Repair of Axpert Inverters : A Journey Started

[BritishRacingGreen]

Hi @Coulomb i have raised some eyebrows a couple of weeks ago claiming that there are igbts that exposes pos temp coeff in certain regions of operation.

I was looking for alternative to the max igbts and in doing so i notice the one used by max has a slight pos coeff. See snippet of datasheet below. 

[P1000]

Looks pretty convincing to me (although the current is quite high for this application):

 

Edited November 20, 2022 by P1000

[Coulomb]

Interesting, thanks.

According to the linked PDF, IGBTs have a cross-over point from negative to positive Vce versus temperature characteristics. 

From https://www.renesas.com/sg/en/document/apn/usage-notes-paralleled-igbt?language=en

As long as the cross-over is at fairly low power levels, it seems to me that paralleling devices is generally safe. But it's possible that some parts would be less suitable for paralleling than others.

So in your diagram, it seems that at low Ic, the slope may be negative, which would make them not share well. But at low collector current, each IGBT can easily handle the whole load on its own anyway.

[BritishRacingGreen]

This image below reminds me of a very old Afrikaans song : 

 twee-mosfets-met-een-skoot-middeldeur-geskiet-fontein

 

[BritishRacingGreen]

hi @Coulomb ,  I received a small 1kw 12V Axpert inverter for repair . Its an Axpert VP 1K-1000W model . Couldn't find any service related manual out there in the wild , but did find the 3KW-24V manual you posted on this forum. Fortunately the 12v machine is not unlike the 24V in terms of function blocks , components, labelling  etc. 

The battery 125A fuse is blown . So I suspected the push-pull Mosfets on the 12V side to be shorted . Alas , not , instead the IGBT full bridge is shorted.

As per diagram Q16 and Q19 shorted on all three pins . So I removed Q16 and Q19 (EDIT : note without replacing them) , supplied the 12V battery input with current limited bench supply and switched on . Now the Axpert reports F05 (over heat or output short circuit) . 

I have checked the bus dc voltage and one IGBT gate drive (Q19) as shown below (the yellow trace is Dc Bus , the green trace IGBT gate drive) :

 

When I switch the machine on via switch , the dc-dc converter ramps up the dc bus to steady state terminal voltage as shown, in short time . The CPU then outputs two cycles (20ms)  as shown . I suspect because it cannot read any subsequent output on L and N  , it assumes there is output short.  Your opinion on this ?

It then aborts the bus activity. What bothers me is the short time it takes to collapse the bus voltage to zero. In contrast , The MAX inverter I worked on will take forever for the bus caps to be discharged.

Also is it correct to assume that the reason for the original dc fuse to blow , is because of the full bridge short circuit.?

I can only assume that there is no bus soft start , there is no component section for it on the main board , and also the above trace suggest there is no initial soft start sequence . So because of the lack of this evaluation , it open up the bridge and as a consequence something gives in , in this case the fuse. 

 

Kind Regards

 

EDIT : I have checked the mosfets drive signals on the battery side . there is definitely burst control from the CPU , which suggest it is actually ramping , which is a soft start?  But then why did it not report soft start failed when bridge was shorted. ?????  

 

 

 

Edited December 12, 2022 by BritishRacingGreen
Extra info

[Andry91]

Guys I don't know if it can help you, I have a diagram of the axpert max 7.2kw model.  There are 29 pages I can't insert them here, write me on whatsapp and I'll send you the file

[Coulomb]

18 hours ago, BritishRacingGreen said:

The CPU then outputs two cycles (20ms)  as shown . I suspect because it cannot read any subsequent output on L and N  , it assumes there is output short.  Your opinion on this ?

That sounds plausible for a simple firmware, as would be found in a 1 kW model. They typically use 8-bit microcontrollers, a few rungs up the latter from the chips in some birthday cards.

18 hours ago, BritishRacingGreen said:

It then aborts the bus activity. What bothers me is the short time it takes to collapse the bus voltage to zero. In contrast , The MAX inverter I worked on will take forever for the bus caps to be discharged.

That is concerning. Maybe the bus capacitors are really small, especially compared to those for a ≈8 kW model.

18 hours ago, BritishRacingGreen said:

Also is it correct to assume that the reason for the original dc fuse to blow , is because of the full bridge short circuit.?

There certainly seems to be a high correlation between fault code 09 (in 5+ kVA models) and MOSFETs being blown. I assume that when the bus shorts, the resultant short circuit on the DC-DC converter high voltage end causes the MOSFETs to blow. I suppose it's possible that on the smaller models like this, the fuse could blow instead. On 5+ kVA models, blowing the battery fuse is almost unheard of.

18 hours ago, BritishRacingGreen said:

I have checked the mosfets drive signals on the battery side . there is definitely burst control from the CPU , which suggest it is actually ramping , which is a soft start? 

There are two different soft starts: bus and inverter. Ramping of the pulses to the full-bridge IGBTs could be because the firmware is performing an inverter soft-start. In 5 kVA models, this seems to take the form of ramping the target AC RMS output voltage by 5 V (I forget the time scale) until rated voltage is achieved. If anything is out of spec during this time, fault codes ensue and the inverter soft start is aborted.

In these smaller models, you say that there is no bus soft start power supply. I guess the bus soft start is then effectively combined with the inverter soft start, ramping up the target DC bus voltage with the target AC output RMS voltage. Maybe on these smaller models, the microcontroller has more control that in our higher power models. For us, the DSP only directs the DC-DC to turn on or do nothing. Is there a SG3525 chip, U9 on our main boards?

Quote

But then why did it not report soft start failed when bridge was shorted. ?????  

Since there is no separate bus soft start process, that explains the lack of fault code 09. I guess the primitive firmware on the limited CPU power microcontroller doesn't do much in the way of fault checking.

[BritishRacingGreen]

@P1000 your dual solder iron suggestion/practise is pure truth.  I am even able to suck the big TO247 igbt clean out of their thru holes. One on the wick, another one directly on the transistor leg. Clean

Edited December 13, 2022 by BritishRacingGreen

[P1000]

On 2022/12/13 at 7:46 PM, BritishRacingGreen said:

@P1000 your dual solder iron suggestion/practise is pure truth.  I am even able to suck the big TO247 igbt clean out of their thru holes. One on the wick, another one directly on the transistor leg. Clean

Well, then I can inform you of the next trick - swapping out one of the soldering irons for hot air works even better...

[Coulomb]

On 2022/12/13 at 6:18 AM, Andry91 said:

Guys I don't know if it can help you, I have a diagram of the axpert max 7.2kw model.

A bit late replying, sorry, but this manual is already in the files section. No need to pester @Andry91 for a copy:

 

[BritishRacingGreen]

H @Coulomb,  The 1KW Axpert inverter I  am busy with requires new IGBTs, as the existing ones have failed  short circuit. As I don't have an exact replacement , I have decided to use 75T65S devices for testing purposes. These IGBTs comes as second hand out of a MAX that had two of them blown out of a cluster of 8 devices. So I took 4 good ones and inserted into the 1KW machine , initially as free standing , no heat sink.

They get to work perfectly as far as the generated sine wave is concerned , I  have only tested under zero load conditions. But they do get hot , not runaway hot , but hot . And they all get hot , not just one or two. After a while you cannot really touch them. So because its so easy to perform rework on the small machine , I added the heat sink , but it still gets hot .

Under the same conditions in the MAX , they run ice cold at idle. The 75T65 is Trench Gate Field-Stop generation , and I cannot find  info on them to decide that their gate drive requirements are more stringent than with the other types 60N65 etc.  But what I have noticed is that on the small machine there is no negative gate drive supply for gate turn off . The buck stops at 0 volt. The turn on voltage remains good , no problem there.  So do you think that these high current devices has a very slow turn off if not driven negative and possibly causes overlap during dead time ?  I haven't tested the collector-emitter voltages at turn on to check for good turn-on yet , that will be my next step . Although the issue do not worry me , because this is just testing and i will replace these devices with in spec ones for the small machine. Its just very intriguing.

Maybe during the failure of the full bridge on the MAX  , the ones that did not fail maybe got damaged , but i fail to see that as the 4 I  am using has very similar static characteristics , eg, resistance and diode drop reading etc.

Kind Regards

 

[BritishRacingGreen]

Coulomb I have a question that I was always afraid to ask , but seeing that it is Holiday and I am sitting with glass of Red and doing a bit of nothing  , I have gained the courage to ask.

Picture my Axpert MAX 7.2 , which is an off-gridder , but cross-dressed as a grid-tier in order to provide ac blending. I infer that it will only limit grid-interaction within the boundaries of its own closet , bar the couple of watts that does slip thru  the upstream grid.

Now I have this theory, dangerous maybe, you tell me . If I cut the load current sensor to the DSP , and somehow insert a programmable gain amplifier  to programmatically alter the gain from 1 downwards as I  please.

In addition I have a digital power meter on the my residential boundary , as to measure incoming power or outgoing exported power. So I introduce a controller that can read from this bidirectional meter , and i can control the PGA gain on the machines load sensor .

Now lets say I have net zero on the meter . So I set the PGA to 1. But if I read  imported power from the meter say 1.5KW  from the grid , I step the PGA gain down in increments until I get to zero import on the meter . My rationale is that i am fooling my poor old max to produce more power than is required to deal with its own load output.

Is there a serious hidden flaw.??? Or shall i start by heating up the soldering iron so long? 

EDIT : on seconds thoughts, i think i should be going from gain of 1 and upwards instead, fooling the expert that the load is higher than it actually is, which will cause extra pv/batt power to be pushed. 

Kind Regards

 

 

Edited December 22, 2022 by BritishRacingGreen

[Coulomb]

13 hours ago, BritishRacingGreen said:

Under the same conditions in the MAX , they run ice cold at idle. The 75T65 is Trench Gate Field-Stop generation , and I cannot find  info on them to decide that their gate drive requirements are more stringent than with the other types 60N65 etc.  But what I have noticed is that on the small machine there is no negative gate drive supply for gate turn off . The buck stops at 0 volt. The turn on voltage remains good , no problem there.  So do you think that these high current devices has a very slow turn off if not driven negative and possibly causes overlap during dead time ?

This is something I'd love to know more about. Yes, I suspect that you can't just put the high current ones into the 1 kW with no negative gate drive; they won't turn off properly or as fast (is my guess).

Also, the components immediately connected to the gates (sometimes 22Ω, sometime as high as 75Ω, often 47Ω) seem to be "adjust to suit installed power devices". Also, the gate resistor is sometimes split into two with a fast Schottky diode across one, but sometimes these diodes aren't installed (and the resistor is 0Ω); I suspect again, batches of inverters might go through with one set of gate parts and one power device, and another batch might go through days later with different devices and components.

[Coulomb]

13 hours ago, BritishRacingGreen said:

In addition I have a digital power meter on the my residential boundary , as to measure incoming power or outgoing exported power. So I introduce a controller that can read from this bidirectional meter , and i can control the PGA gain on the machines load sensor .

I think that the idea has merit. You could organise it so that it never goes to actual zero power, so that you don't burp power to the house meter, which might send alarms if it goes negative (for reasons that are beyond my capacity). So you might be able to push excess solar power into the geyser, for example.

But when the geyser turns off, that's a big load that suddenly has nowhere to go except through the house meter, until it all settles down. Maybe the meter allows some leeway; my impression is that usually it's very little, and some are outright mean.

But you might be able to get advance warning that the geyser is nearly up to temperature, and use that to ramp down the "local export" power.

 

[BritishRacingGreen]

11 hours ago, Coulomb said:

This is something I'd love to know more about. Yes, I suspect that you can't just put the high current ones into the 1 kW with no negative gate drive; they won't turn off properly or as fast (is my guess).

Also, the components immediately connected to the gates (sometimes 22Ω, sometime as high as 75Ω, often 47Ω) seem to be "adjust to suit installed power devices". Also, the gate resistor is sometimes split into two with a fast Schottky diode across one, but sometimes these diodes aren't installed (and the resistor is 0Ω); I suspect again, batches of inverters might go through with one set of gate parts and one power device, and another batch might go through days later with different devices and components.

Yes , I think it may be a good thing for me to build a half-descent IGBT tester , where I can draw 10-15 amps on the load side  at hopefully 12V , but at high current  . I am not too interested to verify the max drain-source working voltage , but most interested to verify  narrow , say , 50uS gate pulses , and to verify hard turn-on , and to check the Drain-Source slew rates  compared to that of the  gate pulse. I assume by using 12V on the source-drain that the ac characterisic tests are not affected by drain-source voltage, whether that may be 12V or 600V. I may be wrong though.

Incidentally , I have found a crude and simple way to test IGBT in machine without  the risk of failure to short circuit and subsequent ungraceful destruction  , by means of inserting only half of the bridge at one time . The inverter actually produces half wave and  reports about 127VAC without error. I then  remove and test the opposite half .  Prior to all of this  of course I test the gate signals voltage level integrity . Then I switch on without any igbt ,I have found that there is a short window of time  between the end of bus soft start and when the inverter reports inverter soft start fail , to test the DSP driven gate signals.  I really want to leave nothing to the imagination when replacing and debugging IGBT's they are (a) expensive  and (b) it is extremely difficult to remove them without causing pcb board damage.

It appears to me that IGBT failure accounts for quite a high percentage of Axpert power chain failures. I reckon you may be able to shed further on this  assumption.

Edited December 23, 2022 by BritishRacingGreen
Extra info

[Shadders]

On 2022/12/23 at 8:48 PM, BritishRacingGreen said:

I assume by using 12V on the source-drain that the ac characterisic tests are not affected by drain-source voltage, whether that may be 12V or 600V.

From what I've read of various data sheets Vce doesn't affect turn on characteristics as you are almost certainly going to be higher than the saturation voltage in any application (SGF80N60 has a saturation of 2.1-2.6V @ Ic =40A for example) so 12V test voltage should be fine.

 

On 2022/12/23 at 8:48 PM, BritishRacingGreen said:

It appears to me that IGBT failure accounts for quite a high percentage of Axpert power chain failures

 Certainly most of the failures on all Axperts I've had a look at had damage to them. Roughly 3:2 I think on full bridge vs dc-dc and pretty much always battery mosfets are gone too.

 

On 2022/12/23 at 8:48 PM, BritishRacingGreen said:

I have found a crude and simple way to test IGBT in machine without  the risk of failure to short circuit and subsequent ungraceful destruction

If I'm understanding your procedure correctly you: 

1) remove all full bridge igbts 

2) power board and test driver gate voltage  -  by triggering opto I assume? 

3) probe the gate with oscilloscope to check dsp switching signals 

4) install half bridge and test  

5) remove half bridge, install other half and test

All testing is done by powering the board through the battery terminals and the controller in place?

[BritishRacingGreen]

On 2022/12/28 at 9:12 PM, Shadders said:

From what I've read of various data sheets Vce doesn't affect turn on characteristics as you are almost certainly going to be higher than the saturation voltage in any application (SGF80N60 has a saturation of 2.1-2.6V @ Ic =40A for example) so 12V test voltage should be fine.

 

 Certainly most of the failures on all Axperts I've had a look at had damage to them. Roughly 3:2 I think on full bridge vs dc-dc and pretty much always battery mosfets are gone too.

 

If I'm understanding your procedure correctly you: 

1) remove all full bridge igbts 

2) power board and test driver gate voltage  -  by triggering opto I assume? 

3) probe the gate with oscilloscope to check dsp switching signals 

4) install half bridge and test  

5) remove half bridge, install other half and test

All testing is done by powering the board through the battery terminals and the controller in place?

Hi @Shadders , I trust you are well , thank you for feedback . Yes my initial procedure I posted is somewhat confusing .

Yes , your summary is correct . The interesting thing is with only a half bridge active , the max report about 127VAC and does not raise any error. 

[BritishRacingGreen]

On 2022/12/28 at 9:12 PM, Shadders said:

From what I've read of various data sheets Vce doesn't affect turn on characteristics as you are almost certainly going to be higher than the saturation voltage in any application (SGF80N60 has a saturation of 2.1-2.6V @ Ic =40A for example) so 12V test voltage should be fine.

Thank you for this valuable info . I  hope to soon create a jig and publish it here on thread.

[BritishRacingGreen]

37 minutes ago, BritishRacingGreen said:

Thank you for this valuable info . I  hope to soon create a jig and publish it here on thread.

My current IGBT test procedure is as follows (please feed back if you see flaws or enhancement or additions which will make this a better experience) :

1a: remove shorted or failed IGBTs , by sucking the 24 lead holes clean and remove the heatsink cluster completely. On 5KW machine there is only 4 igbt's i presume as oppose to max'es 8

1. measure for short circuit on DC bus and BAT bus with all bus caps discharged.

2. Power the SPS source only , without any addition mains battery or pv power. On the max this is easy the SPS power supply is an external module which i unplug , and then feed the SPS supply with 45VDC limited to 1A max. On 5kw machines i assume the mains to SPS psu is onboard , so that will become a challenge for me.

3. Under conditions of (2) the control board should evaluate the bus soft start , and pass it . The bus capacitors  should have a residual voltage of anything higher than 300V (if i recollect).

4. If all is OK , then the inverter will report error F05 (output short circuit). I think the max thinks its shorted because output voltage AC remains 0V because of no igbt working.

5. Restart the power supply but check for valid igbt drive signals on all 8 igbt gate circuits. The timing should be 10ms max and the on and off voltage levels must be checked. on 2 of the igbts there is actually PWm activity within that 10ms .Between the time the bus soft start ends and the F05 is detected , there is only a limited window of time to check these signals. You there need to restart the power supply . 

6. switch off ,  re-instate the standard SPS supply . Connect the variable bench supply now onto the battery terminals , at about 52V and limiting the current to 4A.

7. switch inverter on , and observe no errors other than F05 (to be expected). Bus voltage should be well above 350V when all is ok. EDIT : note that this procedure only focus on IGBT failure , the assumption is made that the MOSFETS  and the BUCK IGBT is ok.

8. Switch off and insert the Upper Left and Lower Right IGBT's and  touch solder it. If if have max , then one transistor per position is fine. (EDIT : NO NO  , you should solder in the other in parallel  as well , as one will defeat the object of the whole exercise!!!)

9 Switch on , you should see no errors , but an output voltage of half the rms of full ac. For the good order verify the integrity of your 2 gate signals 

10 . Switch off , remove the half bridge and solder the opposite half bridge in place.  you should see no errors , but an output voltage of half the rms of full ac. For the good order verify the integrity of your 2 gate signals .

11. You have reached a stage where little has been left to the imagination, hopefully  , so now you can fix the igbts onto the heatsink and solder all 24 leads onto the main board, making sure that you maintain the positions of IGBTS exactly as you had it during above half bridge tests.

12 of course you should now get 220AC after switch on , and it is good to verify the shape of the pure sine wave.

 

 

Edited December 30, 2022 by BritishRacingGreen

[BritishRacingGreen]

 WARNING : Inadvertent connection of PV- or PV+ to earth = unforgiving results

I am still on this journey and truth be told enjoying every moment of it . Over the December period up till now I have been involved  with some Axpert repairs and as promised I will share my experience .  I am at times just lazy to blog, so i want to play some catchup. 

AXPERT MKSIV    

I received this machine from @Steve87 , and reported that the installer / technician had inadvertently touched the PV- to the chassis ground via screwdriver . With the AXPERT machines, this is always a problem because the top edge of the fan plate is uncomfortably near to the screwdriver . Now obviously the technician never turned the machine off . So you can see below the arc damage :

 

Obviously I did not hold my breathe for this this could be extensive damage inside the machine. So before I opened it , I powered on  via bech 50VDC supply , it booted and then raised error F35 (inverter soft start failed) . So this was already an indicator that the DC BUS is ok , but the IGBTS was blown open circuit because the AC output could not be raised .

So opened the machine removed the modules including the main board and ,alas , all 4 IGBT blown :

Furthermore , the directly underneath the IGBT on the other side it looked as follows :

ooh , I thought, this is not going to be a serviceable main board , it will need a new one. But after treatment with a bit of surgical spirits it looks like this :

Which is great , there are several faulty driver components , but the PCB is good , so it is serviceable. Just wonder about the all the deposited carbon and there was no mechanically damaged or burnt component component. Funny.

Ok , so the reason for this failure is PV1 was shorted to Earth .   The IGBT DC BUS has PV- connected to BUS- . Now the output of the full bride is AC and one leg is neutral , which in turn is connected to earth by design. So guess what , the IGBT got shorted and this was the beginning of an ungraceful destruction of the whole bridge.

Incidentally not long ago , @Coulomb and myself noted on another thread that grounding (earthing ) of PV wires is detrimental and can destroy things. So be extra careful when working with PV wires . Another good reason for PV frame to be earthed is to also prevent surges to enter your pv wires. 

I have decided from day one to not select a surge protector that protects to earth as well , because it offers a low resistance path , inviting surges to enter your pv wires. BUT PLEASE , this is not the norm as suggested by many  , this my opinion . But I stand with it . So I only provide surge protection between PV- and PV+ itself , not between PV- and Earth or PV+ and earth.

EDIT :  there is one feed in the  inverter that is isolated , that is the battery . This is because of the inevitable HF transformer required to get the low battery voltage to 400VDC dc levels. But PV is not isolated , as you can see.

 

 

 

 

Edited January 11, 2023 by BritishRacingGreen

[BritishRacingGreen]

REPAIR : MKS2  AXPERT CLONE

The owner  was woken up by a repetitive clicking sound very early on Christmas day. he discovered that his Axpert's display backlight and LED's   was cycling a number of times per second  , on then off. Also the relays was clicking at the same rate.

So he sent the machine to me all the way from CapeTown , I was surprised as the cost of courier was only R165 and it was delivered within 4 days (2 working days) . 

So it landed , and again I connected the now standard 50V current limited bench supply on its battery terminals . Awful , it clicks and clicks and clicks and clicks , and flashes and flashes , etc., etc. Neatly at a deterministic rate. I put my money on two failures, either the control board or the SMPS supply rails , 12V and or 5V.

So its time to open the unit . No physical damage . I  disconnected the MPPT and other peripheral as to only power the control and main board . same story.

So i removed everything and powered the main board on its own , without control board . Steady 12V , -12V and 5V . Plugged in the control board without display . Measured the voltages gain , 5V was cycling between 4.0 and 5.0V  , as much as the multimeter could keep up with it. 

First thought is overload on the 5V , but there was no component or section  getting hot , and neither the 5V linear regulator. So this meant two things , either the unregulated DC not enough or , the capacitor input filtering not enough . unregulated DC on scope was ok peak to peak , but there was severe ripple . 

So I soldered on a temporary electrolytic cap , and problem solved.

On the left is the faulty cap , the one on the right is a new one. It could be the green one is inferior, The new one is Rubycon , good capacitor .

The reason for the solid 5V when control board is out , is because there is also a 100nF film cap on the input , so this actually provided smoothing under no load.

The cap electrolytic has run dry , its has lost its ability to store . The machine is about 7 years old .

So I upgraded all the PSU caps as standard procedure.

I am waiting for 4 RCT Axperts arriving from Malawi next week . Now guess what !!!, all of them  4 has the same problem . They cycle the clicking at slower level , but I assume the smoothing cap is slightly better  , but obviously still at fault . I will report my findings.

 

 

[Coulomb]

Sorry, I meant to comment on this ages ago.

On 2022/11/09 at 11:34 PM, BritishRacingGreen said:

Also , ZD9 is a peculiar glass tube device , leadless. Is there any significance in this  I need to read into ?

That package has a special name, MELF, which stands for Most End up Lying on the Floor, or sometimes Metal Electrode Leadless Face. They seem to have superior reliability in some applications, and perhaps higher peak pulse power ratings.

https://en.wikipedia.org/wiki/Metal_electrode_leadless_face

[BritishRacingGreen]

UPDATE : REPAIR : MKS2  AXPERT CLONE

So once I repaired the machine , I dispatched it back to CapeTown , and the owner hooked it up again . Alas , something weird happened . The machine would power all the plug points as normal , but no lights  during load shedding. Yet when Eskom grid got back , the lights came on !!!!!!!!!!!!  Solar is not For Sissies , was my first thoughts.

So I kept my cool and reckoned if inverter is providing output power within specs ,   there must be a path from the lights that is not returning to the inverter. So I requested to measure the voltage between neatral and earth . 110VAC. I asked him to disconnect the inverter from everything and check static resistance between output neutral and earth . Nothing (!). So I am assuming now that during the phase of inverter fault the inadvertent and frequent clicking affected the bond relay and degraded its ability to bond neutral to earth.

Next I asked him to reconnect the inverter and introduced a hard bond between neutral and earth. Fortunately he has the input ac not via ELD. And the light came on!!!!!!!!!! Yeah , but why ??????

Long story short , it turned out that the lights was not connected to the output ELD, but before it. So that's why this ELD never tripped. It turns out that the lights return path was partially returning via earth  and not 100% via neutral . So he has a nasty fault in his house , and as we speak , they are investigating this issue.

The bad news is , I did not test everything during test and verification cycle. Good thing is this revealed a legacy problem the owner had.

So one thing i have learned , is to even check all ancillary / supervisory functionality of the inverter before you dispatch .

And by the way , and equally important , it is my opinion that as many , if not all circuits should go thru ELD .Earth return is for fault current but which can be detected , not a long term return path. At home I even have my inverter AC Input on ELD , not to protect myself from harm , but to allow it to detect leaks, and low and behold , I get trip on extreme rainy burst, which means i am leaking on the roof to pv. Remember this : we all talk about nuisance tripping , but what prove have we got the ELD is not telling somethin important.

 

Next up : there is 4 machines on its way from Malawi !!!!!!!!!!!!!!!!!!!!!!

 

EDIT : the owner also informed me that in the past he experienced being shocked by plumbing and metal structures . This to me is something to do with a combination of earth fault current and bad earthing .

 

 

 

 

 

Edited January 25, 2023 by BritishRacingGreen

[Coulomb]

11 hours ago, BritishRacingGreen said:

So I am assuming now that during the phase of inverter fault the inadvertent and frequent clicking affected the bond relay and degraded its ability to bond neutral to earth.

Did you end up with one screw left over after re-installation? That seems to often happen, either that or being one screw short.

If you left out one of the critical screws holding the main board down, one with am earth in a circle symbol near it, as well as a screw symbol, then that would explain the lack of neutral to earth connection.

Like the one on the right in the photo below:

It seems that in that part of the PCB, they need all the track area they can get for the heavy current carrying tracks, so they rely on the screw and stud for the neutral to earth connection. Some say that they did this deliberately so that the neutral to earth connection could be disabled by leaving out that one screw.

If that's the case, and the customer is willing to do some disassembly to get to that screw, you might be able to get that fixed without shipping back and forth. Or maybe the customer knows someone handy with screws, but not with electronic repair. BTW, repair of these inverters seems to be keeping you busy!

Edited January 26, 2023 by Coulomb

[BritishRacingGreen]

9 hours ago, Coulomb said:

Did you end up with one screw left over after re-installation? That seems to often happen, either that or being one screw short.

If you left out one of the critical screws holding the main board down, one with am earth in a circle symbol near it, as well as a screw symbol, then that would explain the lack of neutral to earth connection.

Like the one on the right in the photo below:

It seems that in that part of the PCB, they need all the track area they can get for the heavy current carrying tracks, so they rely on the screw and stud for the neutral to earth connection. Some say that they did this deliberately so that the neutral to earth connection could be disabled by leaving out that one screw.

If that's the case, and the customer is willing to do some disassembly to get to that screw, you might be able to get that fixed without shipping back and forth. Or maybe the customer knows someone handy with screws, but not with electronic repair. BTW, repair of these inverters seems to be keeping you busy!

Thank you for your pointers. There may be an excellent chance that a missing screw could be the culprit. Yes the owner will assist me in a remote debug. I see there are also surge protection components that relies on this bond, which worstens the situation. 

Yes, Coulomb the repairs are keeping me very busy, i am enjoying it. Maybe I have just written one too many lines of software code the last decades,  and reverting back to my roots in engineering is so refreshing,  especially now that i am semi retired and have spare time to do so. Also in the spirit of so msny things ive learned from you in such small time, i have decided to start paying forward and help others where it is possible.