Repair of Enecsys SMI-D360W-72 micro inverter

[induzer]

I had read and heard of some issues with these micro inverters of Enecsys before. The company had to close in 2015, if I remind right.
I hoped not to get part of this trouble the upcoming time. Well, for the law of Murphy...

It started just 2 weeks ago, I was awakened by my UPS at 5am in the morning.
Long term short the pv-micro-inverter had tripped one of the 10A MCB's for my Flat, the 6A MCB for the inverter AND (as I should see later) it's internal T3,15A fuse.

I had got it, 8 years old, for about 100€ 3 weeks before.

Looked like some low-resistant-something at the grid-involved parts.
I found a blown high-side Mosfet at the grid-h-bridge, after replacement I decided to start the grid "soft". Right decision, the ac-fuse tripped as soon the inverter got power at the DC-input, with 5 Volts at the AC-Input.

Thermal imaging showed me that the blown Mosfet had killed it's Gatedriver-IC, it statically provided high-level.
As this happens often when a Mosfet dies and the Gate-isolation get conductiv, I also had ordered the gate-drivers along with the Fets from Mouser...

Now I could start testing the inverter with full voltage at AC-Input, without blowing the fuses. I hoped so...

So I connected my 40 Volts DC-Source to the inverter, adjusted my variac to full grid voltage and took a step back. 

First these inverters consump about 30 mA. If the booster-stage for the internal 380VDC-Link starts, the current increases to about 70 mA.

Now nothing further happened ... good !!!
YES, because my variac was conncted to an galvanic separating transformer. And the anti-islanding-protection just did its job.

So I removed the secure transformer and took a greater step back. AC -> Go!

After 3-4 Minutes "security-timeout", which the devices have build in, it increased to 550mA DC.
Sadly on AC-side my Siemens PAC3200 just detected 2 Watts active power 

So it was time for the thermal cam again. The driver of the Fet which is seated in between the DC-Link and the Chokes directing to the grid-bridge just had the same shitty "heat-footprint" as the killed driver I had found at the ac-bride-circuitry.

But there was a second strikingly red thing at the thermal image, I earlier had mis-identified as part of the  internal power supplies. It's an OP-Amp near to the MCU. 40°C just for measuring something... no sir !

Thankfully I got the spare delivered the next day from an nearby distributor.

Replaced the OP-Amp. And again ... DC 40V limit 2A -> Go!, AC -> Go! ... take a step back, gripping the emegency-off button 

The DC-Source's (fan) get louder and relays click happily (current limiting), while the inverter tries to find the MPP of the DC-Source 

A look at the PAC approves ... 80 Watts active power straight to the grid 

Will reassemble the unit the next days, 'cause for real testing I prefere propper cooling.

I can't determine why the unit originally failed. But I'm suspecting the common problems with the power electronics or semiconductors.

All in all the repair took me 7-8 working hours and ca. 25€ for parts(excl. shipping ... don't ask   ).


induzer

[induzer]

Some words about the inner circuitry of the Inverter.

The logic-and-control section (e.g. ZigBee-COM, MCU, gate-drivers) are powered by the DC-Input via a dc-dc buck-converter.

The PV-DC gets boost-converted by a MPP-tracking DC-DC circuit to the 380V DC-Link.

In my opinion, the Fet inbetween the DC-Link Capacitors and the big double-choke in direction to the grid-bridge is for proper anti-islanding reasons.

Because of the statical high-driving the gate-driver-ic of this Fet does, it had the same heat-footprint as the defectively static driving ic for the grid-bridge.
But I had seen others of the inverters where this IC was blown physically visible.
I got irretated by this and accidentially replaced the proper ic 'U21'

The Gate-driver-ICs are often used IRS21814 by Infineon.

Near to the MCU, which by the way is a "high-end" device of the Microchip PIC-series, are placed two 74-TTL Quad-AND ic's and 2 Analog Devices AD8616 R2R Dual-OP-Amps

As far as I can determine the "mpp-tracking booster" does galvanic separation.

So far ...

Hack on,

induzer

[monkeyisland]

The fuse and the mosfet blew after replacement despite no continuity.  I don't have temp camera,  but now with your help at least I know where to look at, Thanks for your input