Kai Deng GTI300W Grid tie inverter reverse engineering

[peter18]

Das ist komsich normalerweise berennen die Widerstände nicht einfach durch. Normalerweise gehen die Mosfets kaputt. Bei mir ca. jedes jahr 2 nebeneinanderliegende.

Deswegen werde ich dort auch jetzt eine bessere Kühlung selbiger einbrignen mit hilfe vom Heatpipe system.
nur einmal ist bis jetzt bei mir ne spule durchgebrannt weil ich ne Stromspitze ausversehen reingebracht habe.
Ist wirklich nichts anderes bei dir noch kaputt? irgendwas was einen ordentlichen Kurzschluss auslöst?

[marcash]

Hallo,

wie hast du die defekten Teile bei dir entdeckt?. lässt sich da das im eingebauten Zustand messen?

ich habe jetzt nur mal die stellen mit brand abgesucht. und da war nur die stelle.

es ist bestimmt ein kurzschluss irgendwo , weil als ich eine neue Sicherung reingemacht habe ist diese sofort wieder rausgeflogen.

[peter18]

gemessen. die mosfets halten wie kondensatoren immer länger je besser sie gekühlt werden.

Das normale kaputtgehen des china moppets war bis jetzt immer.
-Sicherung 6A Sicherungsautomat gekommen.
- Schmelzsicherung in gerät durch (2,5A F)
- und mosfet durch

Mosfet einfach im eingebauten zustand messen. Muss ein hochomiger wert bei rauskommen so größer 10 mega ohm.
ist er irgendwo bei 0-300 Ohm ist der mosfet durch.
dann auslöten und neuen rein.

Das ist bei meinen 600W wechselrichter meist immer anch ca. 1 Jahr der fall.

[vasonline]

Hi,

R65 and C37 are in the opamp circuit (U4), that together with VR1 controls the "insertion current limit"
C37 is probably the same value as C36 (they are not marked).
They then go to the UCC chip that does the work of driving the DC side.

Can you power the GTI with bench PSU, and see if the red light comes on?

Since I installed an external line filter, the AC side seems Ok however.
The inductive spike form switching off fluorescents would kill the AC side FET's w/o the filter.
Knock on wood!

[peter18]

Hi,

Thanks for the pictures!

-> It is indeed a part of the line filter of the GTI.

I have no idea of the value, but I think it will be fine with a similar coil from a discarded PC power supply line filter.
That should be comparable in power.

You could test it (bridge the connections) without the coil in place. It should work If there is nothing else broken.
I do recommend in the long term to replace it, since it has less resistance to noise/grid spikes/etc with half of the line filter missing.
If it still blows, check all 4 AC FETs for shorts.

ps: looks like a powerjack/sun brand GTI. Not KD, but they share the same idea's.

I solderd the new coil in but there is stille cirkut only DC Side plugt everything is fine. if I pluged the AC side (Bang) Fuse ist eliminated into 1000000000 pices. the 3 mosfets 13N50 are ok. What could be broken that there is such a big circut?

[vasonline]

Hi,

In the past I have made a series light bulb.
That prevents the fuse from blowing.

I find it handy to troubleshoot AC power supply problems.

http://www.giangrandi.ch/electronics/sm ... sfix.shtml

There is indeed still a short.

Greetings

[peter18]

Clever Trick, i will try ist by the light bulb

[vasonline]

Hi,

The following might of might not work for you:

Two ideas from me:

- It might be good idea to check all small SMD diodes across the AC FETs.
When a FET went <Bang> on my inverter, ocassionally it would take out one of the little SMD diodes.

- As a last resort, you could remove every AC FET one-by-one and see if the short goes away. (A desoldering station would be really handy! I bought myself one and did not regret it - for this and other projects )

[bazicpl]

vasonline: I found your great job and lots of efforts to reverse this inverter..

I have very similar one but 600W. It is sunyoung.

divider circuit

divider_of_gwv600w.jpg (213.86 KiB) 19153-mal betrachtet

sy-gwv600w inside circuit overview

gwv600w_inside.jpg (298.4 KiB) 19153-mal betrachtet

gwv600w_inside2.jpg (252.15 KiB) 19153-mal betrachtet

I tried to increase its DC max voltage from 60V to 75V. (I have PV panels with Voc=75V and Vmpp=59V)
Partially it was successfull.
I put 3,6V zener diode in series with 100kOhm resistor at the measuring divider to fool measuring alghoritm. Then when current is increased - voltage drops below 60V - around Vmpp.
All input circuits are capable to deal with greater voltage (mosfets are 100V) (to be honest - almost all: capacitors are for 80V only - they must be changed)

BUT unfortunately during one of my further tests I mechanically broke pull-down resistor of measuring divider so voltage much greater than 5V was passed to ADC pin of PIC microcontroller.
After this PIC is alive but "sees" only "under voltage" instead of real present on AIN0 pin
I measured AIN0 voltage before all of my tests: for 25V DC it was around 1V and now there is the same at AIN0
All seems I damaged an ADC.
I am almost sure main PIC is also copy protected so I even did not try to read it

And here is my question - is it possible to adapt your program for main version of PIC?

I would like to make it live again.

[vasonline]

Hi bazicpl

I think there are several posibilities, but they all require quite a bit of work and experimenting.

- You could desolder (hot air gun) the PIC and replace it with a new one.
- You could use one of other analog inputs (temp sensor If you could do without cooling fan), but that requires some rewiring on the PCB.
- It's quite a bit different from my KD unit, so you would need to verify if the pin-out is exactly the same.

(there is a drawing on page 2)

You could even ask for the manufacture for a new pre-programmed MCU. In my dealing with Chines manufacturers
I found that they are not always bad. I had good experiences in the past. That might even be the simpelest solution.


24/09/2016
Newest code update:

What's New:
--------------
- Fixed bug in optcoupler detection. Pullup resistors were not on. Same now as original FW. Something I noticed from the beginning. After re-reading
datasheet and comparing with original I found what needed to be done.
+ Better sine wave. More like original FW. If you want smaller step size: change VR1 from 5K to... 20K (Hardware mod)
+ Switching between sine wave now in software (got tired of updating two sine waves). Verified with scope: Ok.
+ Main ISR performance improvements. More If/else in main ISR. Streamlined the code in the ISR.

_Edited my old posts & removed the old code.