Carl

Yesterday I nearly decided to stop my journey , pack my solar suitcase, delete my forum account, and walk off in the sunset.
I have done so much homework in order to leave very little if any to the imagination regarding the repair of faulty IGBT bridges. And yesterday I started with great confidence on a machine that was destroyed due to inadvertant grounding of a PV terminal.
So I removed the 4 DC-AC converter's IGBTs and heatsink. Then switch on and check the drivers on DSO. All four looked pretty good, voltage levels -5.4 for turnoff and 15v for turning. Waveform shapes pretty good to my own basis of comparison.
Ok next I temporary solder test igbts onto the board  ,in two test phases  of half bridge only as I have described earlier in this topic. Passed, then I solder all 4 in full bridge mode . Switch on . 220vac beautifully. Doesn't get easier than this.
Then I left it running for a while , and because the igbts are not heatsinked, I touch the plastic case with my bear hands to gauge temperature.BOOM !  the abrupt  explosive sound so  heavy that I was completely deaf for a couple of seconds , then partial deafness for a good two hours after the bang.
I was devastated to say the least , this is not on , and somewhere my procedure have exposed to be flawed. I investigated the circuit and found the gate pin of the IGBT that I touch to be loose , dry joint.!!!!
HA HA , I felt better. Here is the thing .once a gate pin is left hanging , guess what ,the valves meet the pistons, timing belt broken.  The high impedance of the gate will  create indeterministic switching.
LESSON 1 when you do temp rigging , check the quality of your connections/joints.
So no I remove the 2 faulty igbts (it's mate in the same bridge leg will also get destroyed. Turn on and check waveform again.passed.
Solder 2 new test igbts in and this time used about 100 grams of solder to secure connections. Switch on .BOOM.BOOM.  now you can understand my opening statement.
I have now lost faith.  But our fathers have not raised scared children , reckless yes, but not scared. This time around there was no dry joints. I removed the 2 faulty IGBTs again , and tested the driver waveforms again and again and again.ha ha .one driver had a shaky -5 volt trace , what's more the waveform edges not so good anymore. Start tracing the driver.now the gate series resistor is 47R . But it measures around 30k.Shit , so the igbt went faulty due to the first BOOM  and blew this resistor   . But not open circuit , enough resistance to fool the DSO, but this resulted in very bad switching performance if the new IGBT that went during BOOM 2. So basically the we cannot switch the IGBT on abruptly and we cannot switch it off in good time.And that's what caused the BOOM 2. 
LESSON 2: please check this resistor even though your driver waveforms looks ok.
There was no BOOM3.  I feel better now.lots of lessons learned.

The 'assassinated' main PSU temporary repair now looks even more scarier after I repaired the -12V . This mess will only be addressed when the repair has met critical mass , i.e. if the board will become serviceable.
It reminds me of my engineering mentor back in the day . He insisted you must not over design or over repair something . If you design an F1 car , it must be able to win a race , but just after passing the chequered flag , the car must overheat, the gearbox must disintegrate , the suspension must fall apart , and the wheels must fall off with only a thin layer of rubber left on tires. Otherwise its over-designed and over-engineered !

Scope : 
I have been inspired by @Steve87 to embark on repairing , as he facilitated  3 x faulty MAX7.2 machines  for which I am very grateful . We repaired one machine , but that was really just repairing on module level (faulty  MPPT) . The second machine has two AC side IGBT devices shorted , but the rest of the main board seems ok , it gives fault code 09 (soft start fail). The third one was as dead as a doornail and this is the one that prompted me to repair and document my progress in order to help myself and others , because already I have so much to thank to @Coulomb and @maxo for some powerful partial / full schematics and in general the documentation available on the AUS forum of Coulomb and Weber .   So this machine  I am going to use to  repair in stages , as you will see it has some spectacular nasty faults/failures . It would probably not be economically viable to repair this board , but the focus is on learning how to repair other boards. In the process it will probably take me weeks to repair this one , if it can be repaired at all , but the end goal is learning and to possibly attract more gurus and expertise on repair/support.
The journey started when I switched on the Machine 3 for the last time  ,  then smoke escaped from the machine itself . I decided to disassemble it . The Machine still looks very neat inside , I believe its not an old machine as there is no dust layers etc etc and really looks new .   game on.
 
Chapter 1a :  The Baby Face Assassin 
So I tried to check for visual burns and damage , and finally saw a power diode that had burn marks . This is the diode that feeds the ac side igbt driver transformer and is fed via the SPS 12V secondary transformer . so I reckon this could be a shorted transformer primary causing the diode to burn. So I de-soldered the diode , and switched  the main board on via the SPS module , in anticipation that I have relieved the main power supply from a short circuit.  But , instead , all hell broke loose when I switched on again  ,  devices just started to explode and stink and smoke . Something on this board was just shooting from the hip for fun to make a fool of me. So I remember that old Eagles lyrics : ' .... I had to find the passage back to he place I had been before . Goodnight says the night man , you are programmed to receive , you can check out any time you like , but you can never leave ..."
The blown devices were all electrolytic capacitors on the 5V , 12V ,15V and -12V rails. So it became apparent the the SPS power supply on he main board  was a fault. Before of all of this I had removed the control board and as many removable modules , as I possibly could. Fortunately no damage o the boards themselves . Even the capacitors did not open up their shells , although the explosions at times had been horrendous. 
So I had to drill down and study the partial schematics . I redraw the sps power supply as shown below :

 
The SPS+ primary supply comes from three sources really , that is mains , battery or PV , they are wired OR together , that is not shown on the schematic. I limited my damage by introducing a variable 60V bench PSU courtesy of @Steve87. Current limited it to 500mA and set dc to 40V.   removed faulty capacitors and tried to isolate load circuits as far as i can , e.g. removing filter inductor to mosfet drive supply , cutting a track carefully to the bus soft start circuit etc.  
Switched on again , and long story short  , the SPS flyback converter around TX9 and U10 was just shoveling out as much DC on the 12V output as it received from the primary .  I measured the duty cycle of U10 and found it to be at 50% (50% by the way is the max limited by UC3845 type ). The 12V output was roughly as high as the supply DC input. So there is the cause of our explosions , 12VDC probably went as high as 50V !!!!. To limit this I merely turned down the DC input to about 16V in order to minimize my losses further. Note that the U10 chip will only bootstrap at 35V , so you must have that initially then you can ramp down again . In the future I will wire OR a separate 12V supply to the U10 VCC to bypass the bootstrapping.
So I studied the voltage closed loop feedback circuit . It basically samples the 12V output (R209/R210 on the schematic) and then controls a precision shunt regulator U9 , which in turn   controls the diode current thru opto-isolator U8 . U8 transistor is connected to the COMP input of the PWN controller , which in layman's terms will decrease duty cycle the more the COMP pin is pulled towards ground (GND) . Initially I checked if the COMP input works by shorting U9 output to ground. Indeed the PWM controller switched off. So at least some of the feedback circuitry is working . Then I discover that the VREF input of U9 is fixed at 0.00V . Which is shouldn't because of the R209,R210 divider across the 12v output . But it is just a simple divider , so I removed U9 in the hope that it has failed and pulled down the divider . Again I was made a fool of , as that divider midpoint stayed at 0V . In other words there was no feedback voltage provided to  decrease the duty cycle of the PWM . I tested both R209 and R210 . R209 showed 10K , R210 showed 10.5K in-circuit. The resistance of 12V down to GND is about 500 ohms , so this was a giveaway that R210 is open circuit. But how the hell, this is a 38.3K resistor how can it go open circuit .  Out of desperation I soldered a 39K resistor across R210 pads , with the original one still in situ . 
Switch on again , and the SPS was regulating at exactly 12.15VDC . I varied the input and even took it all the way up to 56VDC . Perfect . Short circuited the 12v output , limited perfect.
Can you believe a tiny 38.3K resistor R210 going open circuit !!!!!!!!!!!!!! . This resistor will henceforth be known as the Baby-Face Assassin.
Need to do a lot of cleanup around the SPS  before I am going to address the load circuits and repairing each section bit by bit . 
 
 Why such a big margin on the open loop transfer ? . Also there is no transzorb or voltage crowbar . I am going to breadboard a programmable crowbar around the TL431 and a triac and have it onto the 12V output as long as  am repairing the rest of the circuits. Below you can see my point , pulse width is 800ns to regulate at 12V from 56V input. That is about one fifth of 50% pulse width , which will indeed equate to around 60V with max duty cycle.
 

Yes that is normal , because the initial bootstrap is only there for short while. In that period the SMPS should produce , and amongst other , will generate the +15VDC supply over C75 , which will power U10 via D49  even after the bootstrap via D50 falls away. But because your SMPS does not deliver , when the bootstrap via D49 falls away , everything collapses.
 
The following test is provided to statically test the SMPS negative feedback loop. While no attempt is made to verify the linearity of the transfer function , it is adequate to test the extreme low point of feedback (resulting in highest output duty cycle) , as well as the high point of feedback (resulting in lowest possible output duty cycle) . To do so , we will manually power up the U10 and support circuitry with a power supply of 15V , and manually inject a variable 12V supply on the +12V output rail in order to fool the U10 that it is 'regulating'.
1. Remove all ancillary connections external from the main board , including controller board , fans , mppt , battery , everything.
2. You will require 2 x 12V variable bench supplies and an oscilloscope . Both of them must be current limited and both must be able to vary up to +15VDC.
3. Connect psu 1  positive to D57 cathode and negative to BAT- and set to +15VDC  . Current limit to 100mA or so. This will provide a steady VCC  on U10 pin 7 and power the chip .
4. Check there is steady +5V on U10 pin 8.
5. Check with oscilloscope that there is frequency and 50% (max) duty cycle on the gate of mosfet Q36 .  
6. Connect psu 2 set to 11.7 V between D54 cathode and GND . Set current limit to 800mA. This will false feed the +12V bus. At 11.7V the U10 output duty cycle should still max out at 50%. Check that the voltage drop over U8 diode is well below 1.1V . PLEASE NOTE : by injecting this 12V supply remember that the 12V loads on the main board is still connected . So please make sure that you dont push this voltage higher than say 12.8V.
7. Increase the psu to 12V and measure the TL431 bias voltage over R209 . It should be very close to 2.5V . Increase the psu to 12.5V and make sure the bias voltage is above 2.5V. at this stage TL431 turns hard on and current flows thru the U8 diode. The voltage across this diode should saturate at approximately 1.1V. This will in turn turn the U8 transistor hard on on the SMPS primary side . In turn this will pull the COMP pin 1 of U10 well down to 0V . As a result of this , U10 must correspond by outputting  its lowest duty cycle on its output  to the mosfet gate.
8. If all of this passes , the feedback loop circuit is pretty much in order.
... I will continue with this in another post  whereby you can bring the SMPS up with starting with a small 10V voltage at its dc input and working our way up to 40V if all goes well ...
 

The Guys over at Custom Lithium are building replacement packs for these Nissan leafs 40Kwh with 300kms +, given the average cycle life of NMC cells that would yield a life span of 600 000kms to 80% (2000 cycles with each cycle giving 300kms ) Lets work on worst case that would mean less than half the life span for our climate and or our current conditions ( say the battery only gave 1000 cycles ) thats still 300 000kms to 80% remaining capacity . Not bad i think .

Email me: [email protected]

This has been done before by people much smarter than me;
  Fixed Adj. 2 seasons Adj. 4 seasons 2-axis tracker % of optimum 71.1% 75.2% 75.7% 100% The table above shows that there is hardly any value at all in adjusting panel tilt each season over a fixed tilt array, however there is a very real advantage to a 2axis tracker over fixed array.
The above numbers don't really tell the whole story - for a fixed array to theoretically match the annual output of a 2axis array you would need to add 40% panel capacity, but the larger fixed array would still give low output in the morning and evening with a huge output in middle day, whereas the 2axis array would give a much smoother output throughout the day.
This smoother output would deliver significant benefits in real life over the simplistic result of math on a piece of paper showing equal annual output. To have your output pick up quickly in the morning and extend PV output late into the afternoon would have significant benefits in an off grid situation (and most other situations as well).
 
 

Hi Carl, Yes, would like that Leaf for the 20km school run. Will send detail by PM.
Have excess production on my solar PV to charge with.
Am in Gauteng but flying down to Cape for the holidays. However, will have to send the leaf by trailer...
 

Current range that it charges to is between 55 - 65 KM. Range is calculated after every recharge and is based on previous trip conditions.
The battery pack is a standard 24kWh pack and a new pack will give you 180 KM range in summer and around 160 KM in winter. That is for moderate driving.
Our long term average for over 2 years of driving is 7.5 KM / KWh (That works out at around 40 cents / km charging from the grid in Cape Town)

Yes, it will work. I have a 12V 100A LiFePo4 connected to a 720w Mecer inverter. 
However, the charge switch MUST only be set to 10A and not 20A. If you do that, the inverter will eventually fail. This is well documented in the mybroadband thread. 
Other small catch is that if you drain the battery enough, then the battery BMS will turn off the battery. When the power comes back the inverter won't turn on as there is no battery voltage and hence cannot start charging the battery. Solution is to use a voltage source that is high enough e.g. 13.8V and connect it momentarily to the inverter battery terminals. This will wake up the BMS and allow the inverter to turn on and charging to commence. 
For safety I only use a power supply that is isolated e.g. ye olde transformer based psu. 
Yes the Mecer does charge the battery at below the ideal 14.0V but so far so good and if necessary I will hook up a 14.0V (isolated) PSU 
to the battery to let the BMS balancer do its thing. 14.2 V is probably even better. 

@BritishRacingGreenyou have inspired me to attempt a repair on a faulty Growatt SPF 5000 ES inverter. With MPPT board removed and only battery connected i noticed surface arcing around the Ceramic caps and the DC to DC Bus Side IGBT's located on the bottom of the middle heat sink. There is a fair amount of black dust on the board and some moisture on the fan unit and on the board.  All the IGBT's and Mosfets test ok on the diode test function so hoping that damage is limited or best case no damage. Inverter came up with warning error 62. Please advise what is the best way to clean the board.

I have a nasty solar related problem , but an interesting one also. I passed on my old MKS3 (only 2 years old though) to my son-in-law , he added a Pylon UP5000 and used it as a backup system for a couple of months. Last week we added 4x555W panels to the mix (one string supply (170-180V  working voltage) .
I encountered a nasty failure mode in that the  PV will supply high loads or charge flat battery without any problems . When the PV demand is in the order of 1kw and higher , no problem. But when the load is only like 300watts and the battery is full, then the system becomes unstable . The battery voltage rise slightly , and of course affects a large charge current from the MPPT . when that happens , the PV drops again , and everything repeats at about 3 second interval. Add a large load , hairdryer at 1.5kw , everything is stable again. or add a flat battery , everything is stable again. weird. What worries me is that sometimes the AC output also fluctuates somewhat.
The MKS3 MPPT was just working fine when I decommissioned it at my own site.
So here is my theory . I think the MPPT regulation is faulty . I have seen it on the DC bus voltage curve on the HomeAssistant log. So when there is no load on the MPPT , the DC bus shoots up .  But when there is load , the regulation improves for the wrong reason , dampening the rise in DC bus voltage. Could be a faulty MPPT current sensor , I am taking wild guesses. 
There are suggestions it may be a panel related issue , but I have my doubts . When a panel is weak , or we have dry joints in the PV cables , we would experience just about the opposite  behavior as described above .
Fortunately , courtesy of our dear friend   @Steve87 ,  I have an MKS4 which I am going to test tomorrow at son-in-law site , in order to make sure in what subsystem of the installation the fault lies , panels or inverter .
So I will keep posted on my findings . One thing I have learned , is that my PV  commissioning procedure must be revisited in order to ensure that  PV operates adequately at various load conditions.
The interesting thing is that no error is offered , I think the poor regulation  is still within the DSP's windows of forgiveness , otherwise it would have raised a DC voltage too high error or so.
I hope its hardware  related and not require firmware upgrade . I am sh#t scared of firmware upgrade , so far I had no need to touch that , it makes me wake up at 3 in the morning , eyes wide open. Then I realize we have a mentor in @Coulomb , nightmare is over and I fall asleep again. 
 
 
 
 

This rebuild has been a major success. After about 10 cycles I am happy to report that the Hoppecke has been fully restored to produce & store energy again. 
The function of top or bottom balancing cannot be stressed enough. 
 
 

Yes, its a very unforgiving failure,  because the neatral is also grounded, the full bridge converter now gets shorted, and mostly leads to the full bridge IGBTs getting destroyed. 

Chapter 5 : Beginning of the End 
I am definitely not out of the woods yet , not sure yet if the fan controls and relay switching logic , control board and display will work , but I decided last night to introduce the control board and display. I have no way of verifying the operation of the control board , other than to make sure that there i no bad resistance between rail supply inputs etc.
So I inserted the control board and realized early that one can actually misalign when docking it to the main board , which I reckon could cause havoc , and undo all our hard work. I wired up the bare minimum , including the display .
Switch on , and after the usual startup timer timed out , it gave me Warning 01 , fortunately that is fan locked , so i plugged in the 3 fans . The machine then started up  without any further warning or error . obviously the battery icon flashing as well because I have not connected battery or grid to the power chain.
I was satisfied with the progress and decided to  add battery power  in the early morning , which will be an indicator whether all my hard work was worth the while.
No PV  modules added as that is the very last thing i will do in the final analysis . PV1 metrics does show up on the display though , but that is hogwash. I have seen it before on other machines as well when pv mppt module removed .
Today I decided to take the big step to add current limited DC power  to the battery port and bring up the power chain. A bit nervous though .I have two 30V variable supplies I cascaded to set to aggregate of 51V.
Here is me pre-charging the MAX battery input with suitable resistor.
 

and here is me going firing  51V DC , attacking the MAX full-on , no more 'Mr Nice Guy' :

 
The MAX submitted  immediately , produced about 380V on the BUS , opened its front porch with a beautiful sine wave 228VAC .  I also checked the IGBT gate drives and saw the magic in action . One half bridge had a continuously changing PWM to produce sine , wheras the other part of the half-bridge was constant at 50% duty cycle @50HZ.

Have I won the war ? not yet , but many battles yes , still has lots of work lying ahead . So the journey continues ....
 

Chapter 3 : The AC IGBT Drive Section
 
Haven't even started on a schematic trace here , hope to add it during next week.  But for me , the show already stops at the very beginning . The 12 SMPS flyback transformer provides an ac signal  which is diode half wave rectified , which in turn drives the primary of TX7 , the IGBT drive transformer. So I thought i will put the diode back because my rail supply is back to 12V again, in the hope that the igbt transformer was just incapable of coping with the faulty high voltage. Solder D20 back on , switched on and smoke came from the transformer . Ai, Ai , Ai. Brand new difficulties and possible cul-de-sac lying ahead. 
So in retrospect , when the SMPS failed and decided to throw max voltage on the bus , this TX7 transformer actually failed first and clamped the rail. Thats the reason why initially when i removed the hot diode D20 , all hell broke loose as i described in chapter one !!!!
I hate transformers , probably because I know too little about them.  The worst news is this transformer is Voltronics proprietary , so no ways I can even think about getting a spare . Only alternative is reverse engineering the device , and re-building from scratch. If you see what Coulomb and Weber do on their AUS forum , its clear that these guys fathers have raised them not to be scared. So inspired by them and others , I decided to rebuild this show-stopping sun-of-a-gun , how difficult can it be?.
Very fortunately , a member by the name of Holmes on Coulombs AUS site has done something similar , so I had a good idea what was lying ahead.
This time I did not have the luxury to cut the devices legs to desolder, so my solder wick skills was tested again , this time I managed well , but it takes a long time , and you need to be patient.
I fired up my gas-braai and cooked the transformer in boiling water for about 30 minutes. That softened up the glue sections , so I was able to dissasemble the split core halves. Next I removed the Mylar tape layer and started to unwound the the coils , obviously documenting it carefully . I am going to make proper notes of my rough sketches and publish it here at a later stage.  
Rewinding these devices is not rocket science , but one must be careful to keep the windings flat , or your next layer will become unmanageable. There is one primary winding , 21 turns of .3mm enamel copper wire , then there is three secondary windings each 35 turns , each winding section is nicely insulated with about two layers of Mylar tape. oh , and of course testing for isolation between windings as well (insulation).
I was not confident enough to just plonk it back onto the mother board . So I bench tested it , in order to both check  voltage gain in terms of turns ratios as well as polarity of winding in relation to each other . It passed all test surprisingly with flying colours .
So I soldered it onto the board , switched on , and I got the secondary voltages that the drive section desires . Next post I will post the drive schematics and show the waveforms of the secondary winding in more detail.
 
EDIT : oh , I forgot , all did not go plain sailing , when I first bench tested the re-winded transformer , the core got hot !. Long story short , there was a nasty air-gap formed by me not putting the split core properly together. So this is a forward mode transformer , it does not store and release energy like a flyback transformer (which is really a multi winding inductor) .  So the air gap caused inductance to lower , huge magnetizing current flow in primary.  So I made sure to mechanically secure the split halves by a healthy layer of Mylar tape.
 

The maximum demand charge is actually based on the highest average demand (KVA or KW) during a 30 minute window period starting at midnight on the beginning of the month.
So the billing period is broken down into 30 minute blocks and then you get charged for the block with the highest demand.
Maximum Demand = kVA = kVAh/T      (Where T = integration period of 30 minutes)
So from this you can see that the starting current of motors will be an insignificant contributor to your maximum demand charge.
So the trick is to reduce your energy consumption for any 30 minute block period.

The maximum demand charge is actually based on the highest average demand (KVA or KW) during a 30 minute window period starting at midnight on the beginning of the month.
So the billing period is broken down into 30 minute blocks and then you get charged for the block with the highest demand.
Maximum Demand = kVA = kVAh/T      (Where T = integration period of 30 minutes)
So from this you can see that the starting current of motors will be an insignificant contributor to your maximum demand charge.
So the trick is to reduce your energy consumption for any 30 minute block period.

Circuit Breakers can typically either have a thermal / magnetic or an electronic trip unit fitted. Some points to consider are:
1) Electronic trip units are far less affected by ambient temperature.
2) Thermal magnetic trip units typically have to be derated for ambient temperatures above 40 degrees C. (Eg. at 50 degrees, delta T = 10 degrees, 5% current derating)
3) Trip units are sometimes interchangeable for AC and DC circuit breakers. As @plonkster pointed out the thermal protection of the trip curves are identical since the bimetal strips are responding  to the RMS value. The instantaneous protection against short circuits are influenced by the "ferromagnetic phenomena" which in tun is affected by the circuit breaker design and the way the poles are connected. ABB have a coefficient called "KM" that the instantatneous trip value for AC has to be multiplied with to obtain the DC values. This coefficient varies between 0.9 - 1.3 
So coming back to @viceroyproblematic circuit breaker.
1) I would suspect that the breaker may have a high internal contact resistance which is causing  localised heating near the bimetal strip.   
2) Other factors such as undersized cabling (@Chris Louw), poor crimp connections, incorrectly tightened etc. will also cause localised heating resulting in a derating of the trip setting.  

Circuit Breakers can typically either have a thermal / magnetic or an electronic trip unit fitted. Some points to consider are:
1) Electronic trip units are far less affected by ambient temperature.
2) Thermal magnetic trip units typically have to be derated for ambient temperatures above 40 degrees C. (Eg. at 50 degrees, delta T = 10 degrees, 5% current derating)
3) Trip units are sometimes interchangeable for AC and DC circuit breakers. As @plonkster pointed out the thermal protection of the trip curves are identical since the bimetal strips are responding  to the RMS value. The instantaneous protection against short circuits are influenced by the "ferromagnetic phenomena" which in tun is affected by the circuit breaker design and the way the poles are connected. ABB have a coefficient called "KM" that the instantatneous trip value for AC has to be multiplied with to obtain the DC values. This coefficient varies between 0.9 - 1.3 
So coming back to @viceroyproblematic circuit breaker.
1) I would suspect that the breaker may have a high internal contact resistance which is causing  localised heating near the bimetal strip.   
2) Other factors such as undersized cabling (@Chris Louw), poor crimp connections, incorrectly tightened etc. will also cause localised heating resulting in a derating of the trip setting.  

Technically this is called "Current Level Discrimination" and is a technique linked to the staging of the Long Time (LT) tripping curves of two serial-connected circuit-breakers. (Quoted from the Schneider Electrical installation guide)
An interesting point is that overload protection of a cable can be implemented either at the source or at the load side but short circuit protection has to be implimented at the source.
@plonksteris spot on.

Technically this is called "Current Level Discrimination" and is a technique linked to the staging of the Long Time (LT) tripping curves of two serial-connected circuit-breakers. (Quoted from the Schneider Electrical installation guide)
An interesting point is that overload protection of a cable can be implemented either at the source or at the load side but short circuit protection has to be implimented at the source.
@plonksteris spot on.

Technically this is called "Current Level Discrimination" and is a technique linked to the staging of the Long Time (LT) tripping curves of two serial-connected circuit-breakers. (Quoted from the Schneider Electrical installation guide)
An interesting point is that overload protection of a cable can be implemented either at the source or at the load side but short circuit protection has to be implimented at the source.
@plonksteris spot on.

@Jaco de Jongh, Ah this is something different, still not eddy currents though.
Eddy currents by their very nature are current flow internal in a metal and are effectively flowing in a short circuit so there is very little voltage. 
This is induced current, but no real potential build up, think of a lump of metal as being made up of windings of a transformer but each winding being shorted out. Plenty internal current rushing about and plenty heat but no voltage build up to speak of. 
I think your dealing with the capacitive effect of an electrostatic charge which does have plenty voltage, but normally not much current.
Like the little shock you get when you touch the door handle from some carpets.
You have all the components a large plate being the panels and the roof combination, the other plate being the ground, and high frequency MPPT switching.
The higher the frequency the lower the capacitive reactance (it is inversely proportionate to frequency) so it could probably develop enough current to deliver a nice shock.
Mmm, interesting.
 
 

Technically its a fault voltage that we come into contact with that will drive a current through us.
Our impedance is determined by many factors such as the current path through the body, amplitude of the applied voltage, duration of the applied voltage, frequency of the applied voltage, contact surface area, pressure exerted and skin temperature.
IEC 479-1 Table 1 (for the highest risk 5 % of the population) shows that the total impedance between dry hands varies from 1750 ohms at 50 V rms to 700 ohms at 1000 V rms. The impedance of an arm is very similar to a leg. So the values mentioned also apply for the current path from one hand to one foot. The more typical situation is where the current path is from one hand to both feet and then you can use 75 % of the bove impedance value. 
So start by measuring the fault / touch voltage.