Hacking GTIL On-grid Inverter

I am trying to hack the Sun GTIL inverter's dc controller board (which manages the input side of the GTIL blue inverters) which uses a Microchip dsPIC33FJ64GS606 PIC. Am doing this because one of mine died, and I can not get a replacement board from the manufacturer.

I think the chip is still ok, just lost some of its code. The problem is that the PIC does not now implement the MPPT function to havest max energy from solar panels. It just drifts from 20-200 watts at random. The inverter was making correct power (about 1550 watts) just before the PIC lost its mind. Does anybody have any experience using Microchip's MPLAB IDE tool? Or does anybody know how to clone one PIC from a known good PIC.

I have the PIC spec sheet, and have ordered the Microchip MPLAB PG164100 Debugger.

thanks

That does not necessarily sound like the code is wrong as it may be the input A to D conversion which is reporting fluctuations resulting in the tracking going all over the place. The reason for the variations may be due to subtle changes in voltage under load (MPPT) which with a faulty capacity may end up reporting a larger V drop than what it should be, The caps may seem ok from a static test or prope measurement, however with a scope they should show a lot of ripple.

completelycharged said:

That does not necessarily sound like the code is wrong as it may be the input A to D conversion which is reporting fluctuations resulting in the tracking going all over the place. The reason for the variations may be due to subtle changes in voltage under load (MPPT) which with a faulty capacity may end up reporting a larger V drop than what it should be, The caps may seem ok from a static test or prope measurement, however with a scope they should show a lot of ripple.

Click to expand...

So maybe I should just change out the input caps?

Potentially.

If you have access to a scope it should show up. Could be the main input caps or in the the step down circuit before the A to D.

Usually the component that fails is the caps and mainly due to a unit either being run a long time or at a too high a temperature which leads to early failure.

completelycharged said:

Potentially.

If you have access to a scope it should show up. Could be the main input caps or in the the step down circuit before the A to D.

Usually the component that fails is the caps and mainly due to a unit either being run a long time or at a too high a temperature which leads to early failure.

Click to expand...

My old Tektronix scope died, but I have a new 100 mhz on the way. I have a 125 watt 48 VDC bench supply I use for testing these inverters. When the inverter is working properly, it produces 110 watts output to the grid, and the MPPT function loads the power supply to about 48 VDC. This inverter, which is not working correctly as previously described, produces about 75 watts to the grid, but the MPPT function only loads the bench supply to about 53 VDC. The output power is very steady, only fluctuating +/- 5 watts. Tried to attach video, but file size too large.

suntan901 said:

completelycharged said:

Potentially.

If you have access to a scope it should show up. Could be the main input caps or in the the step down circuit before the A to D.

Usually the component that fails is the caps and mainly due to a unit either being run a long time or at a too high a temperature which leads to early failure.

Click to expand...

My old Tektronix scope died, but I have a new 100 mhz on the way. I have a 125 watt 48 VDC bench supply I use for testing these inverters. When the inverter is working properly, it produces 110 watts output to the grid, and the MPPT function loads the power supply to about 48 VDC. This inverter, which is not working correctly as previously described, produces about 75 watts to the grid, but the MPPT function only loads the bench supply to about 53 VDC. The output power is very steady, only fluctuating +/- 5 watts. Tried to attach video, but file size too large.

Click to expand...

I received 10 new 1000uF 100 V caps of Chinese origin. No way to test them, as my capacitance meter (in my LG DVM)







only goes up to 2 uF. So I made a "supercap" out of 8 of them and installed across the input of the inverter, per the photo. On the bench, the inverter performance is the same, 75 watts +/- 5 approximately. Normally 110 watts on the bench, power supply limited. Realizing that the caps really need to be very close to the actual SMPS mosfets as any lead length increases ESR, I tried to see what kind of noise appeared across the input of the inverter using a primitive "scope". Results inconclusive, like random noise almost.

But it seems to me that the additional capacity should have changed the behavior of the inverter on the bench. Am I wrong?

The additional capacitance would have made some difference if it was the input buffer caps.

Can you clarify :
"On the bench, the inverter performance is the same, 75 watts +/- 5 approximately. Normally 110 watts on the bench,"
vs "It just drifts from 20-200 watts at random"

"This inverter, which is not working correctly as previously described, produces about 75 watts to the grid, but the MPPT function only loads the bench supply to about 53 VDC." - I'm guessing the power supply is a transformer and bridge so the 53V vs 48V is just lower loaded so the power supply voltage is closer ot 68V peak ?

completelycharged said:

The additional capacitance would have made some difference if it was the input buffer caps.

Can you clarify :
"On the bench, the inverter performance is the same, 75 watts +/- 5 approximately. Normally 110 watts on the bench,"
vs "It just drifts from 20-200 watts at random"

"This inverter, which is not working correctly as previously described, produces about 75 watts to the grid, but the MPPT function only loads the bench supply to about 53 VDC." - I'm guessing the power supply is a transformer and bridge so the 53V vs 48V is just lower loaded so the power supply voltage is closer ot 68V peak ?

Click to expand...

I was unclear. The 20-200 figure is not on the bench, rather connected to 2 kw worth of pv panels. The 110 watts is what the inverter(s) produce when working properly, fed by my bench "supply", a 48 VDC li ion battery charger. When working properly, the inverter(s) MPPT function load the bench "supply" to 48 VDC. Don't know what's in the bench "supply", guessing a smps with regulator.

I have ordered "good" caps from Digi-key, 6000 hour rating, guess I will install and see what happens. Afraid it may be a waste of time to install the cheap ones, not knowing the quality.

Ok, got it. The charger is just being loaded up to the CC limit at V max which pulls the voltage down to 48V.

"75 watts +/- 5 approximately" vs "20-200 watts at random"

At the moment I would still suspect it is to do with the input side, the reduction in variation with the additional caps could be promising. What are the values of the capacitors in the existing unit ? I have a similar unit (GTIL) however it is a 1kW wind inverter, so the caps may be different compared to the solar units. I think my unit had 100V caps on the input side.

completelycharged said:

Ok, got it. The charger is just being loaded up to the CC limit at V max which pulls the voltage down to 48V.

"75 watts +/- 5 approximately" vs "20-200 watts at random"

At the moment I would still suspect it is to do with the input side, the reduction in variation with the additional caps could be promising. What are the values of the capacitors in the existing unit ? I have a similar unit (GTIL) however it is a 1kW wind inverter, so the caps may be different compared to the solar units. I think my unit had 100V caps on the input side.

Click to expand...

My 2 kw GTIL uses 10 1000uF @ 100 V caps on the input. The Input side (smps boost from 48 VDC to about 800 VDC) is a total of eight 250 watt smps boost converters. The DC Controller Board drives all eight using a PIC33FJ (with 4 analog inputs and numerous PWM outputs), half dozen more IC's, and 5 isocouplers, 4 of which seem to drive the eight boost converters in pairs. Each 250 watt boost converter uses 2 MOSFETS as a switch.

I guess I muddied the info on the bench testing, but let me clarify; The addition of 8000uF across the input terminals made no difference when testing on the bench. Inverter still produces 75 +/- 5 watts, and the bench "supply" is only loaded by the inverter to about 53 VDC. When this same bench "supply" powers a properly-working inverter, the inverter produces about 110 watts, very steady, and loads the bench "supply" to abput 48 VDC.

I have not been able to see the part numbers of the other ICs on the DC Controller board, because the conformal coating obscures the part numbers and must be scraped off very carefully, but I suspect the a to d is actually within the PIC, not a separate device. I tried swapping out the entire DC Controller Board with one from a working inverter, but they are different versions (v5.5 and v5.4), and substituting causes the inverter under test to shut down after 1 sec, and will not restart.




Drawing is of PIC33FJ, exact part used in GTIL. Red or blue lines indicate actual connections in the GTIL DC Controller board.

What I was noting was the 20-200 random vs less variablity on the bench.

The DC HV boost arrangement is a standard approach with high frequency inverters. If the DC bus voltage is stable then that cuts all of the input side issues out.

On the unit, can you pull up the screen that shows the voltage waveform and have the working unit do that side by side to see is the faulty unit is displaying any distortions ? If the A-D input/signal has issues it may be showing up on the waveform readings ?

completelycharged said:

What I was noting was the 20-200 random vs less variablity on the bench.

The DC HV boost arrangement is a standard approach with high frequency inverters. If the DC bus voltage is stable then that cuts all of the input side issues out.

On the unit, can you pull up the screen that shows the voltage waveform and have the working unit do that side by side to see is the faulty unit is displaying any distortions ? If the A-D input/signal has issues it may be showing up on the waveform readings ?

Click to expand...

Checking the DC bus voltage is a good idea. I had not got that far yet, still trying to get the MPPT to properly load the input DC supply. As for the output waveform, none of my 3 GTILs has ever had a voltage display that I really believed. They all show an imperfect sinusoid, from day 1.

suntan901 said:

completelycharged said:

What I was noting was the 20-200 random vs less variablity on the bench.

The DC HV boost arrangement is a standard approach with high frequency inverters. If the DC bus voltage is stable then that cuts all of the input side issues out.

On the unit, can you pull up the screen that shows the voltage waveform and have the working unit do that side by side to see is the faulty unit is displaying any distortions ? If the A-D input/signal has issues it may be showing up on the waveform readings ?

Click to expand...

Checking the DC bus voltage is a good idea. I had not got that far yet, still trying to get the MPPT to properly load the input DC supply. As for the output waveform, none of my 3 GTILs has ever had a voltage display that I really believed. They all show an imperfect sinusoid, from day 1.

Click to expand...

I replaced the 10 input caps, 1000uF @ 100 VDC, and the result is the same. 75 +/- 5 watts. Wish I had a dc controller board for this version of Main Board so I could substitute.

Well, it turns out that the input capacitors WERE bad, after all, but the dc controller board was also. So, this one is FIXED, after replacing both. Thanks to all for the advice. I never would have suspected the caps, as they have only about 2500 hours operating time.