Tejota wrote: ↑Thu, 08 Mar 2018, 04:07
[ Many photos... ]
Thanks for the photos, Tejota.
Some notes:
- The main (U1) firmware is 72.00. I'm guessing that this is effectively "72.100", hence later than 72.70. I wonder what "parallel compatibility group" it will belong to. [ Edit: we now that 72.00/10/20 is separate and later than 72.60/70/90. ]
- The SCC firmware is up to 04.12. I wonder what updates are in there, if any.
- New parameter Ev (34), is presumably the maximum tolerated battery voltage before you get an error. Nice to see that adjustable.
- There is another new parameter (39), with no name at the left, and the present value is AdS. Do you know what this parameter is for? Did the inverter come with an up to date paper manual? [ Edit: Duh. Not new parameters. They are described in the 5048 manual, as Weber reminds me below. ]
- It's good to see the fans installed right way around (blowing hot air upwards) (thanks, @PaulVK!)
- The capacitors have changed from the originals with the gold colour; obviously the original 63 V models would not do.
- The IGBTs seem to be different part numbers, but that doesn't mean much.
coulomb wrote: ↑Wed, 07 Mar 2018, 06:23
Does anyone have first hand experience of this?
Yes. I lost 4 (four) pip4048ms on cluster. Full disaster. I dont know what happened....
Ouch!
Working conditions: 1 pip4048 (master) manufactured August 16 (no heatsink on top) + 3 Axpert 5048 (4kW) (slaves) manufactured Mars 2014 (heatsink on top).
I assume that the slaves would have been Axpert
4048s, being 4 years old now.
At midday global solar production about 9 kW 160A (40+40+40+40 more less) to battery...
So: battery voltage might have been near its highest.
A little cloud effect.... not important.
Well, it might be. Pips/Axperts tend to overshoot the desired battery voltage by sometimes over a volt for several seconds.
I only have dissambled one shorted slave: 12 mosfets shorted, 2 mosfets fifty fifty, and 2 mosfets correct. Buck IGBT (2) shorted !!!
Interesting about the buck IGBTs being shorted. That seems to indicate that there was a surge that travelled from the inverter circuit back to the battery circuit. So maybe something happened to the inverter of this slave, surge blow up MOSFETs at the DC side; the leads of the failed MOSFETs all fused open circuit (hence it was able to start up and show an error 09), and when the battery terminal voltage suddenly increased, the inductive kick-back sent a spike of voltage to the other three battery terminals, killing all their DC side MOSFETs (and possibly more).
No visual destructive effects.
Does this refer to the buck MOSFETs only, or to all? My theory above relies on open circuited but failed MOSFETs, which usually means the leads are blown, a very visual effect. You did say this was a "shorted slave", so I'm guessing you means none of the MOSFETs or IGBTs had visible failures. In that case, I really can't explain why the buck IGBTs blew. Is it possible that the two failed IGBTs were not the buck devices (Q31, Q32) but instead were two of the high-side DC-DC IGBTs? (See my
Power Topology post.)
Any explanation of this disaster ?? Mosfets and capacitor were original on failed units.
Summers in north east Spain are about 2°C higher than Brisbane Australia, and we've had inverters fail after a few years when the capacitors are factory originals. The factory capacitors are rated for only 2000 hours at 105°C, so Spanish summers would take their toll. So that's my guess.
Ps: Yes I bought an axpert 5048 pf1 64V until I repair failed units if I can.
If you don't already have one, consider buying a desoldering station; it will be very useful for repairing 4 failed units. Weber bought one a while ago, and it's been wonderful for removing those endless MOSFETs, and sucking out solder from capacitor holes. Good luck, and don't forget to check out all those gate driver parts as well.
[ Edit: Added ", if any" to "I wonder what updates are in there". ]