I've blown a very reliable 'classic' PJ inverter 🙁

8 hours ago, Paul said:

Well its been a bit of a sad day as I was completing ugrades to our little off-grid home in rural Uganda. I had took a classic PJ 3K 24V - 230V which had worked reliability for us in the uk before I upgraded there to a larger inverter last year.

The little inverter worked perfectly until I accidentally dropped a tool when working above it on something else. Foolishly I hadn't got around to insulating the DC terminals and the tool made contact between the +24v terminal and the case. There was a very small spark and the inverter cut out.

Now when switching on, the green LED comes on, there is continuous beep for a few seconds, then the fan operates for a few seconds, then it shuts off.

There is no drive to the FETs but none appear blown and their green LEDs are on.

The only anomalies I can see are the 5v regulator on the control board is extremely hot and the MCU is hotter than I would expect. The 5v regulator (the wonky TO220 package with a small aluminium heatsink) measures 12v on its input pin and 5v on it's output pin.

I note that there is an 'earth' terminal on the control board which has a connection to the chassis via the AC outlet (yellow wire), so I guess this is how the damage occurred, my incompetence aside! I have not investigated how or what this earth terminal connects to yet.

Anyone have any idea what else I can test to determine whether it is repairable or will have to be consigned to the PJ scrap bin? This would be a real shame as this inverter powers everything we need in this little house and happily sustained continuous loads of 2000w, which is better than the new generation of PJ/upower 6k inverters!

//content.invisioncic.com/g308908/monthly_2022_12/20221215_121249.thumb.jpg.b72e93673b9fd192132b94acd7668a96.jpg

//content.invisioncic.com/g308908/monthly_2022_12/20221215_121303_1.thumb.jpg.d48804fe01edeede4d3dbc5483ac4eca.jpg

//content.invisioncic.com/g308908/monthly_2022_12/20221215_121319.thumb.jpg.a897930bde9756529cda48cb9fd21c4a.jpg

5v on the output of the 7805 (plus the lights coming on then going off again) would seem to indicate that the main CPU is not damaged--though obviously something connected to it isn't working right (as evidenced by it getting hot).

Normally if a microprocessor gets very hot when powered up, it'll either short out the power supply rail (i.e. 5v rail reading ~0.6v)...or it just won't do anything at all.  Your case with the 5v rail being fine (if seriously loaded!) and the fan/buzzer switching on and off (by CPU control) is quite unusual, to say the least!

Unfortunately, I really can't provide any further diagnostics--as PJ inverters do not provide any diagnostic info, and I am not very familiar with their design anyway.

I honestly have a hard time believing a DC-to-chassis short could be responsible for an issue like this.  The fact that there was a small spark indicates that yes that incident probably was the cause--but there should be no possible way for any significant amount of power to flow between the DC and AC isolation.  Unless the main transformer isolation has failed over time?

Because the ground terminal on the control board literally just goes to several 10nF (0.01uF) decoupling caps to the AC output rails--per U.L. grounding requirements.  There is no way that a main battery DC potential should cause any sort of damage when connected to the chassis--or any output lead.  (Or else the inverter would have been fried long, long ago!)

Thanks Sid for the reply. Now you have clarified that the 'earth' terminal is just capacititively coupled to the AC output rails it also seems improbable to me that the device would die simply from momentarily having 24v DC connected to the  chasssis 'earth'

The inverter had been working well up to this point, powering loads such as an induction hob, kettle with no abnormal behaviour or noises that would suggest transformer breakdown. I had also been using a clamp meter earlier to measure DC current when AC loads were applied and the DC figures were as expected for the given AC loads - typically 85-90% efficiency so I don't think there is an issue with the transformer. Presently, the control board does not even send any signals to the MOS board, and disconnecting the MOS board cable, nor the AC output leads from the control board makes no difference to the behaviour at all.

I will have to take it back to the UK and do some more thorough investigation with my scope.

Here in Uganda the only inverters one can get are ridiculously overpriced Victron units, or much-worse-than-PJ cheap chinese HF rubbish.

I didn't sleep at all the night after I broke this inverter as we hadn't been to Uganda since late 2019 and the off-grid solar was the final part of the build completion as it would mean we could cook using the plentiful free electricity from the sun rather than contributing to local deforestation by using charcoal, or using bottled gas which required many miles of driving to exchange when empty.

We don't spend a lot of time here - we mainly live and work in UK and my wifes relatives look after the land and animals in our absence, but now we have a baby and a toddler being able to boil water/milk and cook quickly is important.

The only saving grace is that at least I wired all the lighting to run off the 24v DC output of the charge controller and there are a couple of USB ports on the charge controller for charging stuff.

It is possible that the CPU got damaged--but unfortunately, with no diagnostic info, it's basically impossible to figure out what's going on.

Yeah, I've shorted each output AC terminal of a PJ inverter to either of the DC input terminals without any issues.  So that's why I'm suspecting a transformer insulation breakdown to be the root culprit--which would not necessarily impede the normal operation of the transformer.  It just would mean that the DC input and AC output would no longer be isolated--and a short between DC input and AC output would result in significant power flow (and likely some damage).

But that doesn't explain how chassis grounding could have caused something of this sort.

It is possible that the CPU got damaged--but unfortunately, with no diagnostic info, it's basically impossible to figure out what's going on.

The picture show  all the mosboard LED is green  so all the FETs are  good  but can test with a  multimeter  easy  .     That  leave the  control board  which is not  available      or the transformer  which Paul say is good .       Only the rev 11.3  control board  is  available  and the riibbon  do not match .      

UNLESS....it's possible that if the transformer mounting to the case is a bit too firm (or resulting from jostling/shaking during any shipping of any sort), and the enamel has been worn off of one of the transformer coils...that would connect DC-side power to the chassis irrespective of grounding.  And in THAT case, you'd get a spark/instant shutdown if either battery line was shorted to the chassis.  And in THAT case, damage to FETs is extremely possible...and if they fail, the damage trail can easily go right back upstream through the totem pole low-side FET drivers, through the gate control circuitry, and right into the CPU itself.

Gotta love a FET driver circuit that isn't electrically isolated....GS inverters use a fully isolated/protected FET driver circuit, and damage of that sort never goes beyond the FETs and/or driver board.  (DC bus overvoltage incidents are an exception to this, though!)

If you have a DMM (digital multimeter), you can do a quick continuity check between the AC output terminals and the chassis.  There should be no continuity between any of the AC output terminals and the chassis.  (OK, maybe 1 megohm or so--but that won't cause a spark!)  This likely won't be the issue.

If that passes, you can do a continuity check from either of the smaller FET heatsinks (where the big transformer wires attach) to the chassis--there should be no continuity.  I'm kinda betting there'll be a dead short here--which would completely explain everything.

...if there's a dead short on the FET side, but open circuit on the AC output side, then the transformer isolation is good--just there's mechanical damage to the outside windings (which is the FET-side tap; AC output windings are the deepest layer).

The poor little thing did endure a 10 hour flight with transfer (although well packed), so what you have said Sid is very possible. I have dmm with me so will test today and report back

Well, I found no low resistance paths or shorts between AC outputs L/N and chasis.

Measuring between either FET board heatsink for FET board mounting screws gives a similar result of ~2M with DMM in resistance mode, and oddly 1.3v in diode test mode.

If I remove the 'earth' lead from the conrol board, then there is infinite resistance and no registration of any diode voltage between all FET boards and heatsinks.

Something that did seem odd to me was this component called 'D27-1' in the middle of one of the attached photos. It has no meaninful markings so I unsoldered it from the control board to test it in isolation. It has infinite resistance, and does not register anything when DMM is in diode test mode, in either direction. So I have no idea what it is or what it's purpose is. One side of it appears to be connected to the 0v terminal of the 5v regulator (LM7805)

Unfortunaltly, I note that the CPU becomes too hot to touch within just a few seconds of powering on, so I suspect it is fried and will have to be consigned to the scrap bin sadly.

<a class="ipsAttachLink ipsAttachLink_image" href="/monthly_2022_12/275465605_20221218_155305(1).jpg.367b93b6c0873a1e9f95cb76748bffb0.jpg" data-fileid="1974" data-fileext="jpg" rel=""><img alt="20221218_155305 (1).jpg" class="ipsImage ipsImage_thumbnailed" data-fileid="1974" data-ratio="79.28" style="height:auto;" width="946" data-src="//content.invisioncic.com/g308908/monthly_2022_12/172125253_20221218_155305(1).thumb.jpg.3a4520f0f3356f0cc88cb9d1cb7ef2bf.jpg" src="/applications/core/interface/js/spacer.png" />

<a class="ipsAttachLink ipsAttachLink_image" href="/monthly_2022_12/20221218_162222.jpg.7c6f44c5635475c3b6eaef32e0d6e4b4.jpg" data-fileid="1975" data-fileext="jpg" rel=""><img alt="20221218_162222.jpg" class="ipsImage ipsImage_thumbnailed" data-fileid="1975" data-ratio="222.55" style="height:auto;" width="337" data-src="//content.invisioncic.com/g308908/monthly_2022_12/20221218_162222.thumb.jpg.0b87a6547c3684d6ea5f408e6629a178.jpg" src="/applications/core/interface/js/spacer.png" />

<a class="ipsAttachLink ipsAttachLink_image" href="/monthly_2022_12/20221218_155246.jpg.501f1c4ca40c74b6a9fc7abc8a2dc54a.jpg" data-fileid="1976" data-fileext="jpg" rel=""><img alt="20221218_155246.jpg" class="ipsImage ipsImage_thumbnailed" data-fileid="1976" data-ratio="70.8" style="height:auto;" width="1000" data-src="//content.invisioncic.com/g308908/monthly_2022_12/20221218_155246.thumb.jpg.1e1440a2f67c7e0a7482ce8f7fd8263f.jpg" src="/applications/core/interface/js/spacer.png" />

<a class="ipsAttachLink ipsAttachLink_image" href="/monthly_2022_12/20221218_153848.jpg.4cb80cbad687ae6bc0e37324c9eb0d0f.jpg" data-fileid="1977" data-fileext="jpg" rel=""><img alt="20221218_153848.jpg" class="ipsImage ipsImage_thumbnailed" data-fileid="1977" data-ratio="97.91" style="height:auto;" width="766" data-src="//content.invisioncic.com/g308908/monthly_2022_12/20221218_153848.thumb.jpg.faf2af87364c5abfefecba9472ae1663.jpg" src="/applications/core/interface/js/spacer.png" />

<a class="ipsAttachLink ipsAttachLink_image" href="/monthly_2022_12/20221218_153829.jpg.c31923e9ea27786d02fb8455cc97ab3b.jpg" data-fileid="1978" data-fileext="jpg" rel=""><img alt="20221218_153829.jpg" class="ipsImage ipsImage_thumbnailed" data-fileid="1978" data-ratio="97.91" style="height:auto;" width="766" data-src="//content.invisioncic.com/g308908/monthly_2022_12/20221218_153829.thumb.jpg.98491e836586fc4f785ac79e38b9161a.jpg" src="/applications/core/interface/js/spacer.png" />

Something that did seem odd to me was this component called 'D27-1' in the middle of one of the attached photos.

IT is a diode  and  if it  can be replace   and maybe work .     I never repair a  control board   but just replace it with  another one   .    

Yeah, I'd agree with you there.  If it was just getting a bit hot to the touch after 30 seconds or so, that could be explained by I/Os being shorted/overdriven by failed circuitry.  But if it's getting smokin' hot, then it's probably damaged.

On the (undamaged) v3.6 control board I have, it's marked "16C".  That in an SMD database appears to cross-reference to a ~16v TVS (transient voltage suppressor) diode, roughly corresponding to the generic SMBG16C (bidirectional 16v TVS): https://www.electronica-pt.com/db/smd.php?ref=16C

Why it's bidirectional instead of unidirectional is beyond me--as this means it won't protect against negative transients, at least from the 0.6v absolute max typical on most chips!

Even if it wasn't blown apart, your meter would not register any voltage/continuity in either polarity.  FWIW according to a continuity check, said diode (D27-1) is across the input leads of the LM7805 voltage regulator.

Just thinking of ways to make a 'usable' inverter again out of this - could I ditch the PJ boards completely and use one of these to drive the existing transformer:  https://www.ebay.co.uk/itm/124495744402 . Input capacitors look a bit on the small side but I could add the big ones from the old PJ board close to this boards inputs to help.

I see them and similar looking boards on aliexpress but heard a lot of people saying stuff from them not turn up and difficult to get refund.

Yes, you likely could replace (most of) the PJ board stack with that one.

The main issue with those boards (based on the EG8010 ASIC) is that I do not believe they are capable of running the sine wave into distortion.  That's only a problem because PJ seems to like running a high transformer primary voltage (generally 18vAC for 24vDC input -> 230vAC output inverter), which gives you less than zero headroom for transformer losses.  Yes, it seemed to reduce PJ inverter blowouts--but only because the FETs were switching less and less of the AC cycle!

mathematically, 24v / square root of 2 (1.414) = 16.97vAC absolute maximum possible output at full sine from an EG8010 board.  Assuming a transformer ratio of 12.7, that'll give you 216vAC output maximum at no load with a 24.0vDC input from the battery bank.  And that output voltage will sag quickly under load.

Just thinking of ways to make a 'usable' inverter again out of this -

NO NO  I   post in this forum that  it  work for  100 watts with my ASL 9.0 transformer  .         The voltage spike by high inductive load on shutdown will kill the FETs      The over current  protection  do not protect the EG8010  but will kill the FETs on shutdown .               The new rev 11.3 control board with a  rev 11.3 complete main board and matching ribbon cable will  work  if your  transformer primary is between 16vac to 18vac .    The cost is 300 dollars for the 2 items from Powerjack in China  and  ask that they  test before shipping .     I bench test  with 4  FETs  and should have 2  LED on steady  before  using .     

lotta nonsense.

It's possible that the EG8010 boards could be more susceptible to damage resulting from disconnecting of the battery--but that would be only because the greatly reduced DC bus filtration caps.  Any transformer will "kick back" when suddenly released--and if that "release" is not controlled, it can easily cause system damage/failure due to an overvoltage on the mainboard's DC bus (i.e. >200v for a few mS on a 48v system).

So in summary, it might work, but not very well. As this is only a piddly little '3kW' unit anyway it's not worth buying the boards. I'm back in the UK now so I'll just keep an eye out on ebay for another working secondhand unit of similar spec to take out when I go back to Uganda in a couple of months. This one was originally purchased secondhand (but barely used at all) for £130