PIP inverter repairs and hardware modifications

[Howard]

Hi All,
I'm new to the forum.Great to be able to contribute. Paukvk and I purchased eight PIP-4048WM VERSIONS(2015). One(mine) has been on line for about 2 years just out of warranty.Paulvk for about same time.
HAs been running well until last week or so when loaded up to 82A through batt/solar mode.Noticed smoke burning white smoke -blowing form fan exits-shut down and took off line immediately in view of stress,blowing Mosphets.Although no failure was noted but we suspected caps.(Underrated 3300uF 65v electros)

After calling GPower without much help was totally frustrated why main board could not be removed and was going to ask you?

Quite amazingly as I was working on the same problem at the same time on the same day(Yesterday) as you!!
Thanks soo much for the post because that is ONE sneaky screw under date code designation label!!!

Have done visual -no dc checks of Mosphets but noticed one of 3300uF 65V electros closest to rear of inverter light brown colour slight leakage component side between pcb and cap crimp rim diam.
So will research new caps,check mosphets and advise. Will probanly put this one back on line and replace cap on other PIP
Thanks again!!

[weber]

Welcome Howard. Yes., that Coulomb is one coul doud.

You're lucky you only got the sweet white smoke of Gandalf the capacitor and not the foul-smelling black smoke of Sauron the MOSFET.

Coulomb and I recently did an exhaustive search of Digikey, Mouser, RS Components and Element14, looking for suitable upgraded capacitors with an 80 V or more rating. This time we allowed for capacitors up to 22 mm in diameter, as previously one could only fit 18 mm. But we did not find any new option with low enough ESR or impedance, and high enough ripple current. So our recommendation for the capacitors is exactly the same as it was 2 years ago.

Here's the data sheet for the existing capacitors. I had to request this from the capacitor manufacturer. It seems it is a "special", i.e. not part of their standard line.
Jamicon WL-series 3300 uF 63 V 105°C 2000 hour

Here's our recommended upgrade.
United Chemi-con EKZN800ELL 182MM40S, 1800uF 80V 105°C 10,000 hour
Digikey part 565-4129-ND. Also available from other suppliers.

Here's my post from 2 years ago, in which you can read more details about this choice, and the slight hit taken on impedance and ripple current in return for voltage and endurance.
viewtopic.php?title=pip4048ms-inverter& ... 332#p56103

The existing MOSFETs are the 75 V IRFB3077.

A few weeks ago I listed 5 potential choices for upgrading these MOSFETs to 100 V.
viewtopic.php?title=pip4048ms-inverter& ... 332#p64007
We've since settled on the SUP70040E.
Digikey part SUP70040E-GE3-ND. May be available from other suppliers.

[Edit: We have since used the CSD19535KCS]

As with the capacitors, we've taken a slight hit on current ratings and on-resistance in return for the voltage rating, but the intention was that no spec should be significantly worse than that of the 75 V IRFB3307 used by the manufacturer two years ago.

I note that each inverter uses 4 of the capacitors and 16 of the MOSFETs. You may choose to replace only the capacitors, but you should definitely replace all four at once, even if only one has failed. The others won't be far behind. And since it is such a pain to get the board out, you might consider upgrading the MOSFETs at the same time.

Same goes if only half the MOSFET have blown. You should replace all 16 of them, and the 4 caps. [ Edit: but see this post about only replacing half of them under certain circumstances. ]

Yesterday at my place, Coulomb pre-emptively upgraded the MOSFETs and capacitors in his PIP, to the above choices. I have just heard that he has reinstalled the PIP and it is working fine.

But of course we cannot guarantee this upgrade. You do it at your own risk.

[ Edited Coulomb: removed then replaced degree symbols in preparation for phpBB conversion ]

[Howard]

Hi Weber ,
Thanks for that and first message ...Yep will do .Plan to do all that.Concur on mosphets and overall view.
We will need about 32 capacitors for all units.We were thinking about Nichicon for caps but have to check specs & bulk order?
You guys using earth straps on wrists?
Thanks again..no worries yes realize the risks.

[weber]

Hi Howard. We didn't bother with anti-static wrist straps because the humidity is so high here in Brisbane at the moment. Let's face it, it almost always is, except for a few weeks in winter. But they are a good idea.

We looked at Nichicon capacitors. We did not find any 80 V (or higher) Nichicons in 22 mm (or less) diameter with radial leads, with a low enough high-frequency-impedance and a high enough ripple current, along with a 105°C endurance rating of 2000 hours or more. If you think you might have found one, please point us at it.

I note that these are the important parameters in this application. The capacitance is utterly irrelevant.

[Howard]

Hi Weber,Thanks again for that.. Will advise on Nichicon caps either way-
Will get to parallel mode yet.

[Howard]

Quick answer on Nichicon is no stock ..but can make volumes.

We are investigating these capacitors that this was not an O/volt issue but an ESR,temperature,(yes aging too) as they are mounted right next to heatsink.
It was as load ramped up to 82A from batteries inverting to 240VAC 4KW load that cap/s died.Have logged .pdf report if anyone is interested.

Will keep you all posted. Thanks!

[Truka]

Hello everyone .
After so much searching I did not find people as wise as here.
I come to ask for your help to repair my PIP.
When I changed a fan (with the problem already mentioned here) to another with the colors of the different wires, I shorted the wires and damaged the board.
I was very grateful to know the reference of the damaged components:
-D40 + C38 + ZD7 + R158 + ZD8
Thanks and greetings to all.
     Truka.

[coulomb]

Hello everyone.

Welcome, Truka.

I come to ask for your help to repair my PIP.

Sorry to hear you need to do that.

When I changed a fan (with the problem already mentioned here) to another with the colors of the different wires, I shorted the wires and damaged the board.

Interesting. But the parts you mention below seem to be part of the IGBT power supply; I would not think of these as related.

I was very grateful to know the reference of the damaged components:
-D40 + C38 + ZD7 + R158 + ZD8

I was curious as to what that part of the circuit was doing. So I did a quick trace of it; see attached.

Warning: I've probably left parts out of this partial schematic.
Use your browser's controls (View Image, Open Image in New Tab, etc) to see the schematic in detail.

The Zener voltages and part numbers are a bit of a guess, based on my reading of tiny marking codes, and other markings on the parts that could be the real marking codes. I think it's making a negative supply of about 5.6 V, and a positive supply of up to 18 V, relative to the IGBT emitter, and isolated from the source of power (I can't find where those wires go; it has low resistance so presumably it's another transformer winding).

This circuit is repeated two more times. One of these power supplies runs three opto-coupled gate drivers; the other runs one. QB2 is one of the upper IGBTs of the 230 V inverter output full bridge.

The RS1J diodes are moderately fast 600 V rectifiers, with a tᵣᵣ of 250 ns, made by Vishay. Other similarly fast rectifiers, with any decent voltage rating should do. These are SMA (DO-214AC) sized devices.

[ Edit: Added warning; source -> emitter; added DO-214AC package size. ]
[ Edit: C11 -> C38. ]
[ Edit: Redrew so the power flow is left to right; indicate multi-winding inductor rather than transformer. ]
[ Edit: The schematic is now an attachment. Added schottky diodes. Replaced with up-to-date version with left-to-right flow. ]
[ Edit: Removed extraneous schottky diode; show lower transistor of U3 is MOSFET with diode; added SMD codes and part numbers ]

[Truka]

Coulomb
Very grateful for the precious help.
I also do not understand if the fan connectors sit low on the board, and the burned components are on the top.
I can not find the capacitor C38. Thanks for your help.
Greetings to all.
    Truka.

(I do not understand how to send photos.)

[coulomb]

I also do not understand if the fan connectors sit low on the board, and the burned components are on the top.

I think this could be a little mangled in the translation from Portuguese. I think you are remarking that the fan connectors are at one end of the main PCB, and these components are near the far end, so what could the connection be?

I'm thinking that somehow the voltage to the transformer primary (at the right end of my partial trace schematic diagram) was too high. Maybe it was because the fans were shorted out, and it was trying to raise the voltage to the fans, and this raised the voltage to other parts of the circuit as well. Or maybe it was some sort of transient.

Hopefully, the two zener diodes in series did their job, sacrificing themselves to protect other parts of the circuit. However, I'd check the other zener diodes and associated parts: ZD1,2, and ZD3,4. They are nearby; you can see them in the photo below.

I can not find the capacitor C38.

I don't understand how you know that C38 has failed, if you can't find it. Did you get these parts designators from some repair manual or similar?

[ Edit: perhaps you mean you don't know how to find a suitable replacement. It is a bypass capacitor, so its value will not be critical. Any multilayer ceramic part of the right size (M1608? some suppliers still use the old imperial size 0603). Something like this (Mouser Portugal; I hope I got that right). ]

C38 is highlighted below; this is from an older (2014?) model:



The T350 opto coupled gate driver is one of five in a group, but I assume you already know where these parts are, to give their designators.

C24 is on the other side, with the 15k bleed resistor R158 across its leads. I'm surprised to hear that R158 is burned; its value is 15k; assuming it can dissipate 400 mW (total guess), it would need more than about 77 V to overload it. At that voltage, plenty of other components would burn up. Hopefully XD7 and ZD8 failed shorted, saving other components, but that would also include R158.

(I do not understand how to send photos.)

Years ago I wrote a post about how to attach photos to posts; I can't find it now. Maybe it's time to write another one.

[ Edit: I did, and you can find it here. ]

[Truka]

Coulomb, you're the best.
I mean, I did not see the C38 capacitor in your circuit.
Yes, the board is 2014.
thank you

   

[coulomb]

I mean, I did not see the C38 capacitor in your circuit.

Oh! My bad. I copied a capacitor from another schematic, and it happened to be C11. I then forgot to change its designator to C38. It's the capacitor across ZD7.

Sorry about that, corrected now.

[coulomb]

Weber and I are preparing to do a MOSFET and capacitor upgrade to another PIP-4048MS. This one was manufactured in February 2017, and has different MOSFETs and capacitors to the ones we usually see. Wow. Could they have seen the light and upgraded to our recommended specifications, making such upgrades unnecessary from now on?

The MOSFETs are Texas Instruments CSD19505KCS. This is nearly the same part number as one of our recommended MOSFETs, the CSD19535KCS. The main difference is the MOSFET rated voltage; they are using 80 V compared to our recommended 100 V. With the lower voltage, they get better on resistance (2.6 mΩ versus 3.1 mΩ), but earlier models used the IRFB3307, which are 5.0 mΩ). So they've gone from 75 V to 80 V MOSFETs. This is probably adequate if the layout in the battery side of the inverter is exceptionally low inductance, but I don't believe that this is the case.

There are actually two different brands of 48 V bus capacitor installed in this particular inverter. One pair is Jamicon TK series 3300 μF 80 V capacitors. The 80 V is another improvement, in line with our recommendations. However, the TK series is "general purpose", though with 105°C rating. The rated life is only 2000 hours, compared to 10 000 hours for our recommended part. It also has a low ripple rating, around 2 A [ edit: at 120 Hz; see below ] (it's hard to find real data on this part; it may be a custom manufactured part). The original capacitors were rated at 4.57 A, so they've gone backwards in ripple current rating. It seems to me that in this application, ripple current rating is important, so again, their upgrades don't seem to have gone far enough.

The other pair of capacitors is Jainghai CD294 series 2200 μF 80 V. Again, the voltage is good, but the life (2000 h) and ripple current rating (2.1 A [ edit: at 120 Hz, see below ]) are poor. The CD294 doesn't seem to be a high ripple current series. The capacitance value (μF) is irrelevant in this application.

So the capacitors have improved in voltage rating, but worsened in ripple current rating. I don't know if this is overall an improvement or not.

[ EDIT: My bad! I've again been caught by the various manufacturers quoting ripple current rating at different frequencies, e.g. 120 Hz versus 100 kHz. Our recommended capacitor has a 120 Hz ripple current rating of 2.3 A, not much better than the Jianghai at 2.1 A, or the Jamicon at some 2.3 A. Without specifications at 100 kHz, it's hard to tell which is better. The mere fact that our recommended capacitor has a 100 kHz ripple rating, and that it's a member of a low ESR series, is indicative (but by no means proof of) better high ripple performance. ]

In summary, the improvements are basically welcome, but in our opinion don't go far enough. Still, it will be interesting to see if and how much the reliability improves with the 80 V MOSFETs and capacitors.

[ Edit: qualified "inverter" to "battery side of the inverter" ]
[ Edit2: Removed non-ascii characters in preparation for a forum conversion to phpBB ]
[ Edit3: Re-inserted UTF8 characters in preparation for testing another forum conversion to phpBB ]

[paulvk]

I have a ESR meter on the way.
I will measure the various caps and see what the results are.
Note it can measure at 100kHz
Looking at two cases of smoking caps high current is involved.
The vacuum cleaner with commuter in a shorting position (almost a short).
Howard with probably nearly 90amps on 48v side.
In my opinion the ripple current is the problem, draw a lot of current from the batteries the ripple becomes too much for the caps heating them.
Once the caps are gone high ripple & voltage spikes, because of this the MosFets fail.

I can only recommend that the resistance to the batteries be kept very low.
I have 500mm of 35mm2 cable to my positive 1 meter long 100mm2 buss bar which the batteries positives are connected (I have a positive earth system) and 1 meter of 35mm2 to my negative shunt on the other side of that is 11 25mm2 cables from the battery banks via circuit breakers.
There is more than 30kg of copper in the cables!

[coulomb]

I have a ESR meter on the way.
I will measure the various caps and see what the results are.

Excellent. I hope you saved some failed capacitors to check your theory. I'm pretty sure it will be validated.

[weber]

I am aware of at least 3 different failure modes involving battery-side caps and/or MOSFETs. They might be called slow, medium and fast failures.

The slow mode has no obvious venting of the caps, but has the MOSFETs fail after 3 months or more of operation. This seems likely to be due to the caps having high ESR due to reaching the end of their 2000 hour life due to running constantly near their maximum temperature (due to ripple current) and near their maximum voltage.

The medium-speed mode involves venting of caps. If this is noticed early, the MOSFETs may not fail. It seems obvious to me that it is caused by heating of the caps to well beyond their rated temperature, by ripple current. Is there any alternative hypothesis as to what would cause the caps to vent?

The fast mode is the one observed with the vacuum cleaner, which was all over in less than a second. The MOSFETs failed as soon as the vacuum cleaner was turned on. The motor didn't even spin. The same vacuum cleaner had been used many times previously, and has been used many times since. The inverter had been in service for less than 3 months and had been only lightly loaded for most of that time, and was lightly loaded at the time of the failure.

However it had been a very hot summer and the inverter was not in air conditioning. So the capacitors may have aged significantly and so increased their ESR, and so although not at end-of-life they may have been less capable of controlling an unusually high voltage spike when it arrived. It is only in this case that it would be interesting to measure the ESR of the caps from the failed unit and compare it with new ones.

[weber]

In the vacuum cleaner case, the cabling between the battery and the inverter consists of 2 metres of 100 mm² (2 x 50 mm² in parallel) and 800 mm of 50 mm².

[paulvk]

I now have the ESR meter and Howards 4 caps
They measure close to 2700uF on my capacitance meter and between 0.026 to 0.024 ohms on the ESR meter.
Now the question is are they really 3300uF very strange they all measure very close to 2700uF , my meter is home made but is accurate.
Will be interesting to see what I get for the ones in my inverter but this may take a couple of weeks.

[coulomb]

Now the question is are they really 3300uF very strange they all measure very close to 2700uF , my meter is home made but is accurate.

I think it may be because 3300 - 20%(3300) = 2640 μF, so 2700 is just over the -20% lower tolerance limit for capacitors this size. I think that perhaps they can control the capacitance way better than when the standard tolerance was set at ±20%, so they can save material by aiming near the lower end of the tolerance range. Evil, but legal.

I've sometimes noticed that capacitors read low, and often near the low end of their tolerance range.

Of course, it could be that they all suffered a reduction in capacitance from new, along with an increase in ESR that contributed to the failure. I note that the 2013 originals and replacements are supposed to be 15 and 17 mΩ respectively, and you measured 24-26 mΩ. I'm guessing that the increase in ESR (25/16 = 1.56) could be a problem.

[coulomb]

Last project day Weber and I started repairing his PIP-4048 which had blown its battery side MOSFETs. Experience has shown that when this happens, other parts driving the gates tend to fail as well. It helps to have a schematic for these, so we traced out the following back to U9. [Edit by weber: C152 is 2.7 nF]

Battery MOSFET driver.png (63.88 KiB) Viewed 1418 times

To see in full detail, you may need to use your browser's controls, such as View Image or Open Image In New Tab.

Many parts of this schematic are repeated, as indicated by the coloured rectangles. For example, transformer TX5 has two seconary windings; the other winding connects to another gate driver for the MOSFET at the opposite corner of the full bridge. The B output of the PWM chip connects to another set of transistors, which drive transformers TX8 and TX11. TX8 has two secondaries that connect to two gate drivers for the other corners of the full bridge on the battery side.

The low side battery MOSFETs are on the outside; the high side battery MOSFETs are in the middle. The primary of the high frequency transformer is the output of the MOSFET full bridge. The drains of the low side MOSFETs, the sources of the high side MOSFETs, and the inputs of the high frequency transformer are connected by a flat bar between the main PCB and the heatsink for the MOSFETs. Tabs from these two flat bars bend down at 90° to the bar, and poke through the PCB like thick square component leads [ edit: see this post for what they really look like ] :



[ Edit: I had somehow completely failed to acknowledge Weber's contribution to this. Some parts, like the existence and purpose of the flat bars, were completely his discovery. My apologies. ]
[ Edit: Transformers TX10 and TX11 drive the gates of the MOSFETs of the full brige at the 400 V end of the high frequency transformer, not the 230 V inverter IGBTs. Thanks to Weber for pointing that out. ]
[ Edit: Added more text attempting to explain what is replicated. ]
[ Edit: Added D55, renamed OP1 to U19, thanks to Weber's checking. ]
[ Edit: Added D26 and D15 to the schematic. ]
[ Edit: Added rectangles showing what gets repeated; several diodes now Schottky type; Q23 -> Q26 and other minor updates ]
[ Edit: Added R14, C5, 3 capacitor values, pin 1 connected to pin 9 ]
[ Edit: Eliminated evil four-way joins. ]
[ Edit: Removed sentence starting with "I have not traced out the secondaries of transformers TX10 and TX11". That part of the circuit is traced now. ]
[ Edit May 2018: Updated to re-include R14 etc. ]
[ Edit Sep 2019: Value for C152. ]

[coulomb]

While I'm in a partial schematic posting mood, and also Maxo on the South African Power Forum is having trouble with this power supply, I thought I'd post this partial schematic trace that it seems I hadn't posted before (belatedly for Maxo, unfortunately).



As usual, use your browser's controls (View Image, Show Image in New Tab, etc) to see the schematic in detail.

Edit: It seems that this power supply is simply not present in later models; this trace is from a 2013 model. This may explain why earlier models could run the LCD display and update firmware with utility power but no battery.

Edit 2018/August: To see if you have this power supply, the easiest way is to look for the presence of transformer TX6 on top of the main board:

TX6 location.jpg (258.47 KiB) Viewed 3986 times

Edit: Redrew schematic to make it clearer.
Edit: Added Q63 and associated parts.
Edit: Various corrections, e.g. ZD15 is 12 V.

[Howard]

Hi Coulomb,
Great to have schematics!
I have just done the cap mods.Replaced MFR PART#EKZN800ELL182MM40S
1800UF 80VDC 105C UCC 10,000 HR ELECT.CAPACITOR the decided best option on the PIP4048.
Had some old blank pcbs I could use.Paralleled caps to get value .They all came in when measured above 3300uF.I mounted caps upside down on a PCB and screwed to some unused heatsink holes and wired them to mother board or main board.
Question?
Do we know how thick wire should be Cap.pcb to mother board.Any idea of series DC current the 3300uF'sor 1800// sink?( ..not to sure where caps. are in the Battery charge circuit ?
Ive only used approx.6 strand hook-up wire? Distance approx 90mmx2(+/-)
This is thicker than the original capacitor tinned copper leads
Do you/we think this is sufficient?
Using caution thanks -ready to restart PIP.

[weber]

Hi Coulomb,
Great to have schematics!
I have just done the cap mods.Replaced MFR PART#EKZN800ELL182MM40S
1800UF 80VDC 105C UCC 10,000 HR ELECT.CAPACITOR the decided best option on the PIP4048.
Had some old blank pcbs I could use.Paralleled caps to get value .They all came in when measured above 3300uF.I mounted caps upside down on a PCB and screwed to some unused heatsink holes and wired them to mother board or main board.
Question?
Do we know how thick wire should be Cap.pcb to mother board.Any idea of series DC current the 3300uF'sor 1800// sink?( ..not to sure where caps. are in the Battery charge circuit ?
Ive only used approx.6 strand hook-up wire? Distance approx 90mmx2(+/-)
This is thicker than the original capacitor tinned copper leads
Do you/we think this is sufficient?
Using caution thanks -ready to restart PIP.

Do not restart your PIP.

You can't position the capacitors any further away from the MOSFETs than the original ones were. It doesn't matter how thick you make the wires, they will still have too much series inductance, which will stop the capacitors from doing their job, by making the total impedance too high at high frequencies (i.e. during fast switching transitions).

And you do not need to have the same capacitance as the original. So you do not need to parallel two of the 1800 uF to replace one 3300 uF.

As I mentioned here: The impedance and ripple current ratings are what matter in this application. The actual capacitance is irrelevant.
http://forums.aeva.asn.au/viewtopic.php?title=pip4048ms-inverter&p=64360&t=4332#p64360

So you must simply solder one 1800 uF United Chemi-Con capacitor directly in the place where each 3300 uF Jamicon capacitor used to be.

[coulomb]

Here is the schematic trace for the final "power supply". It's actually the Bus Soft Start circuit, which Weber and I previously thought was done by slowly increasing the pulse width of the main DC-DC converter.



This shows that while TX2 and TX6 look similar, they are quite different in operation. When enabled by the DSP, this power supply will charge up the two 470 μF capacitors on the higher voltage bus (usually about 8x the battery voltage, 336-480 V).

I also updated the schematic trace for the primary power supply. Unlike most of the other power supplies, the primary power supply can only be disabled by the processor. The reason is clear; otherwise, you'd have no way of starting the machine!

But the fact that it can be turned off under processor control answers another mystery that has been bothering me. When the power switch is off but there is power from the SCC, it seemed to me that the processor always should have power. But it can sense the power switch, and after a half minute or so delay, the LCD light goes off and I can't seem to talk to it. I'd say it's because it's sensed the power switch is off, and after shutting things down in an orderly manner, switches off the main power supply via opto U8, which will kill its own power. [ Edit: I no longer believe it shuts down the main power supply if there is PV power available. ]

Earlier models used to be able to keep running from mains power, via the utility power supply involving TX6 (schematic trace a few posts earlier). But that power supply doesn't exist on later models, so turning off the power switch really does end up powering down the inverter. [ Edit: if there are no charging sources, i.e. no PV power and no AC-in. ] It never seemed to be consistent to me, but now I know why. My experience was based on a mixture of older and newer models, which actually behave slightly differently.

[ Edit: Added "via opto U8". ]
[ Edit: Target bus voltage is 336-480 V. ]
[ Edit: clarify that it's the primary power supply that can only be disabled. ]
[ Edit May 2019: Switching off only results in total power off if there are no charging sources. ]

[weber]

Thanks Coulomb! You have done a great job of enabling suitably-qualified people to repair their own PIP-4048MS, or equivalent Axpert, by tracing, drawing and posting these schematics for the sections that most often fail.