Replies: 33 comments 112 replies
-
|
Cool project—thanks for sharing! We are hacking someone else's CAN Bus here, so I, sadly, don’t have all the information I would like. However, you are absolutely right when you say that the power-up sequence "randomly" determines which r4875g gets assigned which "address". Given your system, I assume you will have some automation that can handle more than one ESP32, allowing you to assign different "names" to each. Going for 4xESP32 is probably the quickest route, but I sympathise with your notion that this design wasn’t the original intention of the CAN Bus system. :) But hey, as I said, we are hacking! If you, or anyone else following along, figures out how the units decide their "address" and how to identify them using a message to all (revealing their serial number or another unique identifier), please share it here. I’ll do the same if I discover more. Out of interest, do you have the two 30kWh batteries connected to a common busbar, or do you maintain two independent battery banks? canbus:
this is actually confusing, since the 0x108180FE is actually the CAN ID of the ESP and really should be differnet to the IDs that the r4875gs are using, I should change this to make it clearer. To trigger a response to get the r4875g to reply with the current staus the following message is sent here:
the -on message is then used to analyse the reply from the r4875g The Setting of new values for voltage amps etc is done in these functions, where you would need to add functions with additional IDs accordingly: 
|
Beta Was this translation helpful? Give feedback.
-
|
Thanks for your detailed reply! Interesting that my assumption that "the power-on order determines which address is used" was correct. I wasn't sure about that because it leads to a real limitation for the CAN bus idea in the first place. It would never be possible to use two or more identical devices on a CAN bus and identify them exactly?! For my situation this limitation would be somehow acceptable, because both my battery racks are feeding the same bus bar (the same off-grid solar system). So I could always work with the broadcast address (0x108080FE) to control them in parallel (on/off, voltage and current limit settings). But all the monitoring data (current values for volts, currents, power, temperature) will not longer be useful. In case I'll not be able to find a solution, I think I may install one ESP32/CAN tranceiver for each rectifier. In my situation, it would not be a problem to add 4 more ESP32 (via MQTT) into my existing home automation (OpenHAB based). I found this github project where a candump showing unsolcitied messages sending out by each rectifier. Maybe the data in these messages could be different for each rectifier (serial number, etc.) and may be used to identify the physical unit - but not sure?! Until now, I don't have the rectifiers (think about ordering them by mid of December and will be delivered by mid of January). As soon as the will be there, I plan to do some experiments with 2 units in parallel to see if there is a way to identify them. Btw. My solar project (39kW inverter power, 60kWH LiFePO4 battery, 18kWp PV) is described in detail in the DIYSolarForum (see my build thread there). This is how it looks like since it's in use (since May 2024): |
Beta Was this translation helpful? Give feedback.
-
|
Cool system, thanks for the link. You can play with the inverters on the bench anyway; it’s hard to break anything via CAN (at least I didn't manage to). |
Beta Was this translation helpful? Give feedback.
-
|
I found some bits of the puzzle: If you send all zeros to ID: 0x108040FE (current setting, every 5 seconds) and have only one unit connected it replies with all sensor values on ID: 0x1080407f. If you have more than one R4875G units connected to the same CANBus they reply with 0x1081407f, 0x1082407f, 0x1083407f... etc. still not sure if the same unit replies after reset, but the numbering system of the IDs returned at least, is somewhat logical. [12:29:00][D][canbus:069]: received can message (#1) extended can_id=0x1081407f size=8 |
Beta Was this translation helpful? Give feedback.
-
|
And while I was at it... if you send zeros to ID 0x108240fe (hoping to get a reply only from the second R4875G and setting the '-on message' to ID 0x1082407F, expecting to hear only from unit 2) you get exactly that; a reply only from unit 2. The log also shows the can decoding and number crunching of the sensor code. [13:03:35][D][canbus:032]: send extended id=0x108240fe rtr=FALSE size=8 |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
|
This is good news! Thanks for the investigation! Now I know I can use just one ESP32/CAN tranceiver to control and monitor all 4 rectifiers. Because in my situation they need to be adjusted to the same values for (startup) voltage, current, etc. it would not be a big problem if the address order would change at any cold start (but I hope not). For monitoring it would be nice if the order would always stay the same. Looking forward until I'll receive my devices to do some tests by myself. |
Beta Was this translation helpful? Give feedback.
-
|
I've made some compilation tests based on your code which supports two rectifiers and would like to discuss some details if you don't mind. I still don't have the rectifiers, so these are only "dry" tests with the ESP32 code. Be lenient, I am ESPHome newbie - until now I've only used Tasmota and ArduinoIDE but never ESPHome... a) sensor data via MQTT I was not able to access the "substitutions" of the from inside a lambda (do you know this could be done?). Because of that, I've defined a "global" I was able to access from lambda code. Here is the global definition I've used for testing: Initially I've tried this in the "substitutions" section, but as mentioned, without success. b) Status LED to signal CAN send and CAN receive With the following definitions: Maybe there is a better (central) location to put the LED flashing code to just let the LED flash if any CAN data is received... could not find one so far.? The location where I put it in my example requires to add this line to all four "can_id" sections in the "on_message" for all four devices. I'm not sure where to put the code to make the red LED (CAN sending indicator LED) flash each time. It should indicate any sending of CAN data, independent if it's triggered by the servlet GUI or via MQTT or something else. This would be the line to do the call: Maybe you'll have an idea. |
Beta Was this translation helpful? Give feedback.
-
|
Welcome to ESPHome, I was also looking for a visual to monitor if my CAN BUS is up and that's why there is a log message generated here: But to your points: The Lambda functions are a little picky about what they can see, and what is seen from them outside the function, and yes globals work. With regards to your LED CAN BUS indiactors: I would do it the lazy way and have the RED LED flash whenever the ESP sends the string of zeros by calling your script here. interval:
obviously the RED LED will flash even if your tranceiver is faulty and is just an indication that this specific -canbus.send function is active. But chosing green to monitor received message IDs that depend on messages being sucessfully sent is making up for it:) |
Beta Was this translation helpful? Give feedback.
-
|
Thanks a lot! Just one other question. Have you ever tried to enable/disable the hibernation mode of the unit via CAN command? The protocol spec shows this commands: enable unit (no hibernate): 0x108180FE [8] 01 32 00 01 00 00 00 00 As far as I understand, the CAN communication is always working so it should be possible to wake the unit with a CAN command, even if it's disabled (hibernated). I'm also not sure if the hibernation mode is available in all models G1/G2/G5 (I only find hints about "standby/hibernation" in the G5 manual, but nothing for the G1). I've read in a forum that the fans are off in hibernation mode (which would be very nice if they are quite if not used) and the consumption would be <=5W. |
Beta Was this translation helpful? Give feedback.
-
|
Yes Hibernation and Wake-Up are indeed features of the R4875G1, they are accessible by MQTT by sending the following text string (all caps): "PRESS" CAN ON Button (wake-up) Command Topic: ~/button/001_can_on_button/command CAN OFF Button (hibernate) Command Topic: ~/button/002_can_off_button/command I have never managed to accurately measure the power consumption in the hibernate mode, but it is very small and indeed the fans are off less than 5W sounds reasonable to me as it keeps the control electronics running and some of the LEDs are also on. If you want to hear what the FAN can do when it gets angry you can also send "PRESS" to: ~/button/fan_full_speed_button/command to calm it down again send "PRESS" to: ~/button/fan_auto_mode_button/command the fan command works only when the R4875G1 is ON, it doesnt work during hibernate, but this could be implemented because there is a seperate CAN message that can access full speed fan in hiberante state, but I think that's overkill. hope this helps. |
Beta Was this translation helpful? Give feedback.
-
|
That's great! Thanks a lot for clarification. Your're doing a great job here and perfect support! |
Beta Was this translation helpful? Give feedback.
-
|
It's me again - sorry... But I just want to let you know (maybe you already know it), there is maybe a way to find out the status of the unit (on or off/[hibernate/standby]). Maybe a sensor could be implemented to reflect the current status beside the already existing two on/off buttons. In the protocol spec for receiving data (0x1081407F), one of them is documented as:
But I found in a comment in a forum post that these "Unknown" values may reflect the current power status. |
Beta Was this translation helpful? Give feedback.
-
|
I found something else about the amps sensing and setting. There are different opinions in some forums if the factor for the amps should be 20 or 30 or something in between. Regarding to this code and this discussion, it looks like the amps are defined in percent from the max. amps of the unit (with a muliplier of 1024) instead of the absolute amp values. It sounds like this makes sense because for a max. 50 amp unit (R4875G2) the multiplier would be 1024/50=20.4, for a max. 42 amp unit (e.g. Emerson r48-2000e) it would be 1024/42=24.4 and for the max. 75A unit (R4875G1) it would be 1024/75=13.7. I found in your code a factor of 15 (even if your comment states a factor of 20), e.g. from your "# Amp Setting": For the amp sensors, the multiplier would be in this case 75 / (1024 * 100) = 0.0007324 instead of your used value of 0.0009765625 (which would be the correct multiplier in case it would be a max. 100 Amp rectifier - which does not exists). Maybe I'm totally wrong... |
Beta Was this translation helpful? Give feedback.
-
|
maybe you should create an issue for that? i don't use code of this project yet, but quickly scanning through what i'm using right now with my R4850G2 there also seems to be 20 used - and i feel it's fairly accurate.. |
Beta Was this translation helpful? Give feedback.
-
|
Agreed, so now I have made it much easier to change the scaling factor. Please try this easy to adapt YAML example if you want to try out different scaling factors. If you learn something new come back here to report please. My understanding right now is that the amp scaling factor (used to set MAX AMPS and used to read the MAX AMP setting) is |
Beta Was this translation helpful? Give feedback.
-
|
OK, game on. here is what I know; there are only two power states that I know of.
the R4875G1 likely reports this like this, not 100% sure though: ON State correlated message: [07:46:26][D][canbus:069]: received can message (#1) extended can_id=0x1001117e size=8 Hibernate State correlated message: [07:46:27][D][canbus:069]: received can message (#1) extended can_id=0x1001117e size=8 so yes you can listen to can_id=0x1001117e and filter for these two , and see if this is actually always true. Your first question is regarding the amp setting, which, as you rightly point out, is a little odd. Here is what I know: The R4875G1 has a nameplate maximum DC current of 75A. [Edit from the Future: now we now know that for my R4875G1s the 100% is at 69A and this MaxAmp-value can be read via CAN also] You are correct that the AMP setting is a percentage of the maximum current capability of the R48xxGx units, and that’s why there is so much confusion around this scaling value, because the different units have diffent max AMPS. To make it more confusing: If you connect pins 9 and 10 together on the board edge, the R4875G1 is set in such a way that 100% equals 60A. Since I only have R4875G1s, I didn’t bother adding a variable for the factory nameplate maximum amps (which would make the software much more universal, of course). If someone with a different variant is happy to test my code, I’ll look into that. They would also need to be able to load the unit to its maximum amps and measure its output current. hope your project goes well, and don't hesitate checking in when you have more questions. |
Beta Was this translation helpful? Give feedback.
-
|
OK I have added to this file [Control-two-R4875G1-with-one-ESP32-example.YAML] an experimental bolean on/off sensor based on the logic that can_id=0x1001117e becomes can_id=0x1002117e for the second R4875G1 (and the logic holds): it works, but I am not 100% sure we are looking at some unrealted messages that happen to correlate with the hibernate state, but I am confident enough to include it in the example code. I just called it ON/OFF. the MQTT topic for the bolean is: |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for these details. It's really impossible to find official information about the protocol. I think you have the best knowledge on this at the moment in the community because of your experiments and knowledge of this device. Thanks again for your effort to develop and share your findings and code. I will provide feedback as soon as my 4 units have arrived and I have done the first tests with several parallel units. |
Beta Was this translation helpful? Give feedback.
-
|
some freshly squeezed info about can_id: 0x1001117E message content: An error can be reproduced when two chargers are connected to the same can bus and only one of them is connected to a battery. If the charge current (difference) is big enough, the disconnected unit throws -> error mode x[4]: 0x1; x[5]: 0x2C and switches off. RED LED steady until unit is reconnected to load. So, in other words, putting the R4875s on one can bus is fine as long as their outputs are also in parallel with each other. So to fix the issue of getting an error whenever you need to isloate the output of one charger, I have added ON/OFF buttons for each individual charger to the dual charger YAML here. This update also has the mentioned online/offline binary sensor replaced with a text sensor, so it can also show ERROR.
|
Beta Was this translation helpful? Give feedback.
-
|
some more progress... I found a way to extract the UNIQUE IDs... and board type to set scaling factors automatically. And because it's available also, the manufacturing date... YAML example, anyone with a couple of R4875G, R4850G, R4830G if you want to test this. |
Beta Was this translation helpful? Give feedback.
-
|
Freshly registered to share my findings with you. See also https://www.beyondlogic.org/review-huawei-r4850g2-power-supply-53-5vdc-3kw For one of my marine battery customers, I have build a 3-phase 180A charger based on an original Huawei 1U 19" serverrack with 3 out of 4 slots occupied with 3 x R4850G2's. Over time we ran into erratic start-up issues when plugging in shore power but I'll set that aside for the moment. I'm here to share what I found during troubleshooting those issues, specifically when reverse engineering the 4 slots backplane of the rack. The backplane features a 4 position microswitch of which 3 positions allow setting the 'pull to ground' NEG(-) resistance commonly routed to all 4 slots upper pin 11. The resistance can be set from 0k0 upto 7k0 Ohm in 1k0 Ohm steps. I believe these settings are meant to identify upto 8 individual 1U racks. Even more interesting, the backplane also features a resistor network that varies the 'pull-to-ground' NEG(-) resistance individually routed to slots 1 to 4 with fixed values of 470, 7k0, 13k3 and 24k0 Ohm respectively. Even though I didn't test to verify, not having any experience with esp/devboard programming, yet, I am pretty convinced the combination of the eight(8) common per rack/backplane upper pin 11 resistance settings and the four(4) per slot varying lower pin 12 resistance settings, are meant to create a maximum of 32 addressable CAN-BUS ID's, tied to the hardware position of the slots in the racks. Hope that helps a bit. Cheers, Jan Willem UpCycle Electric |
Beta Was this translation helpful? Give feedback.
-
|
And a bigger boat/battery/grid connection. |
Beta Was this translation helpful? Give feedback.
-
|
@mjpalmowski, I saw your update regarding the max current. I removed the pin9-10 jumper on my R4875G1 and it reports 0x68 == 52A, same as my R4850S1. I use the max current value to determine the current programming resolution. I use full-scale = 1024d = MaxCurrent. So, resolution = MaxCurrent / 1024.0 for each respective R48XX. I verified the programmed value vs actual on both R48xx with my KP184 current load up to 8A. I'm curious what max current your units report with the jumper removed. Here's the full dump of the info and description of my two units in case you want decode what the different info packets mean. R4850S1:
|
Beta Was this translation helpful? Give feedback.
-
|
That's an easy and short answer, with jumper between Pin9 and Pin10 removed on my R4875G1s the max AMPs that can be set is 52A, I found this out by trial and error early on. When I get round, I will wire one of them up without the bridge and see what they spit out on Packet 1 ID: 0x1081507F Byte 5. [future edit: just checked the message on the R4875G1 with jumper 9-10 removed: same as you found: 0x68 == 52A] Also thanks for confirming 1024 using decent instrumentation, much appreciated! I will put this one to rest then, you are really ticking them off the list fast, thanks @JXL-2024. My software accepts only integers as AMP set input. Good to know that the R4850 can report the actual DC current value at a resolution of 50.78mA. |
Beta Was this translation helpful? Give feedback.
-
|
@JXL-2024 e.g.: But I'm not able to request the additional board settings like CAN-id, iMAX, pin 11 resistance, pin12 resistance information. I've tried sending to address 0x1081507E for the first unit (all data zero), but there is no CAN answer on 0x1081507F. What exactly is required to send to get this data?
|
Beta Was this translation helpful? Give feedback.
-
|
@fmeili1 , corrected request id to 0x108150FE |
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
On a different note: any of you guys willing to coach me through my first steps into the world of esp programming? To flatten my learning curve a bit in my very time limited life? I got all the hardware at hand (and then some), just dreading those initial ten (10) very time consuming rooky mistakes characteristic for diving into learning a completely new skillset on an unfamiliar platform. European/Amsterdam time zone preferably. |
Beta Was this translation helpful? Give feedback.
-
|
For me, the ESPHome "programming" is still a bit unusual. I've never programmed via YAML in this way. Instead I usually use the Arduino IDE and do all my programming in C++ for all my other controllers. I still haven't internalized this declarative programming philosophy and need to lookup existing examples while trying to do even simple things and I have no experience with it. So I don't think I'm the ideal person to coach that. But I absolutely willing to help as much as I can! Just ask. |
Beta Was this translation helpful? Give feedback.
-
|
Very good primer: https://www.youtube.com/watch?v=zD2ByKUZHcE |
Beta Was this translation helpful? Give feedback.
-
that's indeed a bit unspecific ;) 'programming' an esp as in 'writing code' for it, in the meantime is very similar to doing it for an Arduino or similar microcontroller. you ideally set up your programming environment on your PC, write your code (in C++ i think) and flash it to the chip. what evolved over the last years are environments like tasmota, esphome and similar. in this you are kind of configuring a firmware in a predefined structure (the yaml in esphome) and the use of libraries for supported hardware. there mostly is no need to actually 'write' a lot of code. with the instructions within this configuration file (yaml) the firmware is then built and flashed to your device within this environment. usually with a lot less headaches than the bare 'programming'. in the yamls of this repository there's indeed still a lot of 'code', but i assume that @mjpalmowski's goal is to get it into a library (/component), so then it would only be a single block in your actual yaml to define the pins it's connected to, and maybe a couple of other configuration options. that said, using esphome is easy and not hard to learn, but you're rather 'locked in' to the ecosystem of homeassistant, or at least that's where it has it's advantages. tasmota might be a bit more open/flexible as it's using mqtt a lot more by design. but code is relatively interchangeable, so if there's an interesting component on the one platform, most likely someone converts it to the other. and yes, the video posted above 8) |
Beta Was this translation helpful? Give feedback.
-
|
Jan/Manos, excellent work. Thanks for joining the effort. I have not looked into reverse-engineering the backplane, nor do I have access to one. I asked about the assembly code/cabinet name because I noticed in some of the Huawei product manuals that certain products allow reprogramming of the SLOT position, (most likely the ability of assigning specific Device IDs to SLOT positions, some LED blinking game that I saw in one of the manuals) while others do not. I was keen to find a service or user manual entry for the DIP switches, but so far, I’ve had no luck. For now, I’d like to share that, in the absence of a backplane—where the units label themselves (Device ID "1", "2", "3", etc.)—it does not appear to matter in which order they are switched on; they always seem to take on the same Device ID. However, this is not certain, as I only switched them on and off a handful of times before I got bored. Each unit has a QR code stuck to the front, which looks like this:
The CAN messages relating to the manufacturing data spit out this: [Board Properties] For the R4850G2 version Boards the report looks similar:
[Board Properties] the items marked in bold are being fitered and used in I am hoping that in auto-mode (no backplane), a specific BarCode always receives the same Device ID unless a unit is taken off or another unit is added onto the CANbus) Just in case anyone would like to see the mess that is the response to all zeros to 0x1081D2FE :) in the log this is all tansmitted in encoded text fragments like this: |
Beta Was this translation helpful? Give feedback.
-
|
Wrt those dip switches, measuring the backplane, pin 11 and 12 specifically but tried all other pins as well (backplane connector is fully occupied), all I could find is that only the first 3 out of 4 of the dipswitches are in use, providing a 3 bit number 0 to 7 that, multiplied by 1k0 Ohm sets the pull-down for the upper pin, on all slots of that backplane. The pull-down on lower pin 12 is fixed per slot. |
Beta Was this translation helpful? Give feedback.
-
|
Any of you guys in the Netherlands by any chance? |
Beta Was this translation helpful? Give feedback.
-
Here is what I could find, see pictures: |
Beta Was this translation helpful? Give feedback.
-
I used to live much closer to you (in Germany), but I moved to Arizona 9 years ago... a bit far to the Netherlands 😉 |
Beta Was this translation helpful? Give feedback.
-
|
Bit to far to drive by for a project meetup I guess |
Beta Was this translation helpful? Give feedback.
-
|
I have now all the components together and started to do the first tests with just one unit. As soon as I finished the preparation for all 4 units in parallel, I can start to do some testing with the possibility to assign a fix CAN address depending on the resistor values. But for now, I found one problem with the output voltage sensor - it just doesn't work and always showing the same value. The output voltage could be changed via CAN (checked via voltmeter), but the sensor always showing the old value and never updates (only when a new version is flashed the value is updating just once). Because I found no reason for the problem, I've created an issue for this problem. |
Beta Was this translation helpful? Give feedback.
-
|
I did reply to the issue, and we will work through it there, but more generally, the YAML can be configured as
It should be configured as one of those only, to keep the workload demanded from the ESP reasonable. |
Beta Was this translation helpful? Give feedback.
-
|
I think you are on to something, maybe it cosiders that if power loss AC and power loss battery side co-incides more than once in a row that the batteries must be broken and resets to a 'battery-safe mode' of 46v. Is the RED LED on when it hovers at 46V? |
Beta Was this translation helpful? Give feedback.
-
|
my units are about a year older and my part no and serial code look different. (sometimes you have to press the "find board" button again to catch the entire message in the logs (The ESP is touching it's limits in this CAN message HEX-to-ascii-text-translate-assembly-function, I only optimised it so much.) |
Beta Was this translation helpful? Give feedback.
-
that would mean the CAN module inside the r4875g1 is not resetting during AC power cycles, what happens if you use a resistor between DC- and DC+ to discharge the output capacitors? maybe that will fix it. |
Beta Was this translation helpful? Give feedback.
-
|
I found it! For testing, I've commented out the "interval 30s" to set the output voltage and output current limit every 30s. This is maybe the reason why you don't see this behavior on your side with every 30s resetting the values. It looks like the units "basic" behavior (without program interaction) is like I've described, but because of the interval resetting these values in the code, it's working as required. |
Beta Was this translation helpful? Give feedback.
-
Oh so glad this wasn't a bug, but a feature! The interval 30s is kind of a dead man's switch, telling the unit it's on CAN control. In this special case it rips the unit out of panic mode :) all good. |
Beta Was this translation helpful? Give feedback.
-
|
Btw. have you success to query the iMAX, CANid, pin11 and pin12 resistance values (like JXL-2024) described above? I'm trying with sending 0x1081507E for the first unit (all data zero), but there is no CAN answer on 0x1081507F to get these values. |
Beta Was this translation helpful? Give feedback.
-
|
wrong reply ID... packet 6 arrives at 0x1081507E all other 5 packets arrive on 0x1081507F |
Beta Was this translation helpful? Give feedback.
-
|
@fmeili1 , Sorry that was a typo in my post. |
Beta Was this translation helpful? Give feedback.
-
|
Just to sumarize what I think we now know: All zeros to ID 0x108150fe -> replies with 6 packets packet 1 with id: 0x1081507F Packet 1 Byte 5 contains the MaxAMPs of the unit x2. for example 00 01 00 00 40 8A 10 2B Packet 6 is sent out (?sporadic) in response to all zeros from ID 0x108150fe it contains the physical SLOT number and possibly Row number. Example: 00 06 03 04 00 00 00 00 Byte 2 Contains the ROW number In this example a resistor value of 24kOhm causes Byte 3 to go to 04, SLOT4 thanks @Jan-at-UpCycleElectric for pushing this into the limelight, maybe the ROW number 03 on PIN11 is an error code? anyone has more data? In conclusion the available data all together suggest:
thanks @JXL-2024 you moved the needle quite a bit on this issue. I salute you and certainly will update the code in due course to reflect these insights. |
Beta Was this translation helpful? Give feedback.
-
|
Some findings from me to add to this:
|
Beta Was this translation helpful? Give feedback.
-
|
Thanks for your input here, I am actually using the awesome openDTU software, I just realized your are part of the team that developed it. Packet number: 2 (ID: 0x1081507F) the only thing I observed is that 00 C0 are likely the final 3 digits of the serial number (barcode) in this case "192" as this tracks with another unit that also has its last three digits encoded at this place. |
Beta Was this translation helpful? Give feedback.
-
Nice! I am contributing for a while now and currently maintaining OpenDTU-OnBattery together with one other programmer. The project has its roots elsewhere 😉
Ha! Affirmative! 😉 My three units report the same. Interesting. They put the first 12 chars in ASCII in that message block and the last 4 decimal digits in binary. Very strange. Thanks for the info that I should not rely on the second byte with the same value to read max amps. Had no time yet to fiddle more with the units... |
Beta Was this translation helpful? Give feedback.
-
|
Some thoughts about the board properties (send on 0x1081D2FE, receive on 0x1081D27F). So I'm not sure if the string "BO" is really the identification for "broadcast over". In my case this it's the key "BOM" (with empty value for my untis). It's followed by the "Model" and ends with "/$ElabelVersion=4.". Maybe the result differs between R4875G1 and R4850G2. I don't have a good idea so far to detect the end of the message in a way which may be compatible for all units. In my case I can check for the key "ElabelVersion" to detect the end. |
Beta Was this translation helpful? Give feedback.
-
|
You're missing the last packet which comes in on 0x1081D27E |
Beta Was this translation helpful? Give feedback.
-
|
Ahhh, very valuable information! It works much better comparing to parse some content to find the end - just changed it in my test code and it's working reliable. |
Beta Was this translation helpful? Give feedback.
-
|
I think I found a reliable code to request the BoardProperties without the handling of timeout, out-of-bound and EOF detection via parsing the content (as @JXL-2024 mentioned via 0x1081D27E message).
|
Beta Was this translation helpful? Give feedback.
-
|
Thanks @fmeili1 that's really where this function had to go. I also love the trigger start to replace all those buttons, it should be fully automatic, I will just merge them, your solution is elegant. I have merged your fresh and crisp text import function and put all the set-up stuff into one big magic SETUP button :) see if you like it here: https://github.com/mjpalmowski/CAN-BUS-control-R4875G1-with-ESPHome-and-MQTT/blob/main/YAML%20example_new_auto_board_autoScaling_Factor.YAML mix and match R4850G2 from 2015 (got assigned ID-1, so after this I swap out HV18A0000192 with HV18A0000293 and see what happens) first Pair (ID-1 always reports first): Second Pair: Now, both R4875 boards have been ID-2, let's put them together and see if they remember who was ID-1: Result: The two R4875G1 boards took back their old IDs when put back together. |
Beta Was this translation helpful? Give feedback.
-
|
It took a while until all units and the program runs without issues. It costs me a couple of days to solve a problem which as a result made my ESP32 board really unstable with WiFi and MQTT and also with sending/receiving CAN messages. At first it worked for 2 units, after adding the 3rd, the problems begun and the reason was my own MQTT logging which I've implemented for debugging. I think the main problems where because I've also added MQTT logging in the on_boot section which caused the instabilities. It was so bad (even OTA was not longer working, etc.), that I first thought I've bricked my ESP32 board and so I've used a different one but it showed the same problems. After I've removed all MQTT logging the program now works flawless and reliable with all 4 units. So far I've done only idle tests (no load, but DC-output connected in parallel for all units) so I can't test the ams sensor (scaling) and the amp limit feature until my dummy load will arrive (a cheap resistance 2000W/120V water heating element which have 7.2Ω to be able to do some low current tests with about 7A - high power resistors are too expensive...). After the units had initially assigned their CAN-ID's (addresses) they always stay like this. Because of that, I've physically ordered my units to match the CAN-ID's (addresses). As long as I always have all 4 connected via CAN bus, they will stay the same which is ok for me. I've done many tests with the pin 12 resistors and I can confirm, that they have absolutely no influence to the assignment of the CAN-ID's (addresses) - even if their values could be read via CAN where the reported values are: 1=470Ω, 2=6k8, 3=13k0, 4=24k Now I'm in the final build phase to prepare the final setup (switches, fuses, studs, junction boxes, etc.) to be able to install them on top of my battery racks. Here are some pictures of my current test setup on my desk: |
Beta Was this translation helpful? Give feedback.
-
|
Great work and excellent progress, especially considering this was your first ESPHome YAML project. Thanks to everyone in this chat—your contributions have greatly added to the wealth of information we have on this family of 'Power-Modems.' It's fascinating to see what can be achieved through a bit of collaborative effort. I hope your testing, @fmeili1, goes smoothly and that you'll be charging batteries soon. By the way, when I conducted the maximum output testing on my R4875G1s, I discovered that the units are capped at pretty much exactly 4000W DC output power. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks! Btw. I found that relying on the frequency change >40Hz as a trigger to detect if the AC input gets connected to the grid is not reliable. Instead I found that the DC-output is better to use for that (If the DC output will be >40V, the unit is connected to the grid - if not in "OFF" mode). This is now one of my 3 triggers (beside on_boot and going to "ON" state) to query the max. amp of the unit to get the correct scaling factor. In case you are interested, I can send you the 2 files (one .yaml and one .h file) if they are completed. |
Beta Was this translation helpful? Give feedback.
-
|
Yea, you can just hard code the scaling factor, it's never going to change in your system, unless you play with pin9 and 10. With regards to your code, you can fork the project and maintain the code for 4 units there, you have the perfect set-up to maintain the code. |
Beta Was this translation helpful? Give feedback.
-
|
I've nearly finalized the installation for now. Here are some of the software points, I'd like to mention because of my findings:
Update: |
Beta Was this translation helpful? Give feedback.
-
interesting. still using this https://github.com/mb-software/esphome-huawei-r4850 and values seem fairly accurate (relating input to efficiency), even below 100W, esp. never seen any of that sticking to 0. quickly checking the code it also seems to use 0x73 divided by 1024 |
Beta Was this translation helpful? Give feedback.
-
|
I agree with @drahdiwaberl, The reason it might be sticking is a lack of updates, there is some rumour that the unsolicited Status/Amp CAN-ID 0x1001117E, 0x1002117E, etc, are used by the units to report energy measuremnts Wh, I could not confirm this so far, but my code does read these and at least publishes an independent (alt AMP) value, which interestingly uses the AMP_SET scaling factor and not the universal 1024. The update frequency of this ID is so high that I have restricted it to be read only if updated values are detected: |
Beta Was this translation helpful? Give feedback.
-
|
Finally my DIY Chargenectifier project is finished. Some more information:
I'm pretty happy with the results so far. Now I have to integrate the controlling of the rectifiers into my smart home (OpenHAB) rules. Pics: A pair on each battery rack / left pair with CAN controller / right pair / left AC grid feed (same for the right side)
Thanks a lot for all your help! |
Beta Was this translation helpful? Give feedback.
-
|
Thanks, what started as an engineering hobby project for my own little electric boat during Covid grew into my main business. I specialise in upcycling (pre-used) Porsche Taycan modules, LG_Chem E66A(63Ah) pouch cells. When matched and balanced correctly these modules proved to stay extremely well balanced over prolonged usage and time. No LFP there, technically my current build is 12S Li-NCM712 with a consistent usable voltage range from 3.2V to 4.2V. A single module 2P6S, two modules in series per string, fourteen strings in parallel, with a very specific system design to have all that act reliable as one big single 28P12S battery, de-tuned a bit to a 44,4V 1600+ Ah propulsion battery. |
Beta Was this translation helpful? Give feedback.
-
|
And yes please, a bin for this 7 psu config would be very welcome. I suppose the S3-WROOM-1 the more capable target compared to the C3-MINI-1 and WROOM-32D I have at hand. But could buy another if so recommended. |
Beta Was this translation helpful? Give feedback.
-
oh yeah, that was my first thought when you said 3x32A .. 😵 but maybe starting a seperate thread (or multiple) might be a good idea? ;) |
Beta Was this translation helpful? Give feedback.
-
OK let's do this step by step, I have created a bin (it's a derivative of my latest release, to install it please follow this link if this works with all your boards we move to the next step. It would be great if you could also test whether this AC-amp-limit is an offline limit or if you need to have the CAN BUS connected for it to be active. I invite anyone who is following along to chime in if they know. Many thanks. |
Beta Was this translation helpful? Give feedback.
-
|
Excellent, I'm trying hard to make time, will report back when I have first results. |
Beta Was this translation helpful? Give feedback.
-
|
Some more positive updates about the 4x R4875G1 rectifiers:
I can really recommend these rectifiers - I'm impressed. |
Beta Was this translation helpful? Give feedback.
-
Thanks for your interest! I know about the addressing scheme and I tried both 0x108080fe and 0x108180fe. Currently, I only have one unit on the bus at a time. I will probably want to play with multiple per bus, but that's low priority. After looking at the Excel sheet, I also tried |
Beta Was this translation helpful? Give feedback.
-
Long story short: No joy setting the AC current limit on my unit from 2013...The unit hits the DC current limit beautifully. However the AC current limit is kind of off. Seems to be off by 25% (looked at only two values, 2A and 3A). Maybe the multiplier is 1280? Will try another time.
@mjpalmowski Are you sure about this? I noticed that the command updates the output current limit. So it dawned on me that maybe this is just ja convenience command to actually set the output current based on the output voltage at the moment of processing the command. I did a quick test with a 10S LiPo battery and found that over the course of charging the battery from 38.3V to 41.1V, the output current increased from ~522W to ~565W. The (max) output current value was not updated during this time. It should have been corrected down to keep the AC current limit. The input current rose from ~2.40A to ~2.59A. Maybe this is within some sort of tolerance, but right now I must assume that this command does not actually set an input current limit, but is merely a convenience function to set the output current limit. 🤷♂️ |
Beta Was this translation helpful? Give feedback.
-
|
Hi, if you put the unit that is responding to input current limit setting into hibernation and use a power supply to vary the voltage on the DC terminals you should see the output current limit change. I have tried this with my R4875G1s and their output current limit (when restricted by the input current limit) goes up when the voltage at the output terminals drops. |
Beta Was this translation helpful? Give feedback.
-
|
I could not try this yet. I will need one of the cheap boost converters to be able to, as my desktop power supply only outputs 32V max. To be clear: @mjpalmowski By hibernation you mean sending the 0x32 setting: Today I upgraded my LiPo pack from 10S to 14S to better match a 16S LiFePo4 and the unit's capabilities. It turns out that 1024 is the correct factor for the input current limit also for my unit. It seems this was off due to the low DC output voltage. At least I have no better explanation... |
Beta Was this translation helpful? Give feedback.
-
|
When the unit is in hibernate mode, The rectifier still monitors battery voltage and other parameters when hibernating, you can also change settings, it just won't supply power, if you connect a power supply to the DC output just add a suitable resistor to limit any unwanted currents. Do your tests within the normal voltage range of between 48V-58V. Also as a side note. The rectifier has an output DC power! limit of nameplate power which is 4000W DC for the R4875G1, so you can actually set higher output currents at lower voltages although my firmware does not allow for that (but it's easy to remove the max AMP setting limit in the YAML). Another much easier way to test it (and I just did that with a R4850G2) is to disconnect the battery, set the AC current to 5A and Set the Max Voltage to 58V while the unit is ON. The MAX AMPS are shown as 19.7A To summarise: Some older R4950G2 models do not support input current restriction. However, those that do maximise DC current while restricting AC current. In other words, since the AC voltage is often a fixed number, you can use AC current as an input power limiter. Additionally, for all the boaters, mobile home, and caravan users who have strict current limits at their hook-up points, you can use this function to get your maximum fill in the shortest time without worrying that your AC current will creep up as your battery voltage increases. |
Beta Was this translation helpful? Give feedback.
-
|
I want to report one problem which I was not able to solve but with a workaround it's ok. If I request all sensor values via broadcast the value for "AC in amp" from unit 1 and 2 are NOT updating only for unit 3 and 4. All other values are working with fast updates and correct values for all units. I've checked the code many times and was not able to find a coding problem. After I've switched from broadcast sensor request to 4 individual requests with individual addresses, it's working. I have no idea why this happens because ALL other sensors are updating. I think there will not be any negative effect if requesting individually. I also have the "feeling" that the initial sensor updates (after switching the units ON) now arrive earlier compared to broadcast request. Just want to let you now. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for sharing! I believe you're encountering what I like to call the "CAN Classroom Effect." When you send a broadcast message to address zero (all units), they all respond simultaneously, each with all their parameters. This creates a significant amount of traffic on the bus, much like a teacher asking a question and the entire classroom answering at once. Interestingly, it seems that the same messages consistently get lost, which suggests this might be a bug. (Using a CAN analyzer could help identify which messages are getting through and which are being dropped.) Your workaround effectively spreads the workload over 800ms, giving each unit a dedicated 200ms window to reply. This maintains an excellent refresh rate while significantly reducing peak traffic on the bus. An added benefit is that the ESP32 (the "teacher") has a much easier job processing the responses. |
Beta Was this translation helpful? Give feedback.
-
|
Sounds like a good explanation and to spread the processing time over a longer period is useful anyway. But it's really strange that always and only the same value (AC input current) have this problem when requesting via broadcast. You've wrote "which suggests this might be a bug"... do you think it may be a problem in my YAML code or in the rectifiers firmware? |
Beta Was this translation helpful? Give feedback.
-
|
When I said suggest, I really don't think it is a bug, much more that it appears to be without actually being a bug. I have noticed that certain messages get lost more than others. This might be because there is some sort of rhythm to the code and some messages just end up being sent at busy moments, every time. I think it might be better to use prime numbers as intervals, avoiding this rhythm situation. |
Beta Was this translation helpful? Give feedback.
-
|
A fast "hacked" OpenHAB dashboard (still ugly) just to show the values and change the modes and some settings. Currently charging with manually enabled mode "2" with (rounded) 8kW.
|
Beta Was this translation helpful? Give feedback.
-
|
Huawai engineers would be proud. 新年快乐，致华为工程部！ |
Beta Was this translation helpful? Give feedback.
-
|
Happy Chinese New Year to you all too! |
Beta Was this translation helpful? Give feedback.
-
|
These Huawei R4875G1 units are really fantastic👍 Even after many hours of using them with 40A each, the fans are running slow and the noise level is really low 😀 The CAN bus communication is flawless and the units react very fast to requested changes. For now, I've working with the following charging rules:
All rules are implemented in the ESP32 and even when the smart home system will be down (where I'm doing all the complex monitoring stuff), the solar system will work just with the controller itself. This is more reliable compared to my previous smart home based controlling of the solar system. |
Beta Was this translation helpful? Give feedback.
-
|
Nice to hear that your ESP32 is doing the good work, I agree that local control is the key to reliable operation, the YAML code enables anyone to add high-speed, low latency, low cost, local control to their project, without the fear of breaking the existing code. Adding direct BMS communication, i2c-smart-shunt, CT-clamp options to the local control and using "home automation" as SCADA might be the general direction. |
Beta Was this translation helpful? Give feedback.
-
|
Yesterday, I've finalized the rebuild of the 4 DIY chargeverters (4 AWG and 1/0 AWG cables and using DC breakers instead of the "cheap" battery switch and the MEGA fuse). Done a full load test with 4 x 75A this morning for about 25 minutes. All wires, connections, etc. stays cool (or a bit warm as expected). The R4875G1 working nice, even under full load. They heat up until about 84°C (184°F) with an ambient of 18°C (65°F) and fore sure the fans are at full speed with the full load (and loud because of this). Both racks are looking like this now (this is the one without the ESP32 CAN controller). Thermal images for the one with and the one without the CAN controller at max. load: And some other pictures: Amps for the first and second rack - exact same load! Here are the values from CAN bus via ESPHome via MQTT to OpenHAB (with a fast hacked dashboard): It's unbelievable how these units can spit out 4050W continuous with such a compact volume - I'm really impressed! |
Beta Was this translation helpful? Give feedback.
-
|
OpenHab playing well together with ESPHome firmware controlling such powerful industrial hardware to basically feed your entire house with power, awesome. I think @Jan-at-UpCycleElectric has a good chance that his marine battery project can be rapid charged by these rectifiers also. Bring on 7kW/phase:). Also we have @schlimmchen from the amazing openDTU project (they have been busy hacking the DC to AC side doing some awesome off-label stuff with off-the-shelf microinverters) on the chat now (a very warm welcome). |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
For my solar system, I'm planning to use 4 of these R4875G1 units in parallel. Because CAN is a bus system, I thought it's a good idea to use just one ESP32 with one CAN tranceiver to control all 4 rectifieres.
Regarding the linked protocol spec (XLS-sheet) of the R4875G1, up to 15 units may be attached in parallel to the CAN but. The document states, that each device (node) has it's own CAN address 0x108180FE for the 1st device, 0x108280FE for the 2nd, etc. If using 0x108080FE the settings will broadcast to all units.
But I have no idea how do I know, which unit will use which address (which physical is the 1st, 2nd, etc.)? I can't find any hint about e.g. an address dip-switch to select the CAN address (e.g. 1-15). As far as I know, CAN bus does not have the concept of "client addresses" at all and the identification should be done by analyzing the content of the message itself. But how does this work for such a situation?
E.g. if I look to the yaml part to receive a message for just one rectifier on the CAN bus, it's pretty much clear:
The "can_id" 0x108180FE is the ID of the first unit and the "on_frame" "can_id" 0x1081407F will give me the values.
Regarding of the spec, I would assume to add just another section like this:
(I'm not really sure, if the "on_frame" "can_id" 0x1081407F is still the same as for the first unit or if it must be 0x1082407F instead... if I understand the spec, it should be the same ID for the on_frame?).
How does the 2nd R4875G1 "know" that it should use the 2nd ID (0x108280FE) for sending? Maybe it detects that the 1st address/id is already in use? But if this would be the schema, it would change if the units are powered up in a different order...
I'm a bit lost about how to use it.
The alternative would be to install 4 separate ESP32 devices with 4 separate CAN tranceivers... but I think this is not the ideas of the CAN bus.
Here is the schematic of the whole planned setup (each charger case with 2 rectifiers each will sit on top of a server rack cabinet):
This is the schematic and PCB of the planned CAN controller:
and the used CAN tranceiver (soldered on the CAN controller PCB):
Beta Was this translation helpful? Give feedback.
All reactions