SMA 7 kW Photovoltaic inverter repair

[webgiorgio]

The inverter is 7 kW single phase, transformer-less, SMA Sunny Mini Central 7000TL.
Datasheet: http://files.sma.de/dl/5695/SMC8000TL-DEN102330.pdf
Attached photo of the inverter guts.

The solar power plant consist of 3 identical inverters.
The problem is that one of the three inverters (on phase 3) the power output goes up and down like crazy, oscillating from 4kW to 300W in a period of around 2 s.
Also the current goes up and down, as well as the cosfi (0.5 to 0.9, while should stay at 1).
The THDv (total harmonic distortion of the voltage) is also high with similar periodicity, which makes the filter of the other inverter to vibrate.
It seems also that the inverter overheats in sunny days (more than the other 2 inverters).
I can swap parts between the faulty and a working inverter and compare measurements.

I will contact SMA, which will probably tell me to send the whole inverter for repair.
However the fault seems something banal. The inverter overall is converting some energy from DC to AC, so I would like to attempt a repair.
I have some knowledge of converters (I am electrical engineer) and some basic instruments.
I don't have an isolated differential probe for the oscilloscope, yet.

[coromonadalix]

I should verify all the connections and screws posts, if they are tighten enough ...

I would not play in it even with experience and no schematics ... thoses things are very complicated and hazardous, you may do some damage to take good parts and swap them ..... and they works with high voltages up to 700v .......  thoses are not just converters they are powergrid converters tooo

[webgiorgio]

The inverter is made of 3 boards:
- DSP and LCD
- Inverter
- DC-DC and grid interface relays and filter.

I dismantled all the boards from the chassis and they seem in very good conditions.

The inverter is delivering power in a "uncontrolled" way above 200W.
So, I think the problem is in the control part or in the measurement of output voltage/current.
I think the DSP is messing up, I will swap the DSP board from a good inverter tomorrow if I get the Ok from SMA.

[boB]

You might tap around with a soft-ish insulated tool to see if it might just be an intermittent connection ?

Probably not the DSP itself if 2 of the 3 phases are working well.

[aargee]

Have you tried swapping two inverters around to see if the fault follows the inverter, or if it stays withe the array?

[webgiorgio]

I installed the DSP board of a good inverter on the faulty one, which remains with the exact same problems.
I see this as a good news! The most complicated part to fix is not the problem.

An important observation of today is that all the inverters works well at low power (200W, low Sun radiation).
What could be an inverter fault that is power related?

@aargee I think that the problem is not related to the PV string because if it was there, I would see the power dip only, while I also see the cosfi going low to 0.2. So I think that the inverter is doing a bad job.

Question:
I would like to be able to test the inverter "at the desk". How can I give it a DC supply of around 300 Vdc, 1-3 kW? (besides waiting for a sunny day )
Besides specialized test equipments, what about a rectifier bridge, electrolytic capacitors, and 3x500W halogen lamps in parallel?
As the inverter drains more current, the voltage drop on the lamp increases, reducing the input voltage to the inverter.
On top of this, as the filament heats up, the voltage drop increases even more. This gives me an inverse relation between voltage and current, almost linear, that resemble the I-V curve of a Photovoltaic generator.
The inverter should stabilise at a voltage where the product V*I is maximum.

I probably need to isolate the 230V with a 1:1 transformer, because the inverter would stop when detecting a ground fault on the DC side.

[cs.dk]

I would like to be able to test the inverter "at the desk". How can I give it a DC supply of around 300 Vdc, 1-3 kW? (besides waiting for a sunny day )

Waiting for a sunny day could take long.
I have a 0-600 VDC 2.6A supply if you are near Silkeborg. I don't know if that is power enough though?

[webgiorgio]

I spotted a component in D2PAK that is at 80°C instead of 38°C (on a good inverter).
http://www.ingdemurtas.it/wp-content/uploads/faulty-inverter-SMA.jpg

I think that that part of the circuit is in charge of measuring the string voltage, and measure the R_iso (isolation resistance of the string towards ground).

Besides the power, the harmonic distortion of current and voltage is crazy high!
http://www.ingdemurtas.it/wp-content/uploads/4_2Hz_4kW_20-11-2018_PV_inverter_fault.png

The plant is in Italy, I am there now. Thanks for the offer anyway, good to know

[webgiorgio]

How should I deal with the lacquer? It is difficult to measure with the multimeter with all the components coated!
I saw in a video that is applied spray. Is any solvent available?

How should I proceed?
Reverse engineer (get the schematic) of the part of circuitry with the 80°C component.
Or
Compare voltages with the good inverter nearby?

There are several pads that were used for automated testing during production. Shall I take advantages of those, and compare the signals with the working inverter?

The connector with the flat cable between boards carries signals, so, it seems a safe place to use the oscilloscope.
I need to find which pin carries the signal that comes from the hot component, and compare it with the well working inverter?

The opto coupler is SFH6156-3.
https://datasheet.octopart.com/SFH615A-3-Vishay-datasheet-10469670.pdf
The marking is on the diode side.

The hot D2PAK is Fairchild, 1A37A8 F08 N90, but I can't find a datasheet for it :/

[Zucca]

Is any solvent available?

Not that I know. Even if I would know one, I have the feeling it would be toxic.

Reverse engineer (get the schematic) of the part of circuitry with the 80°C component.
Or
Compare voltages with the good inverter nearby?

Both, why not?

There are several pads that were used for automated testing during production. Shall I take advantages of those, and compare the signals with the working inverter?

You are full of good ideas, just follow them and report here. 

[webgiorgio]

Thanks for the encouragement

I found a patent from SMA that describes the working principle and the circuit that measures the isolation resistance to ground. With a few versions of the schematic
https://patents.google.com/patent/US20150084654
There is a good correspondence with what I see on the board!

During measuring an insulation resistance for an inverter having at least one half-bridge including two active switching elements for driving an output current, and a DC link voltage, a center point of the half-bridge positioned between the switching elements is connected to a grounding point by closing a grounding switch, and the center point connected to the grounding point is connected, one after the other, to a first ungrounded terminal and a second ungrounded terminal of the DC link voltage of the inverter present at the half-bridge by means of the two active switching elements of the half-bridge to establish a connection between the first and second ungrounded terminals, respectively, and the grounding point. A current flowing via the connection to the grounding point is measured using a measuring device.

I managed to read out the part numbers, they are two 900V N-ch mosfets, two optocouplers, and I guess the "voltage measuring device" as they call it in the patent is simply a voltage divider (that series of 750 kohm 1%). I wonder how it is then isolated towards the DSP 

The 10W vertical resistor could be the "protective resistor" mentioned in the patent (13 in the schematic).

The connection to be produced between the
centerpoint of the half-bridgeand the grounding point for the
measurement ofthe insulation resistance in contrast does not
require a Switch that can be subjected to greater loading, in
particular when a protective resistor with a higher resistance
is provided in this connection

I think that I would just replace the two mosfets (one of which is at 80°C) and see what happens.

[EPTech]

Hi there,

I think you got something there. While the inverter is uncoupled on both the mains and the DC side, measure whether there is a short or low resistance between the GSD pins of suspicious MOSFET.

Happy hunting

[webgiorgio]

Thanks for the advice, I will be back to the solar plant around 18th December to take this measurement.
I also ordered an isolated differential probe for the oscilloscope (Micsig DP10013 TR).

[webgiorgio]

So, I got hold of the inverter again. This time I can take it from the solar field to the lab.
I measured (in circuit) the Drain-Source resistance, and it is like an open circuit (burden voltage of my multimerter is 0.2 V). That's unfortunate, I kinda hoped that that mosfet was shorted. Now I have to find out what is wrong in the circuit that drives it.

What if I pull the Gate leg up from the PCB, and tie it to the Source? To force it to be off.

I will also check weather it gets hot when the AC side is disconnected. I think that the two mosfets are in series and across the PV string voltage, like I saw in the schematic of the patent.

Could it be that the mosftet "fails" only when Vds is high enough?

[webgiorgio]

I could not draw the schematic of the part of the circuit involving the D2PAK mosfets, because I can't read the part number on some small packages.
However, it seems like I've found the guilty component!

On the bottom board there are two large TO-264 N-ch mosfets in parallel (see the photo in the first post).
300V 100A N-ch mosfet, APT30M30LLLG.
http://datasheetz.com/data/Discrete%20Semiconductor%20Products/MOSFETs%20-%20Single/APT30M30LLLG-datasheetz.html
One of the two mosfets is shorted.
I removed it and the inverter run. However, I could only test at low power output (50W).

I made a power supply to simulate the PV string, which basically consist of a 500W lamp in series with the primary of a 230-400Vac 100W transformer. The output is rectified to 550 Vdc (open circuit), and fed to the input of the inverter.

The mosfet in D2PAK that was 80°C is now at 35°C. More reasonable.
The Vgs is 2.5 Vdc and 4 Vdc for the other D2PAK. It seems too low! The gate threshold voltage for both mosfets is 2-4V according to datasheet. So, they are in the partial conduction zone.  It might explain why they are a bit hot.

So, I will replace the two TO-264 mosfets, and I will test the inverter at higher power level (a few kW) once I develop a power supply to simulate a 600 Vdc photovoltaic string. I will open a new discussion on the schematic I came up with and put the link here.

[fzabkar]

Could the D2PAK be an FQB8N90C?

FQB8N90C, Fairchild, N-channel QFET, 900V, 6.3A, 1.9 ohm RDSon, 171W, D2PAK:
http://www.mouser.com/ds/2/149/FQB8N90CTM-526708.pdf

[webgiorgio]

Yes it is! It was a 8, not a 5. It is not even getting as hot as before when I operate the inverter without the shorted mosfet APT30M30LLLG.

I am missing a lab DC source to test the inverter at kW power level.