Coulomb

Possible violation of regulations, or at least norms. In South Africa, I seem to see a lot more double pole breakers/isolators than I do here in Australia. I have no idea about the SA regulations.

But then it's not isolated from the 48 V battery. If everything else is isolated, that might work. I use a Meanwell SKM30C-12 in my system; it's an isolated 36-75 V to 12 V 30 W DC-DC; it runs up to 7 contactors, a BeagleBone Black (via a 5V regulator), and the battery management system. A smaller 5V version would work for you.

Ah, my bad. I took this as an ohms reading, with the inverter off and isolated. Odd that it's intermittent, though. Maybe it depends on a load between the DB and inverter, like the light bulb (circle with an X through it) on @plonkster's diagram. Perhaps your stove, oven, or geyser turning on an off? Or just leakage? Some Axpert models only had the three relays; Weber's speculation is that those omit the inverter output relay. Since you have all single pole relays, and no setting 38, you can't do the proper switchover. I can't see the harm in shorting AC in neutral to AC out neutral. Though doing so does short out one of the "safety" relays.

Yes, but when it does, the other contact connects inverter neutral output to earth. So you get the inverter neutral output earthed either through the incoming neutral, or by relay contact. As you can see from the diagram.

It's doing it all for you. Just don't switch the neutral at the AC input; switch only the active. If your inverter can do the relay via dry contact thing, you have to enable it with setting 38. But I can't see how you'd need this.

I'd say you have one of the Axpert models that has the neutral-out to earth relay (or extra contact, really). This is a normally closed contact, so with the unit switched off, you'll read zero ohms, as you did. And when in battery mode, it's also going to be shorted. But in line (bypass) mode, the assumption is that the incoming neutral will already be earthed, and the earth to inverter-out-neutral is floating. With that in mind, the whole situation may become a lot less mysterious. The models with this neutral to earth "virtual relay" seem to be the ones that come with main firmware version 73.00. From this post:

You could start here: Can I update my inverter, and if so to what? In your case, you'll have to ignore the warnings about not attempting to update clones. I have a theory that some of the clones don't have a bootstrap loader, just a jump instruction to the main firmware. These can't be updated at all, since the loader is needed for firmware updates. It won't hurt to try; if there is no bootloader, then the ReflashTool simply won't do any updating. Cycling the battery power to the inverter should bring it back to normal. But certainly, some clones have been updated successfully (assuming that these really were clones, of course, but I'm fairly certain that they were). Good luck.

No. The Axperts are all "off-grid" (in Voltronic Power terminology), basically a stand-alone inverter, solar charger, and changeover switch in one box. I'm not a fan of paralleling inverters like this, but if you try, my understanding is that the stand alone inverter should be larger (in real power), preferably twice as big (but some would argue that just a little bigger is ok). Even then, you need some luck, because it depends in detail how the grid tie inverter checks for the presence of the mains, and it has to get the answer wrong for the paralleling to work.

Ah. On those sorts of inverters, the neutral is the centre tap, not one of the lives. [ Edit: and you want one of the "lives" to become a neutral for normal 240 V usage. ] Strange that they don't leave the centre tap isolated from the battery negative. That's what I was missing, then. Thanks.

That's the second time I've heard that today. How on earth 🤮 does connecting earth to neutral out blow them up?

The time the fan is running is mainly determined by the firmware (edit: the external heat-sinks have little to do with it). They've changed the fan control a lot over the years. Quite early ones were terrible: running to jet speed at the slightest loads. About the quietest one was I believe 72.70. 73.00 now runs the fans at a minimum of 30% (barely audible, but in a lounge room you might not like it), and they run more aggressively during the day. What firmware are you running now?

This service manual is pretty good, even though almost 6 years old now. https://powerforum.co.za/files/file/30-axpert-mks-1-3kva-service-manual-20130611pdf/ With your experience, you may well be able to fix this one.

Sorry to say, but that label doesn't look genuine: Should be 14 digit serial number, starting with 9293 for a 5 kVA model, and as @Chris Louw says above, the date code should be YYMM (yours appears to be YYYYMM) Use of the ¢ (cent symbol) instead of the ϕ (Greek letter phi symbol) Label is too close to the mounting screw, and there is no small date sticker over the nearest cover screw. The screws should be black, not silver. None of this is totally conclusive, but it doesn't look good.

When I wrote that, I assumed that the estimation happened in the inverter's firmware. But in fact, the inverter doesn't report its AC input power at all, so this inferring must be happening in ICC. I think that your clone may be acting differently to what ICC expects. Typically, AC input power would be calculated as AC output power (which is measured and reported), plus estimated charging power, if the inverter appears to be charging from utility. The inverter reports total charging output current (DC amps) and PV output charging current (DC amps), so AC charging output current can be inferred as (total_current - PV_current). Maybe the total current is reporting as zero, or too low, allowing a negative result. You could cycle through the measurements on your LC Display with the buttons, and perhaps figure out what is going wrong. I wonder if they patched the patched firmware 🤔, or the hardware doesn't work quite the same.

Since 51.8 V is the voltage where balancing starts, you might want a smidge more than 51.8 V (say 52.0 V) for the absorb voltage. Just so your're sure that balancing happens most days. On a totally cloudless day, an Axpert could exit the absorb stage a few tents of a volt under the setting. Even though most days it will overshoot, you can't rely on that 🙃

Those are the latest factory firmware versions for 5 kW (PF1) machines without the 64 V capability. Unfortunately, there is no patched firmware for these, to fix the problems that @Padwan has mentioned. If you happened to buy a 64 V model instead, it would have come with firmware version 72.20, for which patched firmware is available.

I'm not sure, but I think that Axperts work the same way, so it's inexplicable (except maybe they didn't think of a need for it). But perhaps they sync to their own AC inputs instead of ±120° from R phase (L1).

And in an electric vehicle, a single 3.2 V cell that goes high resistance can end up with about -360 V, depending on the situation. That sort of voltage can do bad things, to the cell and its Battery Management Unit. 💀

Definitely. Inverters of any kind need a decent flow of air, and they are supposed to have some 200 mm of clearance around them.

I have a theory that firmware version 72.90 only comes with clones. Could you post a picture of the sticker on the side of the inverter, or follow the steps in this post (just the one above the first photo [ edit: you should now use Do I Own a Clone? ] ). Your mention of "other funny stuff" only strengthens my theory. Edit: however, this post (in Italian) seems to disprove my theory. Sigh. If you have a genuine machine, you could update to patched firmware version 73.00e (or later, if that ever happens). It's unclear whether you will run into error 90 after 6 months if you update to factory firmware version 73.00. If you do, you should still be able to downgrade to factory firmware 72.70, or update to patched factory firmware 73.00e. If you have a clone, you might be able to update to 73.00e; some clones don't seem to have a bootstrap loader. If you need to parallel with 5 kW machines that come with firmware version 74.XX, then you can't run 73.XX, factory or patched. See also: Can I update my firmware, and if so, to what?

Kudos to Victron for that.

The Axperts can't do it. In order for utility charging to happen, they have to be in bypass mode. The inverter runs in reverse to charge the battery. As far as I know, you can't have one in bypass and the other two in battery mode. I can't immediately think of a reason why that would be impossible, but I'd say they won't do it, especially after you tried exactly what I would have and it didn't work. Edit : I assume that you gave utility power to the one that is designated as R phase, i.e. its output mode setting is 3P1. Maybe it also should be the master, so turn it on first. Edit 2: no need for a complete new inverter just to charge the battery, any stand alone battery charger would do. Though it might be more convenient to control it if it is another Axpert. But it would need separate control I would think, as it can't be paralleled with the others.

Yes. [ Edit: assuming it will update at all, of course. ]

My guess is that it coulomb counts, but there is a little drift in the current measurement, plus all the difficulties of guessing SOC, with temperature affecting it, hysteresis, internal resistance, and so on. At a certain voltage, it resets the coulomb counter to zero, and hence the SOC to 100%. So to answer your question, it was 99% but didn't know it yet 🤔. Edit: or it really was closer to 94%, and it expects you to keep it at a higher voltage until the current reduces to a low value, so that the battery really is full and the 100% SOC it reports is justified.

Grid (AC input) power is inferred, not measured. The inferring involves some estimations, particularly losses of the inverter (I forget the details now). It won't be pushing power into Eskom, so don't worry about that. I thought if its calculated power ended up negative, it would report zero. This is the big unknown. From the few photos I've seen, they seem to attempt to copy the circuit and even the PCB layout quite closely, but they may have screwed up, either to cut costs, or through not understanding how the machine actually works in detail.