-48v-42-58v-3000w.86038/post-1805865
• https://endless-sphere.com/sphere/threads/rectifier-huawei-r4850g2-48v-42-58v-3000w.86038/post-1807301
• https://github.com/BotoX/huawei-r48xx-esp32
• https://github.com/577fkj/PowerControl/blob/c6d06935af79fcdbf3e71d1ad7ea4a86cd0c756f/main/src/protocol/huawei_r48xx.c
• https://max.book118.com/html/2022/0414/5230000111004213.shtm
  • seedocs/
• https://tieba.baidu.com/p/7475319470