jbroo

@Coulomb it was not the resistors. Through sheer endless trial and error I've figured out the calibration procedure, for the most part:
Issue PSDF to clear and begin calibration. Issue PMID (still not sure what this does exactly). Turn off mains and load on the inverter and power it off. Power it back on, turn on mains supply and load. Turn off mains supply and let your batteries drain to as low as you're comfortable with, but preferably a few volts lower than your "high" reading. Turn on the mains supply, charging should begin. Issue PBATLnnnn where nnnn is your low voltage, as soon as possible, when the reading on your meter has stabilized. Important: ensure your mains is connected when issuing this. Issue PSAVE. Get your battery charge level as high as possible. You may need to force a charge by using equalization or set bulk charge to max (30V). Issue PBATHnnnn where nnnn is your high voltage - important: ensure your mains is connected when issuing this. Issue PSAVE. If all went well, your battery (and SCC) voltage should now be in line with your meter. If you bumped up your bulk, float and equalize voltages to force a charge, now is the time to set them back to their proper values. Optional: Issue CF11 and PBFnnnn where nnnn is your voltage on the meter with both fans on. Follow this with PSAVE. The crucial point that I found out, is to never try to set the voltages without mains connected - I was doing this when setting the low voltage (turning off mains and waiting for the batteries to discharge sufficiently). This is what was causing my voltages to be out of wack. Always turn the mains on after discharge and set the low voltage as soon as possible thereafter.
Happy tinkering.
 

@Coulomb it was not the resistors. Through sheer endless trial and error I've figured out the calibration procedure, for the most part:
Issue PSDF to clear and begin calibration. Issue PMID (still not sure what this does exactly). Turn off mains and load on the inverter and power it off. Power it back on, turn on mains supply and load. Turn off mains supply and let your batteries drain to as low as you're comfortable with, but preferably a few volts lower than your "high" reading. Turn on the mains supply, charging should begin. Issue PBATLnnnn where nnnn is your low voltage, as soon as possible, when the reading on your meter has stabilized. Important: ensure your mains is connected when issuing this. Issue PSAVE. Get your battery charge level as high as possible. You may need to force a charge by using equalization or set bulk charge to max (30V). Issue PBATHnnnn where nnnn is your high voltage - important: ensure your mains is connected when issuing this. Issue PSAVE. If all went well, your battery (and SCC) voltage should now be in line with your meter. If you bumped up your bulk, float and equalize voltages to force a charge, now is the time to set them back to their proper values. Optional: Issue CF11 and PBFnnnn where nnnn is your voltage on the meter with both fans on. Follow this with PSAVE. The crucial point that I found out, is to never try to set the voltages without mains connected - I was doing this when setting the low voltage (turning off mains and waiting for the batteries to discharge sufficiently). This is what was causing my voltages to be out of wack. Always turn the mains on after discharge and set the low voltage as soon as possible thereafter.
Happy tinkering.
 

They are usually pretty obvious, as they will often be a pair of 3-6 (often 4) medium sized SMD resistors in series, with the ends kept close together (so they get affected about the same by noise), and often they are covered in "silk screen" (I doubt that silk is actually used any more):

Note the battery fuse nearby (they look different these days, sometimes just necks of metal), and near the battery terminals. These have 1004 marking (1.00 MΩ each), more commonly they are in the high hundreds of kilo-ohms.
With your wildly varying results, I think it's worth checking these first. Test with heat and cold, as well as gentle pressure on the PCB.

The resistors themselves cost less than about US$1; you'll pay 20x that for shipping.
So the voltages you entered in the PBATH and PBATL commands are from a trusted multimeter, with the  battery not charging or discharging very much,  and were hours apart?
I'm not  familiar with those. Perhaps the  magic formula that you found by trial and error isn't right in all circumstances?
Maybe I'll get a change to chase up those commands one day, and see if I can sort out what they really do, and possibly, if and when to apply them.

I got this far: in Axpert firmware version 02.43, the CF command takes more digits:
C F n n n m m m
n n is 00, 01, 10, or 11 as per the protocol manual
m m m seems to be a fan percentage, 000  to 100. It doesn't seem to be error checking the mmm parameter.
But I can't see how this would affect battery calibration.

I believe that you have to send the H and L commands at significantly different SoCs, otherwise it can't calculate the offset, only the slope, of the curve.
So when it's nearly full, send a PBATH,  then hours later when it's much lower, send the PBATL.
These are designed for use on the test bench at the factory, where you can adjust a power supply to look like a high or low voltage battery. If you have a suitable adjustable power supply, you could do the same.
I'm surprised it ACKd the second command when the parameter to the first was the same.
Edit: It's also possible that the resistors are now intermittent, in which case this won't help much, and you'll have to eventually repair the resistors.

There are any number of ways to get it. But I suspect this is an XY problem (https://en.wikipedia.org/wiki/XY_problem). Why don't you tell us what you are trying to achieve, rather than how you think you need to go about doing it?
To answer the question anyway:
1. look on your router for assigned clients and their MAC addresses.
2. ping the IP address, then check the ARP table for the corresponding MAC.
3. ping the IP address while running tcpdump -ne 
4. Open the dongle and use a USB-TTL adapter with esptool to query the ESP32 directly.
etc

I use wireguard for VPN. Pihole does the dns, dhcp

Nice to see some familiar software there  I also run Pihole on the same Pi to keep my network clean. Why not add Zerotier, then you can use your Pi as your DNS server wherever you are (and access your LAN remotely)?

Vital to being able to keep within the inverter output limits (and avoid importing expensive grid power unnecessarily) is having the means to display the relevant information in an immediate and accessible way.
That's what I told my wife, anyway.  Really, I'd just scored a bunch of old RC servos on ebay for £5 and I wanted to play with them. 😉

 
In context:

 
It uses 6 servos, a Wemos d1 mini clone and a tri-colour LED.  Dial faces are 3D printed, with a filament change from white to black (which turned out much better than I imagined, TBH).
The servos only had around 90° of sweep angle, so I 3D printed some little holder/gearbox things to get over 180° and provide a shaft for the needle to go on, plus some 'steampunk' needles.


 
All run buy the obligatory D1 mini and held together using the very latest in 'twist and tape' technology.

 
Yee-haw! 🤠
It gets fed the data from 'node-red' running on a Victron CerboGX via UDP packets over WiFi, updating as often as the data changes (approx 3 second intervals).  The LED changes colour according to state, so 'All OK' is green, 'You're getting very close to/exceeding the inverter output limit, so if you don't want to pay for grid, don't turn anything else on' is red, then blue for 'dumping power into the hot water tank'.
It's by no means silent in operation.  I originally made a single-servo version for just PV generation and coded it so that the meter movements were gradual (a kind of PID to get from one value to the next), but apparently it 'sounded like we have cockroaches' when it made it's ticky-scratchy transitions between values, so the big version just jumps between values with a 'zzziiiiiup!' sound.
It's actually quite reassuring to have an audible indication when the sun comes out for 5 seconds, or it starts charging at night.
I'd also put a little radar sensor in there, with the aim of having it detect when there was nobody around, so it could stop displaying to save power/noise.  Unfortunately, that meant that every time you entered the room, all 6 needles would suddenly move at the same time, making you jump out of your skin and spill your horlicks, so we just leave it 'on' all the time now.
There's also a basic web UI, so you can see the same data on your phone or whatever, plus it's useful for troubleshooting:

 
I had fun and learned loads making it - and now it's become a family game to try to cook dinner without making the red light come on. 🙂
 
Libraries used are 'ESPUI' and 'servo', plus 'arduinoOTA'.  Code, FreeCad file for the gearbox & stl file for the arrow-head needle are all in the attached zip.
Power_meters_v2..zip

Nice.
Once the mqtt data is in SQLite , you should be able to graph the data with Grafana.
telegraf is also able to persist mqtt data to a database.
In my case, I also used a python app to save mqtt to a database.
I did try telegraf, but the python code was already written.
I found that using Docker really helps to run the services.

So, after various DIY attempts to monitor my Axpert inverter, I settled on using SolPipLog, which is great for monitoring live data via MQTT, using a Raspberry Pi. I use that in conjunction with the HomeHabit app on my Android device. It allows me to monitor various inverter stats in real time, and also to control some functions (such as charge profile, equalization on demand, etc.). However, the one thing that neither SolPipLog or HomeHabit offer is historical data via graphs/charts. Although SolPipLog does export to EMONCMS, this is no longer free and setting up a self-hosted instance is quite painful.
I really wanted to have some sort of historical data available, so I could see trends (or check why the fans are running - did we have load shedding or is it just hot? 😅).
This prompted me to write my own software, which logs the MQTT topic data to a SQLite database, and fetches it into a Highcharts graph on demand. I got it working but got a lot more involved than I expected, so the end result is MQTT DB Logger Grapher, which is reasonably customizable, and while it's developed for inverter and power monitoring, it can be configured to log and graph any MQTT data.
Feel free to download and use with SolPipLog or any other MQTT device monitoring software.
Desktop web view:

 
Mobile web widget in HomeHabit:

 

So, after various DIY attempts to monitor my Axpert inverter, I settled on using SolPipLog, which is great for monitoring live data via MQTT, using a Raspberry Pi. I use that in conjunction with the HomeHabit app on my Android device. It allows me to monitor various inverter stats in real time, and also to control some functions (such as charge profile, equalization on demand, etc.). However, the one thing that neither SolPipLog or HomeHabit offer is historical data via graphs/charts. Although SolPipLog does export to EMONCMS, this is no longer free and setting up a self-hosted instance is quite painful.
I really wanted to have some sort of historical data available, so I could see trends (or check why the fans are running - did we have load shedding or is it just hot? 😅).
This prompted me to write my own software, which logs the MQTT topic data to a SQLite database, and fetches it into a Highcharts graph on demand. I got it working but got a lot more involved than I expected, so the end result is MQTT DB Logger Grapher, which is reasonably customizable, and while it's developed for inverter and power monitoring, it can be configured to log and graph any MQTT data.
Feel free to download and use with SolPipLog or any other MQTT device monitoring software.
Desktop web view:

 
Mobile web widget in HomeHabit:

 

So, after various DIY attempts to monitor my Axpert inverter, I settled on using SolPipLog, which is great for monitoring live data via MQTT, using a Raspberry Pi. I use that in conjunction with the HomeHabit app on my Android device. It allows me to monitor various inverter stats in real time, and also to control some functions (such as charge profile, equalization on demand, etc.). However, the one thing that neither SolPipLog or HomeHabit offer is historical data via graphs/charts. Although SolPipLog does export to EMONCMS, this is no longer free and setting up a self-hosted instance is quite painful.
I really wanted to have some sort of historical data available, so I could see trends (or check why the fans are running - did we have load shedding or is it just hot? 😅).
This prompted me to write my own software, which logs the MQTT topic data to a SQLite database, and fetches it into a Highcharts graph on demand. I got it working but got a lot more involved than I expected, so the end result is MQTT DB Logger Grapher, which is reasonably customizable, and while it's developed for inverter and power monitoring, it can be configured to log and graph any MQTT data.
Feel free to download and use with SolPipLog or any other MQTT device monitoring software.
Desktop web view:

 
Mobile web widget in HomeHabit:

 

I for one have no problem to charge only to 13.7 while under charging condition and the batteries settle to 13.4V when charging stops. I don't cycle a lot. Currently only for LS and then a pedestal fan 24/7. Normal discharge is to about 13V the next morning. 
Only 11 cycles in 5 months. 

Even if the cut off is only 9V a LiFePO4 is totally discharged at 10V and I doubt it's good for longevity. 
12V is normally 10% SOC. 

Nice clean UI you've made there, well done!
I reckon with these black box type batteries you'll need to keep an eye on the charging current when it hits bulk voltage and see how long it takes to drop to a few amps and then set your equalization time based on that. The CV part of the graph should be about the same time always, no matter how low the battery was discharged, it's the CC phase that'll be longer and the inverter does that part just fine, it just needs a bit more time at CV which can be mimicked with bulk charging time, equalization or increasing the float voltage, depending on the inverter.

@jumper thanks for the insightful reply.
No, sadly my LFP batteries are the "black box" type - there is no user interface of any kind and no feedback from the battery BMS is user accessible. I have to rely on the inverter's SOC reporting which is not ideal.
In my charge curve examples, when charging happens after a load shedding period, the reported SOC quickly climbs to 95% (a guess by the inverter I'm sure) and sits there for an hour or two. When it hits 100% is roughly when that peak starts to drop. It's never ever lingered at the bulk phase (so I never get that flat mesa).
I've just tweaked my charge settings, let me see how it behaves going forward. I've pushed bulk up to 28.8 (from 28.4), and enabled equalization at 28v every 5 days, for a max of 30 min.
I also tweaked the "back to charge" and back to "discharge values" - back to charge from 25V to 23V, back to discharge from 27 to 28V.
Doing a manual equalization now:
  
I'll keep an eye on the charging curve over the next couple of days and report any changes.

That "short" spike is still about 25 min, and the total charging time is just over 2 hours. In my example, the charge exceeds the float value (27.2) and keeps climbing for another 25 min, until it reaches the peak of 27.57V, then drops. I have my bulk charge set quite low at 28.4V. My batteries were minimally discharged since my house idles 200-300W when we are not home.
 
 
I don't think that applies to LFP type batteries, and the BMS should balance the cells internally. It appears that there is no need for absorb time, not beyond a few minutes anyway. Image source.

I'd be interested to see if the above charging behaviour is different on a system not running LFP batteries.

@zylinx I have an almost identical setup to you. I have the same inverter (3kVA, 2.4kW). Mine is the Synapse 3.0V+. It is a legit rebrand of the Voltronic VM-3000+ (MPPT). Judging by your serial number, yours is a legit rebrand also. I also have the same HA11 balancer and 2 x 200Ah LFP batteries in series.

Bulk set to 28.4
Float set to 27.2
EQ set to 28.4
I never see the bulk charge voltage go above about 27.4 during charging (as checked on the HA11 and the inverter's readings).
I've been running my setup for about the last 5 months with no issues. The only time I see the bulk voltage go near or above 28v is when activating equalization (manually or automatically) on full batteries. That is, if the batteries are still charging, and I activate equalization, the charge voltage remains where it is, at about 27.2-27.4. This suggests to me that the inverter is in fact charging at the correct bulk voltage but due to the batteries still charging, it does not reach 28.x volts. If you force this voltage on full batteries via manual equalization, you then see it rise.

In short, I think the batteries are being charged just fine. I'm open to ideas if anyone disagrees.
@zylinx You might want to check my post on voltage calibration for this inverter. I got the inverter to report the same battery voltage (within 0.1v) as my HA11 and my multimeter.

I do have one 09.xx firmware, version 09.24, but it's for an Aerox model. They have slightly different power ratings. I note that this is a Freescale firmware, unlike most VM models, which usually have a TI '28062 processor. This 09.24 firmware is for 1200 W (12V), 2200 W, and 3200 W models. So it turns out that you can't use any 5 kVA firmware for this model, sorry.

As indicated you can never get a voltage higher than Voc when the panels are not open circuit. That is the reason they give this voltage. As you draw more power the voltage will sag a bit. The higher the load at the same other factors the lower this voltage will be. Cloud edge will affect the voltage as the panels cooled down a bit while the sun was behind the cloud. 
For sure no need to worry even if you did over spec. This high output is only for a few minutes while the panel temp increases and the output drops. 
High output always goes with high amps and a bit lower volts. 

Yeah, us 24v Axpert commoners have to worry 😄 I was told elsewhere on this forum to overspec the system - my MPPT can handle 1kw (I know, I know), but it was suggested I could spec up to 10-20% higher, so was planning 2 x 600W panels. Now this has me worried...

That one day in the year when it's extra clear and your panels overproduce, and *poof* 🫣

That is a tad worrying though, if I spec Canadian panels for my system and it's on the verge of what the MPPT can handle, I risk blowing it if they put our more than what their spec sheets state...

Going 2 directions is fine as long as you go parallel which in this case you can as the MPPT range is from 30V which is below the voltage of 1 panel. Series not what I would do.