PIP inverter repairs and hardware modifications

[Howard]

Hi Weber,
OK ..thanks so much for the info
I will be an expert at pulling these PIP'S down.Two more pull downs and two re-assembles to go.
By the way Paulvk and I have put two external 120mm 240Vac fans(running from 240VAC O/p) at top cooling exists with temp.controller probe on Batt.side of heat-sink set to come on at around 39C .It runs super cool now which should extend life of all components in PIP.Hope this helps PIP users and the digital temp LED readout gives immediately heads up situation operation during charging.
Thanks again Weber -will post when Ive done the cap mods. on them and restart.

[weber]

I recently invested in these two tools, which greatly speed up the PIP upgrade or repair process.

[jonescg]

Is that a soldersucker? And the Rolls Royce version at that?

[T1 Terry]

This unit by the look of it https://rhinotools.com.au/product/desoldering-station/#reviews love the fact there are lots of spare parts available and the less than perfect reviews were quickly addressed.
Their cable lugs are well priced too, have been a customer of theirs for quite a while now.

T1 Terry

[Edit: Made link clickable]

[weber]

Yes that's the Rhino Tools ZD-985 desoldering station and the Hitachi DB3DL2 cordless screwdriver. I looked at many online reviews before settling on both of these. And so far, I'm very pleased with both.

It's "Rolls Royce" (or should we be saying "Tesla Motors"?) compared to my previous -- the kind where you have to have a muscly thumb to keep reloading the plunger, and lightning precision and coordination to whip the soldering iron out of the way and jam the nozzle in the right place and simultaneously hit the release button.

But it's not "Tesla Motors" compared to other desoldering stations that cost 3 times as much. It's merely the best of the Chinese cheapies. It got a "good enough" review from Dave Jones of EEVBlog.
https://www.eevblog.com/2013/11/01/eevblog-542-zd985-desoldering-station/

And you might think you can use your cordless drill to do the same job as a cordless screwdriver, but you'd be surprised what a difference it makes to have a much lighter-weight device, and to be able to grip it in this manner:



It only comes with one short Philips #2 bit. The long bit in my earlier photo, that you need to get in among the guts of a PIP, must be bought separately. Standard 6.35 mm hex drive.

[jonescg]

And you might think you can use your cordless drill to do the same job as a cordless screwdriver, but you'd be surprised what a difference it makes to have a much lighter-weight device, and to be able to grip it in this manner:

I tried a cordless screwdriver for screwing the buslinks down on my Li-Po batteries, but I find it's still a little too clumsy for my liking. It's easy to slip and make sparks in short order.

But for everything else, like taking the side panels off the battery pack etc (where there's 25,000 screws) it's invaluable.

[coulomb]

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:

The above is my guess of where the "shorting bars" would go (as seen from below the board), based on Weber's theory. Weber and I were working on a PIP repair today, and had to remove the battery side heatsink to clean away the soot (it can conduct rather well and cause big problems).

So now we can compare my guess with the reality:



I think I did OK, except I wasn't generous enough with area of the metal.

[ Edit: Fixed a muck-up where I ended up with just a complete quote of the original post. ]

[coulomb]

After Weber removed the dead battery-side MOSFETs and cleaned the soot off the PCB, we were ready to try testing the gate drivers. We've been trying to devise a way of testing the gate drivers, other than reassembling the main board in a case, attaching a 48 V battery, and powering up. After all, that's what we did last time, and the MOSFETs blew up. There has to be a way to check the gate drivers without having things blow up.

Having traced some schematics lately, I had a plan. It involves testing the drivers after the old battery-side MOSFETs are removed, and before they are reinstalled. Firstly, we use a current limited power supply. Weber's is limited to 31 V, but it seems to be good enough. We set the current limit to one amp (half an amp would be plenty). Secondly, there needs to be power for the driver electronics, which means powering up the main power supply. Shorting the wires that lead to the main power switch does that. There is a capacitor, C7, that will charge up and cause Q9 and hence Q10 to stop conducting, but once Q36 and TX9 are generating pulses, there are power from diodes D57 and D49 to keep U10 powered.

Thirdly, something needs to enable the PWM chip that generates the square waves for the DC-DC gates, both the MOSFETs on the battery-side, and the IGBTs on the high voltage (~400-450 V) side. Usually the processor would do that, but we'd prefer not to plug in the processor daughter board at this stage. Shorting pins 3 and 4 of opto U19 would do this.



The red oval shows the two pins to be shorted. Using a long pigtail as shown enables a DSO earth lead to be clipped to it, should the need arise to debug the PWM chip. You may wish to use a clip lead instead of soldering to the pins.

[ Edit: U9 is near the middle of the board, near the end with the battery terminals. ]

We didn't want to remove any IGBTs, since they seem to be working fine, and we'd prefer that we didn't have hazardous voltages running about, not to mention IGBTs that could blow up if not driven correctly. But the transformer won't be providing energy with the battery-side MOSFETs not yet installed. We know now that the bus soft start circuit is usually enabled by the processor, so with the processor removed, that won't be charging the big bus capacitors either. So the IGBTs will have no voltage across them, so turning them on, with good signals or bad, won't hurt anything.

So we powered it up, and attached the DSO leads to the gate and source of the four sets of battery-side MOSFET holes. The result:



I don't know the reason for the step change part way through the square wave to the gates, but it looks pretty good to me. These two waveforms are from diagonally opposite pairs, hence the difference in phase. Note the ~1.7 uS dead time when neither diagonal of the full bridge (i.e. none of the MOSFETs) is turned on. The frequency of the gate drive is about 37 kHz.

We also checked the drive to the IGBTs on the high voltage side of the DC-DC converter. These are the four TO-247 devices that are nearest U9. The signals were similar, except for some overshoot at the start of the square wave drive.

It all looks good, so we may be able to get this PIP working again. And nothing blew up! [ Edit: so far... ]

If you try this, make sure you remember to unsolder the short on opto U19 when finished, if you used solder to short the pins.

[ Edit: clarified "neither diagonal of the full bridge" with "(i.e. none of the MOSFETs). ]
[ Edit: "battery side MOSFETs" -> "battery-side MOSFETs". ]

[paulvk]

If you are going to bench test the board before replacing into case and have access to a CRO would be good to see the ripple on the caps.

[coulomb]

Weber and I are working on another PIP-4048 this project day. This one had blown MOSFETs that we had recently replaced. We were interested in how many MOSFETs we would have to replace, given that the replacements would have the exact same part number (but they'd almost certainly be from different manufacturing batches). The service manual says "if one or more of them were damaged, please replace all of them".

This was a definite short circuit situation (there was a flash visible from the side of the cabinet). But it seems to us this implies that two sets of MOSFETs (one high set and one low set) must have short circuited, but this short would have protected the other set from damage. Only one set of MOSFETs showed obvious damage, one of the high-side sets. These showed a short circuit drain to source, and source to most gates. So we removed those first. Some of the measurements of the remaining MOSFETs were being affected by large capacitors. In the end, we decided to follow the advice of the service manual. One should see 11.7 kΩ from gate to source (basically, the four 47 kΩ gate to source resistors in parallel), and the reverse voltage (using diode range on the multimeter now, not resistance) from drain to source should be 0.43 V. In one case, it was zero, indicating a shorted set of MOSFETs. The last MOSFET out seemed to measure OK, but we figured that it would have received a share of the short circuit current, and was worth replacing anyway. With those bad MOSFETs removed, the gate to source resistance and reverse drain to source voltage measured OK. So we feel that it should be OK to retain the other 8 MOSFETs.

So that seems to be the best way to find the bad MOSFETs. We still need to come up with a way of better testing the gate driver circuit, since this is presumably the reason for the immediate failure of the new MOSFETs.

Edit: be aware that when removing the MOSFETs, solder will tend to be sucked away from the 47 kΩ and the 22 Ω resistors connected to the MOSFET gates. It's probably a good idea to check for the 11.7 kΩ across any of the 47 kΩ resistors after installing the replacement MOSFETs. If any of the 47 kΩ or 22 Ω resistors are open circuit, you'll read something quite different from 11.7 kΩ.

Edit: This technique of measuring the gate resistors can't really be used on the IGBTs on the high side of the DC-DC converter, since the pulse transformers have very low output resistance. However, a reading of 22-23 ohms gate to emitter would be a good indication. If all 4 of these IGBTs are good, there should be a high resistance from collector to emitter on all 4 devices.

[Revlac]

HI Coulomb
I had an idea there would be many different factors involved on different applications, anyway I'm still learning these things.

The other night I had a look at the IGBT gate drive on the PIP4048 main board using a microscope and found either a crack or short across the resistor (R139) that drives QA1 if that was a short then i guess the AT350 opto could be shot, i will check that later.
As the manual said check for obvious damage i didn't see this.
(picture didn't post)


I'm familiar with H-Bridge drive as i have had some success and some failure building my own inverter, it really didn't like the bad signal from a dry joint, end result was short circuit drain to source, Just something els to fix when I find the time.

Cheers
Aaron

[coulomb]

Weber and I have disassembled and reassembled PIP inverters many times. Eventually, we wrote down some notes on what order to do things. It's now settled down enough that perhaps it will be useful to someone else.

Disassembly

• Remove the cable cover.
• Remove the front cover; unplug the two cables from the board on the inside of the cover.
• Unscrew the 6 screws from the AC transfer board.
• Unplug the AC transfer board's cable from the main board.
• Remove the red and black wires from the SCC board.
• Remove the 3 white plastic plugs holding the clear plastic shroud. Remove the shroud.
• Unscrew the SCC connection to the PV input connector, and the 5 M3 screws holding the SCC board. You don't need to remove the two screws attaching the SCC board to the SCC top panel. Don't unscrew the 4 screws holding the heat-sink to the SCC board.
• Remove the 4 countersunk screws from the SCC top panel. Unplug the 4-pin plug from the SCC board.
• Remove the SCC board and top panel while unplugging its 2 wire cable from the main board.
• Remove the blank piece of PCB material.
• Undo the two screws holding the processor daughter board. Remove the daughter board; no need to disconnect its cables.
• Remove the two countersunk screws holding the fan panel, from the two sides of the case.
  Note: Now seems like it would be a good time to remove the 4 quick connect terminals to the AC input and output, so as to remove the fan panel. But the quick connects are extremely tight. The one time we removed them we found (a) this required extreme force with pliers, which damaged the insulation, and (b) it made them loose when replaced, such that we don't think they would take 40 A without overheating. So leave the quick connects alone and instead take the fan panel out with the main board when you remove it (below).
• Remove the nuts, washers and lugs from the earth stud.
• Unplug the two small plugs along the bottom of the main board, but not the two larger fan plugs, and the AC start switch plug in the top right corner.
• Remove the 9 M3 screws holding the main board, and the two screws holding the battery terminal tongue. (One is under a date sticker.)
• Undo the nut holding the circuit breaker to the chassis.
• You should now be able to remove the main board.

Reassembly

• Insert the main board and fan panel into place. Ensure that there are no cables trapped under the board: fan cables if removed, power cables to the comms and parallel boards, and the AC start switch cable. Plug in all these cables to the main board.
• Guide the circuit breaker back into place, and replace its nut.
• Replace the 9 screws into the main board, and the two into the battery terminal tongue. The round head screw goes into a hole marked with an earth symbol, near the AC transfer board standoffs.
  The tongue uses a chrome plated hex head and a black countersunk screw (the latter in the date label).
• Insert the processor daughter board and its two screws.
• Screw the fan panel into place with 2 countersunk screws. Replace the items on the earth stud, in the order: lug, washer, nut, lug, washer, nut. Tighten the first nut before installing the second one.
• Plug the SCC cable (the 4-wire cable furthest from the edge of the processor board) into CN4 of the SCC. Plug the 2-pin cable from the SCC onto the main board, beside the AC start switch socket.
• Screw in the SCC top panel; it uses 4 countersunk screws.
• Place the small piece of blank PCB with the side that has three holes under the SCC board. One way up works better than the other.
• Screw the SCC board in place, using 5 M3 screws and two larger screws into the PV input connector.
• Replace the clear plastic shroud. Fix it in place with the 3 plastic plugs. Use the three holes furthest from the blank PCB.
• Screw in the red and black wires to the SCC board. The colour is indicated in words on the SCC board ("RED", "BLACK").
• Plug the 8-pin plug from the AC transfer board into the main board. Screw in the AC transfer board with 6 M3 screws.
• Plug the LCD and LED cables into the board on the inside of the front cover, and tuck the bottom of the front cover behind the lip in the case. Replace the countersunk screws holding the front cover.
• Remember to replace the cable cover once the cables are re-connected.

[ Edit: moved the copy from page 73 to here. ]
[ Edit: many improvements suggested by Weber. ]
[ Edit: Remove and replace the nut holding the circuit breaker. ]
[ Edit: replace items on the earth stud. ]

[coulomb]

I have re-drawn the power topology schematic from this post, to reflect the details of the actual PIP-4048 circuit. You can see the original and modified partial schematics side by side in the original post.

[ Edit: Inverter switches now drawn as IGBTs; renumbered. ]
[ Edit: Redrew IGBT diodes to better suggest that they are internal (same package as the actual IGBT). Renumbered QA1 to QD2. ]

[weber]

Oh joy. Oh bliss.

Some of you may remember back in February, that one of my PIPs blew up when I turned on a vacuum cleaner, although that had been done many times before with no problems.

Well, that PIP is working again at last! And not only working, but running the latest beta patched firmware in parallel with my other PIP. Hoorah!

Many thanks to Coulomb for all the hours he put into fixing it. We learned a lot -- some of which has appeared as the schematics Coulomb has posted above.

You might well ask what took us so long? It was partly that there were more urgent or interesting things to work on, and partly that we tried to fix it 2 times previously and failed! My memory is a bit hazy about the first two times. Coulomb might correct me. I think the first time, we replaced the battery-side MOSFETs and various of their blown driver components. We also took the opportunity to do our usual upgrade to 80 V long-life low-impedance caps and 100 V MOSFETs. But it went bang again on test. We found MOSFETs blown again, but this time we noticed two of the sine-wave-bridge IGBTs were blown too, and assumed that they had been blown all along.

Probably those IGBTs went first when the vacuum cleaner was turned on, and their shorting took out the MOSFETs. We also checked driver components much more carefully, for both the IGBTs and MOSFETs, and replaced some more MOSFET driver parts. We replaced all four sine-wave IGBTs, but only replaced the half of the MOSFETs that were actually blown. This time it didn't go bang, but it didn't work either. We got the usual "09" fault code, which translates to "bus soft start fail". This time we found that the MOSFETs were fine but the new IGBTs had blown. We pored over their driver circuitry again but could find nothing wrong.

We decided to replace the T350 opto-isolated driver chips just in case, so I ordered them, and more of the same IGBTs. These were IRGP4066PBF. Soon after, Coulomb posted a schematic that included the sine-wave-bridge IGBTs, and I noticed he didn't have designators for the associated diodes. So I went to find them on the circuit board so I could tell him what they were, because I knew from the datasheet that the IRGP4066 did not have built-in diodes. But I couldn't find any such diodes on the board. And then it dawned on me. Oh $#!+.

The IGBTs were supposed to have built-in diodes! I checked the service manual and saw that the IGBTs were described as IRGP4066DPBF. I had foolishly assumed the entire string of letters on the end was irrelevant. But it turns out that "D" is real important!

So then I ordered the right IGBTs. Actually I ordered some with a higher voltage rating (650 V instead of 600 V), but the same or better specs otherwise -- the IRGP4790DPBF. I installed them today, along with the new drivers (just in case).

When I was cleaning off the flux around the IGBT pads after soldering, I noticed a thin black line between gate and drain of one of the IGBTs. As I scraped at it, I realised it was underneath the conformal coating. I believe this was the cause of the original failure. I suspect 2.5 mm is barely enough creepage for 450 V dc. I removed the conformal coating, scraped off the black tracking, cleaned with solvent and re-sprayed with conformal coating. We shall see how it fares in the years ahead.

[Howard]

[Moderator note: This is partly a response to this post in the PIP-4048MS and PIP-5048MS inverters topic.]

Hi Coulomb,

Thanks so much for your swift reply!! I had to get back to other more menial tasks , apologies for my delay in reply.

So just confirming 72.70 with no suffix to get my desired settings? The link search just 72.70?

Congrats. on getting the blown PIP up ..at last!! I wish I had known about that lousy track when I had my PIP in bits for cap replacement.
Might be good to have a pic of where that track is ..Pray we don't have to go down that path!

We march on ..Thanks again for your work and help!
Best wishes,Howard

[coulomb]

So just confirming 72.70 with no suffix to get my desired settings?

Yes, official firmware 72.70 has the settings numbered 38 and 32 that you mentioned. It also still has the charge bug and a few other limitations, so our patched firmware such as 72.70b and soon 72.70C has all the features of official firmware 72.70, the charge bug fixed, and several other goodies.

The link search just 72.70?

It used to be easy to find firmware files with a simple search, but for whatever reason the manufacturer is not releasing them widely any more. You can follow the links in the index in the first post of this topic to find official and patched firmware files.

Congrats. on getting the blown PIP [fixed] up ..at last!! I wish I had known about that lousy track when I had my PIP in bits for cap replacement.
Might be good to have a pic of where that track is ..

Actually, it wasn't a bad PCB track that was the problem, it was carbonising between copper tracks. Weber has added a link to this Wikipedia article, where it says:

"Tracking is an electrical breakdown on the surface of an insulating material wherein an initial exposure to heat chars the material, and the char is more conductive than the original insulator, producing more current flow, more heat, and eventually complete failure."

The insulating material here is obviously the fibreglass to which the copper tracks of the Printed Circuit Board adhere.

[Howard]

Hi Coulomb,

All good Installed 72.70b with thanks.No issues ..still prayed!
Setting 38 now available for N earth.
Not sure how setting 32 works ..? I set 32 to 10mins?
Is the Post by you 24th Feb2017ad latest and easiest to understand on absorption/ BC process,setting etc ?

Thanks for info on o/heat issue fiberglass ..Do we know whether this has been modded?

Best ,
H

[weber]

Hi Howard,
I checked, and 2.5 mm is enough creepage distance for 400 V when conformally coated. So the original trigger for the carbon tracking I saw, was most likely a one off event, such as the soldering iron tip accidentally being dragged across the epoxy surface of the printed circuit board between the IGBT's gate and drain pads, during assembly. So it isn't something you need to worry about.

[Howard]

Hi Weber,
Thanks Mate.Are you going into production with the EV MX5?

May be you can give me some ideas on my VR commodore when I get round to it:) Big dreams..
Post SOLAR Inverter panel installs ..Hopes..

[Moderator note: Weber's response to the above is in the PIP-4048MS and PIP-5048MS inverters topic.]

[rezydent]

Hi all,
first please apologize for my poor english.
I am runing a new solar inverter from MPP SOLAR PIP 4048 MS 5000VA DC input 48VDC AC output 230 VAC
There was an error (09).The inverter has stopped working. What should I start with, check what's wrong with it ? I will be grateful for your help, I will be grateful for your help.

[weber]

Hi all,
first please apologize for my poor english.
I am runing a new solar inverter from MPP SOLAR PIP 4048 MS 5000VA DC input 48VDC AC output 230 VAC
There was an error (09).The inverter has stopped working. What should I start with, check what's wrong with it ? I will be grateful for your help, I will be grateful for your help.

Hi rezydent. This typically means the MOSFETs and capacitors on the battery side (right-hand side) of the inverter have failed, and when they fail they typically take out some of the MOSFET driver components as well. See the "Hardware" section of the index post at the start of this thread. viewtopic.php?t=4332

Edit: Since this is a new inverter, if you know you have not mistreated it, you should be making a warranty claim with the supplier.

[rezydent]

Hi all,
first please apologize for my poor english.
I am runing a new solar inverter from MPP SOLAR PIP 4048 MS 5000VA DC input 48VDC AC output 230 VAC
There was an error (09).The inverter has stopped working. What should I start with, check what's wrong with it ? I will be grateful for your help, I will be grateful for your help.

Hi rezydent. This typically means the MOSFETs and capacitors on the battery side (right-hand side) of the inverter have failed, and when they fail they typically take out some of the MOSFET driver components as well. See the "Hardware" section of the index post at the start of this thread. viewtopic.php?t=4332

Edit: Since this is a new inverter, if you know you have not mistreated it, you should be making a warranty claim with the supplier.

Unfortunately, the warranty has expired unscrew the casing and see what has been blown. If I have a problem with this inverter, I will ask more questions.

[coulomb]

One of our clients' PIPs was updating firmware very slowly. We were able to swap its comms board with one from another PIP to get the firmware updated, but the original comms board still needs fixing.

To help find out what was going on, we did a partial trace of the comm board. We were not interested in the USB half of the circuit, so that that was not traced.

I'm not convinced that I have the power supply right; it seems pretty weird. It seems that the design is a little marginal, which might explain why some people have been having trouble with certain combination of inverter/charger and USB to serial adapter.


[ Edit: Changed schematic; C9 was on wrong side of R15. ]
[ Edit: D9 was mislabelled as D6. ]
[ Edit: grammar ]

[weber]

That's a nicely done drawing, Coulomb. There's more than one weird thing about that circuit. It looks like it was designed by an amateur. I count 12 parts that are completely unnecessary since they can either be omitted or replaced by a PCB track, sometimes requiring a change in the value or connection of other parts. Namely: Omit D5, D9, Q3, R12, R14, R18. Replace D2, R1, R6, R9, R10, R35 with PCB track. Increase R3 to 1k0. Increase R13 to 6k2. Connect the LED of opto U2 in series with R13.

You show "To USB circuit". It would be interesting to also see "From USB circuit".

[Tejota]

There is a special-order model of the PIP-4048 that allows up to 64 V of charging, I assume it has different MOSFETs and higher voltage capacitors, and different firmware.

MOSFETS in PF1 5kW up to 64V charging are TK100E08N1

I have seen them through the grid of holes. I dont know if there are 16 or more. I cant watch capacitors.... I would have to open the device invalidating the warranty.

A question about parallelized pips: If mosfets/igbts in one unit of the cluster fail, what happens in the other units of the cluster?

Regards.