Coulomb

Agreed. That should be OK, but you'll eventually want some charging to happen, see later. That's a bit weird. With 147 V zero to peak, I'd expect about 147/√2 = 104 V RMS, or 93 V average modulo (as a typical lower range multimeter would read). Plus a volt or so in each case for the two diode drops. But at low power you'd expect the current waveform to be a bit wonky, so it's not a huge surprise. That's deliberate. You want the mains current to be closely proportional to the mains voltage, so that the mains sees a basically resistive load. In other words, this is a power factor correction system. The battery is the ultimate filter "capacitor", plus there is the nature of the flyback circuit. The "transformer" TX5 is essentially a multi-winding inductor, with an air gap in the core. So when the transistors (Q20-Q22) are conducting, the inductor is fluxing (like charging a capacitor, except it's magnetic field instead of electric charge). During this time, there is no load on the output, as the output diodes are reverse biased. Energy builds up in the air gap. When the transistors turn off, the magnetic field has to "go somewhere", so the output develops an EMF (like a voltage) with the polarity such that the diodes are forward biased, and the magnetic field reduces as magnetic energy is transferred from the air gap into chemical or ionic energy in the battery. When smoothed by the inductance of the "transformer", this looks like a pulsing 50 Hz current into the battery. Even though the battery is very much DC, it has no problem with half-sinusoidal current pulses. In fact, it might even be good for lead acid batteries, tending to cause a small amount of desulphation. For the mains current to be in step with the mains voltage, the DSP has to be monitoring voltage and current, and adjusting the pulse width of the gate pulses such that the pulse width has the appropriate 50 Hz (arguably 100 Hz because each half cycle is the same after rectification) pattern. Amen, brother. Usually, the problem is that the gate pulses are not getting to the transistors properly, so that one or more transistors continues conducting until the core saturates. The inductance of the primary will decrease dramatically when the core saturates, making the primary essentially a short circuit (or a very low resistance, basically just the DC resistance of the coil). This is likely only a few ohms, causing many tens of amps to flow and burning out the transistor. Since Q22 seems to be the one blowing all the time, it's likely that its gate components, or something about the printed circuit for that gate, that is faulty. It would be good to compare the gate voltage of Q22 with that of one of the others. Probably you should add a capacitor between gate and source of Q22's pads to simulate the gate capacitance. Otherwise, its gate diode (D57) will never conduct. Even with this, there won't be the drain to gate capacitance of the real Q22 (I suppose another capacitor could simulate that), so it will never be perfectly matched. But you should see if anything is drastically wrong. To choose the capacitor to simulate the gate capacitance, look up the datasheet. You might have to figure it out from the gate charge. Or you could just guess it at about 1-3 nF. The gate diode is to cause the gate voltage to turn off more quickly than it turns on. Are you testing with a real battery? A current limited lab power supply is a good idea initially, but as some point you need a real battery and AC charging enabled (at lowest current initially!) to get the DSP to generate any gate pulses at all. You might want a 24 V incandescent headlight in series with the battery to limit current.

A partial schematic. But it happens to be one of the better (more complete) ones. Probably published before the clone makers made use of such schematics. As you say it's very useful, especially for these models that have a separate AC to battery charger; most higher power models use the DC→AC converter ("full bridge") in reverse. Yes. Wherever MOSFETs are in parallel (as Q19-21 are), and one fails, it's best to replace them all. If in a push-pull arrangement (this is not, it's a flyback arrangement), it's best to replace the MOSFETs in the same "pole" as the damaged one(s). Sometimes that means replacing 8 transistors (e.g. on the battery side of the DC-DC converter). But that means that the other gate drive passives (resistors and diodes) are suspect, as a MOSFET often fails shorted, including drain to gate. Those were likely OK as they would only carry rated current or less when the switching MOSFET(s) failed. My guess is that the new MOSFETs other than Q22 are probably OK, since you limited the power with the incandescent lamp. I would check (again?) all the gate driver passive resistors and diodes. Also the other components in the unusual snubber(?) arrangements, like D24, D25, D13 and all the passives connected to those. If you have an oscilloscope handy, it might be useful to monitor the gate drive pulses, while you have say a 60 W lamp in series with the AC input, but beware of earthing issues. You may need to use two channels and the subtract option.

More like 300-500 kVA here in Australia. The "low" (230 V) voltage system is way less stiff than the high voltage grid. Though I believe that voltage rise even in the HV system becomes a problem with high PV ownership (around 30% here) and with high solar insolation at times. I'll agree that things may be different in other parts of the world, and I have no idea how bad voltage rise typically is in South Africa.

Can you post a link? There are many schematics, and I can't follow along without the same parts designators.

Ah. Then this may be an Infini (on-grid, in Voltronic terminology), not an Axpert (off-grid). They often (not always!) use a different protocol for most commands. Are there a lot of "^P" (up-arrow P, without the quotes, for queries) or "^S" (Setting commands), or "^D" (Data responses), all followed by exactly three digits, then (in the case of ^P or ^S) one or more letters? If so, that's the P17 protocol: https://powerforum.co.za/files/file/42-infini-solar-5kw-p17-protocol-20160309pdf/ Unfortunately, the above is quite old, and doesn't have the HSTS command, if indeed it's a command. If there aren't many up-arrows in the sniffed data, then it could be standard H18 ("ascii") protocol, with commands like QPIGS etc, with a few extra commands like GOLF (set Grid Output Frequency Low loss point) needed for exporting power to the utility. You could check several of the protocol documents in the Files section of this forum.

I'm not familiar with the HSTS command. So I suspect that this may be a sniffing error, or the traffic is encoded somehow. Here is an example protocol manual:

Please check for a private message.

Yes, I was just looking that up. When you say "Axpert V", for a moment I thought you meant Infini V models, which are very "Axpert like", except that they can export to the utility. But you also mentioned Watchpower, which is used for Axperts only (SolarPower is for monitoring Infinis). It might be better to refer to these models as "VP" models, or perhaps "SOL VP" models. The "VP" seem to to indicate PWM solar chargers, with 105V max PV input, and the panels pegged at battery voltage when bulk charging. "M" is the equivalent code for MPPT models (145 V max PV input, with Maximum Power Point Tracking). Apart from solar charging, the VP and M models are very similar. Yes, I agree with this. In fact, I have a similar situation here, with two PIP-4048MS (roughly equivalent to a SOL-I-AX-5M). The mains voltage here is almost always over 240 V, so I've switched to an output voltage of 240 V as well. I did this (as far as I remember) merely to reduce the perceived thumping that the relays encountered when the mains and output voltages differ significantly. I've actually replaced my relays, but sadly haven't gotten around to a post mortem on the old ones to see how bad (if at all) the old ones were. I don't recall running into fault code 51, but it could be. I just found a note in my phone about fault code 51. Eceman (who used to post on this forum regularly) says that a 2n2 capacitor across the MOSFETs fixes this fault code. It's not clear to me how this would help. He also says that an LM339 comparator on the control board has cured some 3 kVA models . I know that these cause restart cycles (where the power supply voltage collapses and the DSP resets or glitches), causing the LEDs to flash. I guess that with chaotic power supply voltages, all sorts of intermittent fault codes could be thrown up, including fault code 51.

It could also mean that the flats have solar panels installed, and the flats are exporting power to the grid. This often pushes up the grid voltage nearby.

All I have is my firmware update instructions. The power source probably doesn't matter; I would use just the battery and turn off all other sources. I had another look at the 78.xx firmwares just now, and I'm more confused than ever. Hopefully EASun have gotten it right. Good luck.

You have to use the supplied RJ-45 to RS-232 cable. If you somehow don't have it, the pinout is here. The RJ-45 connector is NOT an ethernet port. Most PCs will require a USB to RS-232 adapter. See for example the discussion in the firmware upload instructions, The serial ports operate at 2400 bps. The protocols are documented in various protocol manuals, e.g. this old one. But most people will want to use a program to automatically send and receive those commands, e.g. WatchPower, SolPipLog, etc. The above is all assuming that your EASun is an Axpert made by Voltronic. If it's an early EASun that was a clone, then all bets are off.

Are you changing to 240V with the "power" (really a "load") switch off? Otherwise, the command is ignored. The other possibility is that the firmware isn't saving the voltage setting into EEPROM, so it's there for next time the inverter powers up from nothing. Perhaps change a setting like battery float voltage, then setting it back again? Then power the inverter down completely, power up and check that the 240 V setting has "stuck".

Sorry, I have no 09.xx firmwares at all, apart from one for an Aerox branded one, and it's 1200W minimum.

Yes, as long as they are close in release date. If you post version numbers, I can check.

? How can you have 9.4A flowing with no load connected? At no load, the DC->AC converter is powering just the C of the LC filter, so it will have a very low power factor. That's to be expected. If you mean that you are measuring 9.4A into the AC-in port, then that's very high. Usually it's about 300 VA for a 5 kVA model (about 1.3 A, nearly all reactive). That sounds like the inverter isn't quite in phase with the incoming mains. A totally unsuitable firmware might do that, I suppose.