Hi, i have a max7. 2 with error code 09. There is a short circuit across the dc bus due to 2 legs in the ac bridge being blown to short circuit. Each leg has 2 igbts.  The  igbts from the other two bridge legs passes the static tests as per service manual. 

Is really required to replace all 8 of the bridge igbt's as per service manual, or is it ok to replace the faulty four only.? 

Also, what is the best way of removing the transistors. Desoldering all components  on the hosting heatsink, and lift out the entire subassembly ? Unscrewing the faulty transistors from the heatsink seems to be a challenging if not impossible task. 

Any experience to share? 

 

I think you know you need to talk to @Steve87. One of his favorite machines.

You don't want to try only replace the faulty ones, it's asking for trouble. I unscrew the IGBTs from the heat sink using a Z screwdriver and then desolder them. Removing the heatsink is a lot of effort. You might have to remove a capacitor or two to get good access but I normally manage without having to do so. Once the faulty componenst are off check the DC mosfets and all driver circuits again. To be honest if the inverter is more than a year old I don't recommend repair unless you replace the DC capacitors as well because I've had several repaired inverters fail again soon after repair. They passed full load tests so I can only think that capacitor aging is not buffering the surges well enough.

 

I think, this is a great topic & with reference a new Main board for this machine costs at least R7k & there abouts. So I think if you can overcome the challenges as stated above there is much to be gained. The guys in technical at Segen Solar for this reason just replace the main board with a new one. Mainly because it costs them far less for that board & when labour costs of a skilled technician has to do this it all adds up. 

Great going here @BritishRacingGreen 🍺🍺🍺

33 minutes ago, Shadders said:

You don't want to try only replace the faulty ones, it's asking for trouble. I unscrew the IGBTs from the heat sink using a Z screwdriver and then desolder them. Removing the heatsink is a lot of effort. You might have to remove a capacitor or two to get good access but I normally manage without having to do so. Once the faulty componenst are off check the DC mosfets and all driver circuits again. To be honest if the inverter is more than a year old I don't recommend repair unless you replace the DC capacitors as well because I've had several repaired inverters fail again soon after repair. They passed full load tests so I can only think that capacitor aging is not buffering the surges well enough.

 

Thank you for valuable info. 

1 hour ago, Shadders said:

You don't want to try only replace the faulty ones, it's asking for trouble. I unscrew the IGBTs from the heat sink using a Z screwdriver and then desolder them. Removing the heatsink is a lot of effort. You might have to remove a capacitor or two to get good access but I normally manage without having to do so. Once the faulty componenst are off check the DC mosfets and all driver circuits again. To be honest if the inverter is more than a year old I don't recommend repair unless you replace the DC capacitors as well because I've had several repaired inverters fail again soon after repair. They passed full load tests so I can only think that capacitor aging is not buffering the surges well enough.

 

Given thorough thought to your comments, yes i agree. This whole power circuit is under 400vdc switching dc stress and the handling of 7 killowatts is demanding. 

In my case the machines that is faulty are uncommitted/unallocated so its worth a try to bring them up to live again. 

But, yes, you cannot really carry a warranty with these kind of repairs. Its too indeterministic. 

Of course! , desolder some ac caps next to the one array will provide access. Thanks.! @Shadders do you repair units? And if so where do you live? 

To put into context this entire repair & how the damaged parts looked. These inverters were less than 8months old & were water damaged by the recent floods in Durban. @BritishRacingGreenwas very kind to take up the task of trying to repair & at the same time gain some knowledge on these OG7.2s. The context is important because from the pics you will realise just how robust this machine really is. The warranty is voided for water damaged goods. 

Out of 3 damaged machines 1 is already up & running & I will carry out the factory standard stress tests & observe it's operation over time. 

 

I think most of not most ppl looking at these pics would say that these machines are ready for the rubbish bin & there is no life left in them. 

I will leave the technical intricacies up to @BritishRacingGreento explain. All I can say is well done!!! It's incredible to know that in this community we have such a person with such expertise. 

Here are pics of the working unit after the tare down, clean up & repair. 

incredible 😂😂😂😂

9 hours ago, Steve87 said:

I think most of not most ppl looking at these pics would say that these machines are ready for the rubbish bin & there is no life left in them. 

I will leave the technical intricacies up to @BritishRacingGreento explain. All I can say is well done!!! It's incredible to know that in this community we have such a person with such expertise. 

Here are pics of the working unit after the tare down, clean up & repair. 

incredible 😂😂😂😂

Thanks @Steve87, i must say having the luxury of 3 machines makes for a much more comfortable introduction to faultfinding by comparison, reference and redundance.

Max7.2 is intimidating but i must say i am enjoying this journey. 

One thing to point out is that all pcb's are conformal coated and that helps the defence against water ingress and mud. To what extent the capacitors had moist ingress one cannot tell, but they seem ok. So this gave me the luxury to wash the boards with diluted baking soda, and then bake them dry. 

Unfortunately the power stages of the axperts are all integrated onto one motherboard which is nightmare. Would have been so nice had the drive transistors /heatsink  clusters been modularly mounted seperately,along with their drivers. 

As coulomb has stated in his notes, the main board makes up 80% of the machine. 

Edited September 30, 20223 yr by BritishRacingGreen

@Coulomb hi,  i have noted from you and others that repairing of power stages involving blown mosfets/igbt can be very unforgiving, since the driver can still cause harm when transistors are replaced. The walkthru in the service manual will help to check the drivers but thats very static surface test. 

So one thing comes to mind, take the ac output bridge as an example. Would it not be viable and deterministic to introduce say a 2A fast blow test fuse between the dc bus feed and the bridge? Yes, it will require some surgery to the main board, but i was just wondering. 

These igbt also not cheap here in SA considering you need to replace 8 devices in batches.  And removing them again will break my heart💔

Edited September 30, 20223 yr by BritishRacingGreen

22 minutes ago, BritishRacingGreen said:

this gave me the luxury to wash the boards with diluted baking soda, and then bake them dry

Actually i did not wash them, i scrubbed them in order to minimise change of smaller  transformer moist ingress 

Another assumption i am making is that the canbus, rs232 and rs485 interfaces are galvanically isolated from the machines internal supplies.    Because we  expose them to our devices which is invariably not. 

 

 

21 hours ago, BritishRacingGreen said:

do you repair units? And if so where do you live? 

 @BritishRacingGreen yes I do repair units but methinks Harare, Zimbabwe might be a bridge too far for you 😁

I have found that water ingress isn't too bad unless it has burnt tracks. I find the biggest issue for me is the DC caps, they are not easy to come by and my parts supplier can't source them. I've had to pull from several old machines to get good enough ones.

If the machines are less than 8 months old and weren't running at high power conditions you will most likely get away with the repair not having to replace the caps. Message me if you need any help and I'll try my best.

7 hours ago, BritishRacingGreen said:

Would it not be viable and deterministic to introduce say a 2A fast blow test fuse between the dc bus feed and the bridge? Yes, it will require some surgery to the main board, but i was just wondering.

No, even the fastest fuse isn't fast enough to save IGBTs. Even so-called "semiconductor" fuses.

The best thing is to test the drivers with no bus voltage at all, using bench power supplies to supply current limited +12 V and -12 V and about 20 V for the +15 and -5 V isolated power supplies. I short the input from the control board (it's open collector, so you can't hurt them by shorting them) and watch the gate to emitter voltage change from around -5 V to around +13 V. It's a pain with all the connections you need, but really good insurance. As you say, a set of IGBTs isn't cheap, and blowing up a set because you didn't test hard enough is demoralising, as well as expensive.

7 hours ago, BritishRacingGreen said:

Another assumption i am making is that the canbus, rs232 and rs485 interfaces are galvanically isolated from the machines internal supplies. 

Yes, they are.

7 hours ago, Shadders said:

I find the biggest issue for me is the DC caps,

Do you mean the DC bus capacitors, or the ones across the 48 V battery protecting the MOSFETs?

@Shaddersand @Coulombthank you for valuable inputs.

@Coulomb I would have then embarrassed myself with the fast fuse !  I was thinking in the line of a test jig to verify low voltage driver circuits, thanks for the suggestion,  I am going to build one.  Further to this , I have used your IGBT Driver PSU  and Driver Schematics as preemption to trace the MAX7.2 . And it helped a lot , thank you . I  have the following to note :

1. The MAX in general does not differ from the 5KW models , its based on Voltronic's brilliant patent after all. 

2. There are 3 PSU's , one for AC High Side , one for AC Low side, and another which got to be for the buck (haven't traced the buck at all)

3. The PSU's only differs by the designer using negative logic , meaning as per your 5KW schematic, D19 , ZD7 , ZD8, C24  are reversed in polarity , and of course appropriately routed correctly to the driver ic in terms of polarities. Also for one PSU the 18V zener is larger than all the others . What gets me is there is not a current limiting resistor in the diode and zener path . (transformer impedance helping here ?)

4. There are 5 drivers ,  one each for bridge leg , and the 5th got to be for he buck.

4 The driver is not 8 pin ACPL-T350 . Instead its a six pin device , but I figured its configuration out without even identifying the chip. Your D8 is a three pin device on my board, baffled me , but eventually saw its a dual diode of which only one is used!

5. The IGBT's are 80A rated and there are two of them in parallel for each bridge leg. Is there only one in the 5KW  machine , or also dual?

Here's what I picked up so far during the tracing exercise :  My IGBT's are all sound by virtue of both inspection and static resistance and diode tests. But the one driver PSU has gone south . The 5v6 Zener blown dead short, and the 18V Zener , although still showing 600mV forward voltage drop , does have damage from the outside. Wonder what happened , the IGBT passes static tests , all 8 of them . Fortunately it did not cause a trail of destruction !   Could be the power supply in the upstream direction , because the whole machine is as dead as a doornail.

Next week I am going to replace these Zeners and build the jig to check it out . Learning while repairing !

When I have a half decent partial schematic for MAX/PIP7248 I will publish it for you . Once again thank you for the wonderful research and reverse engineering you have before and probably still are.!!!!

 

Kind Regards

 

 

  

7 minutes ago, BritishRacingGreen said:

1. The MAX in general does not differ from the 5KW models , its based on Voltronic's brilliant patent after all. 

Huh. I should go look it up.

7 minutes ago, BritishRacingGreen said:

2. There are 3 PSU's , one for AC High Side , one for AC Low side, and another which got to be for the buck (haven't traced the buck at all)

No, the buck shares a power supply with the AC low sides. Each inverter proper (DC-AC converter) high side needs its own power supply (all off TX7 in the 5 kVA schematic).

Quote

Also for one PSU the 18V zener is larger than all the others . What gets me is there is not a current limiting resistor in the diode and zener path . (transformer impedance helping here ?)

I found that weird too. But I think it's actually that the 5 V zener is expected to conduct, but the 18 V zener only conducts if the voltage becomes excessive. Or maybe  the flyback design is somehow inherently current limiting; I'm not familiar with them. Voltronic seem to use them everywhere though, so maybe I should study them a little more.

7 minutes ago, BritishRacingGreen said:

4. There are 5 drivers ,  one each for bridge leg , and the 5th got to be for he buck.

Yes.

7 minutes ago, BritishRacingGreen said:

Your D8 is a three pin device on my board, baffled me , but eventually saw its a dual diode of which only one is used!

That's actually quite common. It's not like an SMD dual diode costs much more than a single diode, and it helps reduce the component count. One less reel to worry about.

7 minutes ago, BritishRacingGreen said:

5. The IGBT's are 80A rated and there are two of them in parallel for each bridge leg. Is there only one in the 5KW  machine , or also dual?

No paralleling of IGBTs in the 5 kVA models. I'm surprised that they directly parallel them; usually it's considered bad to parallel IGBTs. MOSFETs do the right thing for current sharing when they get hot; IGBTs do the wrong thing.

7 minutes ago, BritishRacingGreen said:

The 5v6 Zener blown dead short, and the 18V Zener , although still showing 600mV forward voltage drop , does have damage from the outside. Wonder what happened , the IGBT passes static tests , all 8 of them .

Are you saying that you didn't have to replace any of the 8 IGBTs in the inverter proper?

If you did have to replace at least one, then that's the failure mechanism: the IGBT melts, shorting collector to gate, and that puts huge stresses on the gate driver components.

Quote

Fortunately it did not cause a trail of destruction !   Could be the power supply in the upstream direction , because the whole machine is as dead as a doornail.

Yes, if the flyback circuit failed, it could have sent forward a pulse of energy through the "transformer".

Quote

When I have a half decent partial schematic for MAX/PIP7248 I will publish it for you .

Excellent, thanks.

Quote

Once again thank you for the wonderful research and reverse engineering you have before...

I can't take much credit for the 5 kVA schematic; that was 99% Maxo's work. I did tinker with it here and there, suggested modifications, and eventually took over maintenance of it, and generated a PDF, which I find much easier to navigate.

28 minutes ago, Coulomb said:

Are you saying that you didn't have to replace any of the 8 IGBTs in the inverter proper?

If you did have to replace at least one, then that's the failure mechanism: the IGBT melts, shorting collector to gate, and that puts huge stresses on the gate driver components.

Its a bit confusing if one follows this thread , I really have two main boards each with its own fault . The one has 4 IGBT's blown , the drivers looks ok. The other board , the one I  am referring to in your post , have the IGBT's ok , but the the driver PSU faulty (zeners) .

39 minutes ago, Coulomb said:

I'm surprised that they directly parallel them; usually it's considered bad to parallel IGBTs. MOSFETs do the right thing for current sharing when they get hot; IGBTs do the wrong thing.

Yes , and I can confirm that they are hard paralleled like mosfets  , ie.  E-E and C-C , no resistor in each emitter legs as per many bipolar paralleling arrangements. The two gates  also fed from one driver , which I suppose makes sense.

 

 

45 minutes ago, Coulomb said:

No, the buck shares a power supply with the AC low sides. Each inverter proper (DC-AC converter) high side needs its own power supply (all off TX7 in the 5 kVA schematic).

Have to believe you here , probably made a stuff-up , will trace that again.  It may explain also why the transistor enumeration   is like  QA1/QA2/QC1/QC2/
QB2/QB3/QD2/QD3 .

 

50 minutes ago, Coulomb said:

Huh. I should go look it up.

I am mainly referring of the design philosophy , eg,   

@Coulomb here is a shot in the dark from my analogue guru friend at work . He reckons at that amps you can introduce pcb track with solder padding to achieve required resistance for bipolar . I then checked the motherboard and wondering if  these awkward tracks are telling somethin. that is track between  E-E .  Could be totally wrong. Could just be to increase current carrying capability on main pcb. 

 

Edited September 30, 20223 yr by BritishRacingGreen
Extra info

1 hour ago, BritishRacingGreen said:

@Coulomb here is a shot in the dark from my analogue guru friend at work . He reckons at that amps you can introduce pcb track with solder padding to achieve required resistance for bipolar . I then checked the motherboard and wondering if  these awkward tracks are telling somethin. that is track between  E-E .  Could be totally wrong. Could just be to increase current carrying capability on main pcb. 

 

Here is also an interesting characteristic of IGBT. At roughly 70% current and below, the temperature coeficient is negative, and above that it is positive. 

Edited September 30, 20223 yr by BritishRacingGreen

22 hours ago, BritishRacingGreen said:

I am mainly referring of the design philosophy , eg,   

Ah, yes, I remember that now. Long ago I also posted an "as actually implemented" diagram:

 

Edited October 1, 20223 yr by Coulomb

21 hours ago, BritishRacingGreen said:

He reckons at that amps you can introduce pcb track with solder padding to achieve required resistance for bipolar .

Maybe. But they seem to use that solder padding everywhere that there is large current.

I'm staggered that they can get 7.2 and 8 kW on a PCB. There is also an 11 kW model recently. They don't even seem to bolster the current carrying capability with stiff single core copper wire soldered to the PCB tracks, as Elcon/TC chargers do (or did; I don't know what their recent models are like).

Edit: though I guess the really high currents are on the battery side, and for that they supplement the PCB tracks with copper bars:

Edited October 1, 20223 yr by Coulomb