• Have you tried out dark mode?! Scroll to the bottom of any page to find a sun or moon icon to turn dark mode on or off!

diy solar

diy solar

DIY 'Chargenectifier'

In general, I have a few issues or fear future problems or have some disadvantages when using the AIOs grid connection that I would like to eliminate with the double-conversion. I want to list most of them as a summary:

Current issues while using AIO's grid support:
  • At the moment, when the AIOs will be connected to the grid (not the moment when the AIO goes to bypass), the fault F60 (Power feedback protection) occur sometimes. If this happens, I have to reset the whole system with all 6 AIOs. It happens more likely if the load is low and PV production is ongoing. I've implemented complex smart home rules to minimize the risk, but it's anyway not nice.
  • The switching time, if grid is attached and the AIOs going to/from bypass (I only use SBU mode), is short - but it's long enough that my small UPS' in the house are clicking. Sometimes my pool pump will stop/restart when the AIOs are switching from/to bypass. I think for long term usage, my appliances won't like this.
  • update: I've forgotten to mention the following scenario, which produces a complete crash (all AIOs going to fault mode):
    • All units are switched on and system is running in bypass
    • Switch off some units to save idle consumption over night (system still in bypass)
    • As soon as one of the AIO which were in "off" state is switched back to "on" (while the working ones are still in bypass)... boom! All AIOs in fault mode and ALL GFCI's breakers and many GFCI receptacles tripped in the house... luckily no damage on appliances nor on the AIOs... but really bad behavior.
Current disadvantages while using AIO's grid support:
  • The grid battery charging can only be used if the AIOs are in bypass mode. When running in inverter mode, no grid charging is possible (inverter and grid charger using the same H-bridge - it's either/or). It's only a small disadvantage, but it's not longer the case when using double-conversion.
  • The grid charging efficiency of the AIOs is not great.
  • The max grid charging current could only be controlled in pre-defined steps which does not always fit the needs.

Possible problems in the future while using AIO's grid support:
  • I know at least one person who also uses the EG4-6500EX (also 6 in parallel) where his POCO showed him that his system feeds back into the grid (see this message). He also always using SBU mode - usually in SBU a backfeed should never happen - but it does (even only for small amounts and maybe only in the situation when bypass is switched on or off - but it's not really clear until now). I don't want to risk problems with my POCO about back feeding in the future.
  • I'm pretty sure the bypass relays will wear out quickly when they are often switching under high load conditions (which I have often).



For sure there are also disadvantages with double-conversion which I'm aware of and I'll live with them:
  • The house could not draw more than the max. online-conversion power over longer times (depending on battery SOC). In other words, if the batteries are completely empty, the max house load could not be over 16kW because the chargers are "only able to deliver 16kW compared to the possible 39kW when the AIO's grid support would be used (6x 6500W). It's very unlikely that this happens and I will live with this (and I'm not willing to build a 39kW DIY-chargeverter ;) )
  • Additional costs.
  • update: It's not possible to enable/disable battery charging or enable/disable driving the house load separately. Because the rectifiers just feeding the busbars, it does automatically both. Maybe it's possible to play a bit with the rectifiers voltage to do "more or less charging", but it's not really controllable.
 
Last edited:
The house could not draw more than the max. online-conversion power over longer times (depending on battery SOC). In other words, if the batteries are completely empty, the max house load could not be over 16kW because the chargers are "only able to deliver 16kW compared to the possible 39kW when the AIO's grid support would be used (6x 6500W). It's very unlikely that this happens and I will live with this (and I'm not willing to build a 39kW DIY-chargeverter ;) )
This is an excellent point that I'm currently "dealing with" running a single 3KW rectifier.
The other big drawback I foresee is not being able to "quick charge" before bad weather is expected. When I had the grid connected to my AIO I could quickly dump over 8KW/h into my bank.

I think I need to order at least one more rectifier.... I said I'd kick myself for not ordering two...
 
Nice setup! With regards to standby: The R4875G1 has two standby states:
1) Not charging but ON. Here the fans are running yellow LED is off, Green LED is on.
2) Hibernate. Here the fans are off but the Can Bus is up. In Hibernate the unit uses next to no power (as in my utility smart meter reports zero). It also doesn't use battery power in hibernate. Yellow LED and green LED are on steady.
 
Nice setup! With regards to standby: The R4875G1 has two standby states:
1) Not charging but ON. Here the fans are running yellow LED is off, Green LED is on.
2) Hibernate. Here the fans are off but the Can Bus is up. In Hibernate the unit uses next to no power (as in my utility smart meter reports zero). It also doesn't use battery power in hibernate. Yellow LED and green LED are on steady.
Thanks for this information!
I've searched the R4875G1 documentation about standby/hibernate mode but I was not able to find anything about it.

The documentation of the R4875G5 model does at least mention a standby/hibernate mode with a power consumption of <=5W, but it says that the yellow LED is "steady on" if "The rectifier is hibernating".

The CAN protocol spec shows how to enable/disable standby (here it's called "disable Huawei" / "enable Huawei") but does not mention possible differences between the G1 and the G5 models.

Code:
enable unit (no hibernate): 0x108180FE [8] 01 32 00 01 00 00 00 00
disable unit (hibernate  ): 0x108180FE [8] 01 32 00 00 00 00 00 00

Have you tested hibernation by yourself? With the G1 or with the G5 unit?
 
Thanks for this information!
I've searched the R4875G1 documentation about standby/hibernate mode but I was not able to find anything about it.

The documentation of the R4875G5 model does at least mention a standby/hibernate mode with a power consumption of <=5W, but it says that the yellow LED is "steady on" if "The rectifier is hibernating".

The CAN protocol spec shows how to enable/disable standby (here it's called "disable Huawei" / "enable Huawei") but does not mention possible differences between the G1 and the G5 models.

Code:
enable unit (no hibernate): 0x108180FE [8] 01 32 00 01 00 00 00 00
disable unit (hibernate  ): 0x108180FE [8] 01 32 00 00 00 00 00 00

Have you tested hibernation by yourself? With the G1 or with the G5 unit?
I've just got the feedback from the developer. It looks like that all version (G1, G2, G5, etc.) support the hibernation mode and it's already implemented in the CAN control code! Perfect!
 
Thanks for this information!
I've searched the R4875G1 documentation about standby/hibernate mode but I was not able to find anything about it.

The documentation of the R4875G5 model does at least mention a standby/hibernate mode with a power consumption of <=5W, but it says that the yellow LED is "steady on" if "The rectifier is hibernating".

The CAN protocol spec shows how to enable/disable standby (here it's called "disable Huawei" / "enable Huawei") but does not mention possible differences between the G1 and the G5 models.

Code:
enable unit (no hibernate): 0x108180FE [8] 01 32 00 01 00 00 00 00
disable unit (hibernate  ): 0x108180FE [8] 01 32 00 00 00 00 00 00

Have you tested hibernation by yourself? With the G1 or with the G5 unit?
yes, I have tested and use the hibernate feature on the G1 model. It works very well.
 
The original offer was $90 plus $49 shipping for one. I've got in contact and they made a new offer with 4 units with a total of $340 plus $135 shipping. This is the ebay link to the offer (delivered from China).
This is the link for the PCB connectors (they are not included in the R4875G1 order and I had to choose a different seller), 4 connectors are about $55 and free shipping.
Short update.

The seller contacted me and told me that DHL would charge an additional fee of $44 because I supposedly live in a remote area (n)

I live in a city of about 60,000 people and there is a major highway just 15 miles from here... how on earth do they think I live in a remote area...

Ok, after paid the additional fee via Pa*Pa* the seller immediately provided the tracking number. As far as I can tell, the seller seems to be serious and he responds quickly to emails and he sent me a picture from the additional requested fees from DHL.

So the total shipping costs for 4 units adding up to $179. The overall total is about $130 per unit, door to door. All in all it's still a good price for new units.
 
Today, I've received the 4x R4875G1 via DHL.
Perfectly packaged! No damage, brand new, each packed in original Huawei styrofoam like protection case.
I can recommend this seller.
They are surpringly small for 4kW 😃

Now I'm waiting for the PCB connectors to do a test run with all 4 units. ESP32 CAN controller PCB is finished and ready to use. Curious about the result - and the noise level.

Is it ok to mount them on the long small side (with the fan on the top front, LEDs on the bottom) instad of flat laying?
 
Is it ok to mount them on the long small side (with the fan on the top front, LEDs on the bottom) instad of flat laying?
I don't see any issues with it as long as they have adequate airflow, those little fans are screamers and move a lot of air for there size.
 
I don't see any issues with it as long as they have adequate airflow, those little fans are screamers and move a lot of air for there size.
I just hope that I don't need to run the units too often and only with usually less than 50% load and that the fan speed in the rectifiers are temperature controlled.

After playing around how to place a pair of them on top of my battery racks, I think option 3 (side by side laying flat) is the best for heat dissipation.

1733933719194.jpeg 1733933763549.jpeg 1733933821477.jpeg

Instead of building complete cases, I'll try to 3D print some sort of protection brackets on the front and rear while the rear should also cover the connector PCB with the contacts against touching. Planning to put additional rubber shock absorbers under the units to help to reduce the noise.
 
To manage noise it is best to limit their power output to something reasonable (something between 1 and 3kw each). These rectifiers are designed to run pretty warm in any case, mine are kept at just below 75C when operating even at 1.5kw output. Victron uses a similar cooling philosophy, it allows the equipment be cooled at much higher ambient temperatures. (They use heat resistant components). PLA will certainly melt here. There is also a grounding screw at the bottom of the case, you might want to use this, as the edge connector ground lug is held in place by a machine screw, self tapping into... plastic 🫣.
 
I just hope that I don't need to run the units too often and only with usually less than 50% load and that the fan speed in the rectifiers are temperature controlled.

After playing around how to place a pair of them on top of my battery racks, I think option 3 (side by side laying flat) is the best for heat dissipation.

View attachment 262234 View attachment 262235 View attachment 262237

Instead of building complete cases, I'll try to 3D print some sort of protection brackets on the front and rear while the rear should also cover the connector PCB with the contacts against touching. Planning to put additional rubber shock absorbers under the units to help to reduce the noise.
I doubt the orientation matters much between those three options, but watch the exhaust air, it's pretty warm on the CVGC, and blows right on the output Degson connectors. And yeah, noisy!

Can't wait to see how this works for you, this might make for killer golf cart chargers!
 
Thanks @mpeterson and @wpns for both your valuable comments!
I've played around with printing narrow brackets to slide over the rectifiers. I want to use PETG filament to have at least a bit better temperature resistance (it's starting softening around 80-85°C compared to the 60°C with PLA). The brackets should just keep them in place when sitting on the battery racks, no strong support is required.

This is an example for the front bracket which will be slide over the rectifiers with a tight fit. I've implemented some deepenings inside around the brackets and holes for better heat removal - I hope this will work.

When the PCB connectors will finally arrive, I need to design the rear bracket to cover the contacts and enough opening to not block the cooling fan output.

Usually they should run at 1-2kW per unit, but with the option to ramp them up to 4kW each if required.


1734017436846.png
 
Last edited:
OK, I would like to address one of the issues raised by @fmeili1 regarding his double conversion, which I would describe as "the concern about battery yo-yo". The idea is to maintain the batteries at a specific state of charge (SOC), say 20%, to be ready to absorb sunshine in the next few hours or to soak up cheap overnight electricity. Essentially, we aim to "hover" the battery at 20%.

To achieve this, you need to 'Chargenectify' the system as much as your house is drawing. This requires a mechanism that closely ties the rectifier output to the inverter demand.

One approach is to directly tie the rectifier output to the inverter input, disconnecting the batteries. :)

Realistically it involves closed-loop control of the rectifiers, where the rectifier voltage is set slightly above that of the batteries, and the current matches the input current (AMPS) of the inverters, updated every x milliseconds. This could be implemented using CAN-BUS control of the R4875G1 rectifiers.

To achieve this, publish the DC current value (inverter AMP draw) to the AMP-setting topic of the R4875G1. Or if you prefer not to use high-latency Wi-Fi/MQTT, you could consider connecting your DC Shunt to a free GPIO pin on the ESP running the CAN-BUS. This would provide the necessary control values with minimal latency to the CAN BUS...
 
The idea is to maintain the batteries at a specific state of charge (SOC), say 20%
Or just pick a voltage (52V is about 25%SOC) and your grid will never let your batteries fall below that level. I ran that way all summer and didn’t buy that much power just to keep the AC input live and the fans spinning.
 
OK, I would like to address one of the issues raised by @fmeili1 regarding his double conversion, which I would describe as "the concern about battery yo-yo". The idea is to maintain the batteries at a specific state of charge (SOC), say 20%, to be ready to absorb sunshine in the next few hours or to soak up cheap overnight electricity. Essentially, we aim to "hover" the battery at 20%.

To achieve this, you need to 'Chargenectify' the system as much as your house is drawing. This requires a mechanism that closely ties the rectifier output to the inverter demand.

One approach is to directly tie the rectifier output to the inverter input, disconnecting the batteries. :)

Realistically it involves closed-loop control of the rectifiers, where the rectifier voltage is set slightly above that of the batteries, and the current matches the input current (AMPS) of the inverters, updated every x milliseconds. This could be implemented using CAN-BUS control of the R4875G1 rectifiers.

To achieve this, publish the DC current value (inverter AMP draw) to the AMP-setting topic of the R4875G1. Or if you prefer not to use high-latency Wi-Fi/MQTT, you could consider connecting your DC Shunt to a free GPIO pin on the ESP running the CAN-BUS. This would provide the necessary control values with minimal latency to the CAN BUS...
Mhh, sounds like a near perfect, very challenging solution. I doubt, that the amp changing speed may not be fast enough even if you would go the suggested shunt-gpio route. Someone may do a test with the rectifier how long it will take between sending the new amps settings to the unit via CAN until the new amps are active.

The idea @wpns mentioned sounds a lot easier.
 
Agree, simple is better. I think setting the minimum voltage and leaving them on and connected is a solid (and more importantly already tested;) solution. If you want to refine that a little you could wake them up at voltage x and send them back to hibernation at voltage y of course.
 
I've not been here for a while, sorry, but did anyone find a way to change the CAN address of the Vertiv R48 units??

I've been looking into it, and haven't found anything yet. I do believe they get individual addresses automatically like some other rectifiers do. I am looking into how/which addresses they could be. I only have one rectifier at this time; I'm waiting for another one...
 
I've not been here for a while, sorry, but did anyone find a way to change the CAN address of the Vertiv R48 units??
In the meantime I've played with my 4 units in different orders and different resistor values between pin 12 and ground (to "simulate" they would sit in an original Huawei rack with a backplane on slot positions 1-4 with their slot resistor values of 470Ω, 7kΩ, 13kΩ and 24kΩ for the 4 slots - see this github discussion for details), etc..

BUT I found no way to influence the CAN address of the units from "outside".

If you don't change the number of units on the same CAN bus the (randomly chosen) address assignments will be persistent. I first tried with 2 units in parallel and they use the same order after every restart (same CAN id). As soon as I've added the 3rd unit, all 3 addresses are newly randomly reassigned and stayed after this reassignment (even the 1st and 2nd units addresses changed). Same after added the 4th unit.

Because the addresses are persistent after they are once (randomly) assigned for a given setup, I've ordered the units physical location based on their CAN address (units left to right with addresses from low to high).
 
Last edited:
@mpeterson
It took a while to get all 4 units to work in my experimental test setup (with 120VAC, later they will run with 240VAC). I had a lot of very useful messages going back and forth with the maintainer and other users of the ESP32 github project to communicate with the units (R4875G1) and finding out some more details:
  • To bring the unit from it's default 52Amp max. current to it's full 75Amp max. (pin 9 and 10 on the PCB connector must be connected)
  • To automatically determine the exact value for the amps scaling factor of the units.
  • Learned, that's it's not possible to assign the CAN-ID's/addresses in a defined way to the physical units. But if all units are connected in parallel for the first time to the same CAN bus, the automatically assigned CAN-ID's/addresses will stay permanent as long as the number of parallel units in the setup will not change.
  • How to query some passive unit information (type, model, serial number (barcode), manufacturing date, etc.)
  • How to hibernate (OFF) the units (no fan noise) and wake them up (ON) if required.
Btw. My units look like brand new (but old stock ). Two are manufactured in Q3 2019, one in Q1 2020 and one in Q2 2020. The serial number (barcode) of the "oldest" unit from 09/2019 starts with "HV.........." but the others are starting with "LU..........". They look the same and behave the same.

All 4 are working from voltage output and CAN-bus control point of view (no load tests so far). All sensor data are available via MQTT (DC output: voltage, amps actual, amps limit setting, power; AC input: voltage, current, power, frequency; and temperature.
Because all units are connected to the same DC bus, the DC voltage settings and the fallback voltage settings for the units will automatically be broadcast to all units.

Most or maybe all of this information is also valid for the R4875G5, R5850G2, etc.

Now I'm building the final installation setup with switches, fuses, breakers, studs, wires, cables, junction boxes, 3D printed brackets to hold the units, etc.).
I underestimated the costs for the additional parts... So the total cost with everything was $1,300 ($560 for the 4 units with PCB connectors and $570 for all other parts). The DIY project is about $81/kW (max. 16kW) compared to (3x) Chargeverters with $90/kW (max. 3x5.12kW=15.36kW for $1395 - the Chargevertes would have been a bit more (maybe $95/kW) when adding the breakers, wires, outlets, etc. for them also...

At the end it's not a big difference from the price point but the DIY is much more flexible because of the ESP32 programming be in own hands and MQTT integration, etc. is possible... and it's more fun :cool:

Some pictures of the experimental setup (stacked for testing only, later each pair will sit on top of each battery rack).

20250103_075055_resized.jpg 20250103_075115_resized.jpg 20250103_075148_resized.jpg
 
I've done a huge effort to reduce the noise of my six EG4-6500EX AIO's because the master bedroom is behind the inverter wall. Now I have this noise problem again with the rectifiers... luckily they are silent when in hibernation and I only need them from time to time to get grid support (I may shift the usage to daytime). But I think about possiblilities to make them more quiet. I don't plan to use them with max. power, maybe with <50% most of the time - at full fan speed the units are like jet engines :oops:

Maybe some passive damping with an insulated housing... or maybe with an active environment noise reduction circuit (ANR, ENR). I'll try to do some research about active ANR/ENR with a simple analog circuit (DSP seems to be to complicated and it would need too much CPU power for a simple ESP32).
 
I've done a huge effort to reduce the noise of my six EG4-6500EX AIO's because the master bedroom is behind the inverter wall. Now I have this noise problem again with the rectifiers... luckily they are silent when in hibernation and I only need them from time to time to get grid support (I may shift the usage to daytime). But I think about possiblilities to make them more quiet. I don't plan to use them with max. power, maybe with <50% most of the time - at full fan speed the units are like jet engines :oops:

Maybe some passive damping with an insulated housing... or maybe with an active environment noise reduction circuit (ANR, ENR). I'll try to do some research about active ANR/ENR with a simple analog circuit (DSP seems to be to complicated and it would need too much CPU power for a simple ESP32).
I wonder if you can do some kind of ducted hood that covers the entirety of the printed structure. And then on the other wider side some much larger, slower running fans with good static pressure rating. And then disable the tiny screamers?
 
I see one controller for 4 units, good stuff. I think putting them all on the same bus was a good decision, allows the units to communicate for load balancing as they ramp down... 16kW what a beast. And I agree, more control and much more fun than an off-the-shelf-system.
 

diy solar

diy solar
Back
Top