17 minutes ago, BritishRacingGreen said:

Thank you , I revised the diagram above (ver 1xB).

As far as the tx ratio is concerned , my own measurements were consistent with a ratio of very near to 1:8 . However I need to qualify that I have not covered the whole dynamic range of voltages , I stayed below 120V on the BUS side. Also my converter was never under any form of load, so dc measurement always reflect the peak voltages on the filter capacitors.

Coulomb , the schematic snippet below refers :

 

It appears that the circuit built around Q52 and Q53 reshapes (rebuilds) the transformer output signal. D37/C112 and D31/C113  establishes a new positive and negative rail supply , and Q52/Q53 produces a new signal from the rail feeds. Why would this be important for the MOSFET gates , whereas the IGBT gate drives have a very similar drive circuit , but without this Q52/Q53 arrangement.?

Below is a snippet of IGBT drive circuit :

 

 

 

 

Edited September 3, 20232 yr by BritishRacingGreen

Remember that currents on the IGBT side are 1/8 of those on the MOSFET side.  So expect similarly lower gate charge (and probably a slightly more relaxed switching time requirement).

Apparently (for the devices in use) this is enough of a difference to allow for direct drive of the IGBT, but require a totem pole drive for the MOSFET.

7 hours ago, BritishRacingGreen said:

As far as the tx ratio is concerned , my own measurements were consistent with a ratio of very near to 1:8 .

Actually, just checking an Axpert Max manual, I find that the maximum CV/Absorb/bulk voltage is 62.0 V, and a VM III manual showed 61.0 V. So my memory is bad (I thought that all the new models went to 64.0 V), and it's possible that the transformer ratios have changed. Indeed, I recall seeing that the ratio is 7.0 for the 145 V max PV Axpert King, but checking the same code just now in an Axpert Max firmware, I see that the ratio is 7.3. So they're apparently not restricted to integer ratios. Though this part of the firmware may have reason to fudge the ratio a little.

Well, thanks for setting me straight.

Edit: Though it seems that all VM IIIs, even the latest models, are 1:7.0 .

Edited September 3, 20232 yr by Coulomb

7 hours ago, BritishRacingGreen said:

It appears that the circuit built around Q52 and Q53 reshapes (rebuilds) the transformer output signal.

Mmm. Maybe.

2 hours ago, JustinSchoeman said:

Apparently (for the devices in use) this is enough of a difference to allow for direct drive of the IGBT, but require a totem pole drive for the MOSFET.

I've never really understood gate driver theory. I think that the totem pole is just for driving all those fat gates in parallel. Maybe the extra resistor and capacitor are just to decouple the prodigious gate current spikes from the rest of the circuit, that would otherwise transmit back through the transformers like TX8. I don't see it as really reshaping the transformer output, but like I say I've never understood the theory.

2 hours ago, Coulomb said:

I think that the totem pole is just for driving all those fat gates in parallel

@Coulomb and @JustinSchoeman , thank you for your inputs . It has prompted me to investigate a little further .  Yes in the case of the battery side , there are 4 MOSFETS in parallel , and for sure the  net input capacitance of the gates are now 4-fold.  I now believe the totem pole circuit is a re-shaping device , with its major asset being the 1uF storage capacitor in each of  its supply lines.

 

My assumptions are based on the fact that the FET gate input of the mosfet is in essence a high impedance and therefore  a voltage controlled input. However it also has a relatively high input capacitance , which does bring about current spikes when the gate is switched  .  

It is now my theory that the high output impedance of the transformer output is not sufficiently low enough to switch the gate voltage fast enough. So the wave is reshaped , but the 'supply' rails are now augmented with two storage capacitors , one on positive , and one on negative.

When the caps are discharged and the circuit is switched on , the first gate pulses will be poor , as the transformer will have to supply both the gate drive signal and charge the capacitor. But once the short switching time has lapsed , the transformer supply has 'ample time' to charge the cap. Now when the next pulse arrive , we have both the transformer output as well as the charge on the cap in order to combat the gate input capacitance. Again after the switching time , the cap has again ample time to recharge , at the cost of the gate drive having a slightly lower voltage , but still well above the gate threshold for saturation. 

I think I have proven my theory after taking an oscillogram of the gate signal versus the capacitor voltage.

The green trace is that of the voltage on the capacitor on the positive rail . Clearly you can see the discharge during gate switching period, and the subsequent replenishing of charge for the balance of the pulse duration.

The same of course will hold true for the negative rail.

This is a shot in the dark , so I welcome your thoughts.

 

 

Edited September 3, 20232 yr by BritishRacingGreen

That is pretty much it.  There is enough energy in the entire cycle to rapidly charge the gate capacitor. But not enough instantaneous power.

So, D37, R47 and C112 form a filtered power supply. Q52 is a class B amplifier (current amplifier) running from this power supply to supplement the transformer power during the switch on phase. The other components are a mirror for the switch off phase.

The lower currents on the HV side mean that you have to move much less charge to turn the devices on and off, and you can get away with driving directly from the transformer.

The diode and capacitor are the two half-wave power supplies, positive and negative with respect to the MOSFET source. So the only slight mystery is the purpose of the resistors. I now think that they are there just so that the transformer output doesn't get too loaded just charging the capacitors, as well as the gate capacitance. So immediately after switching, the capacitors supply most of the current to charge the gates, and the resistor makes sure that the bases of the bipolar transistors get a high enough voltage to turn on hard. If the resistor wasn't there, the transformer would be loaded charging the capacitor.

I suppose that does amount to reshaping the transformer output signal, so we are actually in agreement.

21 hours ago, JustinSchoeman said:

So, D37, R47 and C112 form a filtered power supply. Q52 is a class B amplifier (current amplifier) running from this power supply to supplement the transformer power during the switch on phase. The other components are a mirror for the switch off phase.

Yes my term 're-shaping' is probably stretching it (no pun!) , I think a better term will be a buffer amplifier , although its very application-specific.

21 hours ago, JustinSchoeman said:

The lower currents on the HV side mean that you have to move much less charge to turn the devices on and off, and you can get away with driving directly from the transformer.

Yes , there is only one IGBT per drive , so that does make the difference.

 

19 hours ago, Coulomb said:

The diode and capacitor are the two half-wave power supplies, positive and negative with respect to the MOSFET source. So the only slight mystery is the purpose of the resistors. I now think that they are there just so that the transformer output doesn't get too loaded just charging the capacitors, as well as the gate capacitance. So immediately after switching, the capacitors supply most of the current to charge the gates, and the resistor makes sure that the bases of the bipolar transistors get a high enough voltage to turn on hard. If the resistor wasn't there, the transformer would be loaded charging the capacitor.

I suppose that does amount to reshaping the transformer output signal, so we are actually in agreement.

Yes , agree , the resistor initially confused me , but I also think its to do with not loading the transformer too much.

Thanks @Coulomb / @JustinSchoeman this exercise for me was super important  in order to document a procedure for identifying acceptable gate drive patterns during debugging. I am going to be honest and direct I did not know IGBT /MOSFETS from a bar of soap really when I started with solar 2 years ago , and what I have learned so far on this forum is awesome. And knowing there will be even more to learn as I progress on this 'journey.

I do have a new question or two regarding the gate drive on the igbts. My scope reading shows switching spikes that goes beyond 20V (max ratings  for these igbts/mosfets) .  as you can see there are protective zeners on the transformer output that is suppose to clamp to 18V + 0.6V = 19V in both positive and negative directions  , which I see as too close for comfort , given tolerances of these zeners. But that is for another day .  Cheers

 

Edited September 4, 20232 yr by BritishRacingGreen

4 hours ago, BritishRacingGreen said:

My scope reading shows switching spikes that goes beyond 20V (max ratings  for these igbts/mosfets) .

If this is with the IGBTs in place, then it's very hard to decide whether a spike on the scope is real or not.

If it's without the IGBTs in place, it's still challenging, with fast changing signals everywhere. I find I get the least worst results when using the spring earth contacts to get a really good local earth for the scope probe. Something like this:

The spring part attaches to the scope probe earth area, and you have to carefully manoeuvre the point of the spring to a good local earth / reference point. It can be a real pain, and of course, so few times will you be able to earth the emitter of the IGBT for the measurement. There is a bit of skill to this, almost an art. I'm no artisan here.

Ah. I remember the 'good old days'...  You generally aim for a 10ns rise/fall time on the gate. So this signal has strong frequency components in the 100MHz - 1GHz range, and at these frequencies every track, wire and lead is a significant inductor and capacitor. So you end up with dozens of resonant frequencies which you need to damp, and PCB design becomes critical.

Then you add a 10pF scope probe and add a whole new set of potentially resonances. But how do you tell the difference between 'real' transients and those introduced by the scope probe itself?

I remember this actually being one of the hardest parts when designing a switching regulator.  I have never had to debug someone else's design - but if I had to, I would just check component values and connections. If everything checks out I would assume that the actual drive waveform is correct and I am just seeing measurement induced transients.

Does anyone have a schematic of the MPPT board shown in the pic . It is from a Kodak OGS 5.6 which I think is an Axpert , OEM by voltronics ? 

and here is the other end  of the board 

On 2023/09/05 at 1:26 AM, Coulomb said:

If this is with the IGBTs in place, then it's very hard to decide whether a spike on the scope is real or not.

If it's without the IGBTs in place, it's still challenging, with fast changing signals everywhere. I find I get the least worst results when using the spring earth contacts to get a really good local earth for the scope probe. Something like this:

The spring part attaches to the scope probe earth area, and you have to carefully manoeuvre the point of the spring to a good local earth / reference point. It can be a real pain, and of course, so few times will you be able to earth the emitter of the IGBT for the measurement. There is a bit of skill to this, almost an art. I'm no artisan here.

I have drilled down somewhat on this issue .  the 'spike' appears as a short spike , but I  stretched out the signal wider in time ,  the images below refers:

 

The Y scale is actually 10V per division

It is  not apparent in the image, but the bottom of the spike is actually clipped by the zeners. I have removed one zener , and the spike goes below -20V.

Note that this is with an IGBT fitted , far more damped . If I remove the igbt , the spike oscillation can clearly be seen, as below: please note this is without zener clamping:

 

and here is the same wave , but zener clamping restored. you can see the shaving at the bottom.

 

 

This to me is good news because one can observe the zener clamping on the scope without actually have to remove one to prove its ok.

So although the pulse reaches very close or on the +-20V absolute maximum rating of the igbt , the circuit under repair is correct , and I will abide by that. I have performed the same test on two different 5kW machines and they both expose the same waveforms.

 

EDIT : It is noteworthy to see that when the igbt is fitted , its capacitive load increases the switch off period from 48nS to 960ns.

EDIT2: This data is of course also useful  if we want to check whether  an  alternative / replacement IGBT is suitable in the machine.   

 

Edited September 6, 20232 yr by BritishRacingGreen

9 hours ago, Hedley said:

Does anyone have a schematic of the MPPT board shown in the pic . It is from a Kodak OGS 5.6 which I think is an Axpert , OEM by voltronics ? 

Infortunately no schematic or even partial schematic is available.

9 hours ago, Hedley said:

and here is the other end  of the board 

yes it is rebranded Axpert MKSIV .

7 hours ago, BritishRacingGreen said:

I have drilled down somewhat on this issue.

Excellent work, thanks for that. I'm intrigued by the asymmetry: the positive going edge doesn't seem to want to oscillate / overshoot nearly as much. Maybe there is something on the driver side causing that. Maybe it's as simple as better bypassing of the +12 V rail than the -12 V rail. And/or the +12 V rail being far stiffer, having to generate far more current than the -12 V rail.

Any chance of verifying that hypothesis? Looking for noise on these rails?

Edited September 7, 20232 yr by Coulomb
Added last sentence.

6 hours ago, Coulomb said:

I'm intrigued by the asymmetry: the positive going edge doesn't seem to want to oscillate / overshoot nearly as much. Maybe there is something on the driver side causing that.

D32 bypasses the gate resistor for discharge (turn off) - so turn off is faster and less damped. You need to manage dI/dt during turn on, which is why you have R91. Don't have the same problem during turn-off, so bypass R91 with D32 for turn off.

23 hours ago, BritishRacingGreen said:

yes it is rebranded Axpert MKSIV .

few posts  back  to refered to the fact that you can actually bring up the  dc to dc with 3vdc  on battery  ...no objection 

since its  fixed  ration    then it will  just  spit  what  ever  input  it gets  to  that  input x8  

 

but  i wonder  how you will  give the  sg3525 its  12  and  -12?

and   this new test  i  would imagine that it requires    cpu board to be out  ?

 

 

On 2023/08/23 at 6:58 PM, Coulomb said:

Yes, pretty much.

With utility charging, there are actually three full bridges (H arrangements of effectively 4 switching devices). Each of these is capable of bidirectional power flow. However, two of them (either side of the transformer) are not phase controlled; they are "dumb" bridges that adjust power flow so as to maintain the same voltage ratio. All the switches (transistors) in the DC-DC converter are driven with a fixed waveform: almost a square wave, with a small gap where neither the upper or lower switches (transistors) are turned on.

So by adjusting the phase between the inverter output and the mains, power flows either mains to DC bus or the other way around. To charge the battery, the firmware arranges for power to flow from the mains to the DC bus. The higher DC bus voltage causes power to flow through the other two H bridges into the battery.

Maybe you're ready now for the final piece of the design: the buck converter. What happens if the mains voltage is high and the battery voltage is low? It will charge the battery, yes, but way too fast! The inverter can easily push the 48 V battery up to well over 60 V, which is no good for the battery. So to match the voltages without wasting a lot of power, there is the buck converter. This allows for some tens of volts to exist across the buck converter without wasting much power, and so by varying the pulse width of the buck converter's switch/transistor, you get to regulate battery charging current.

The firmware has a lot of work to do. It has to get all these pulse widths just right, making sure that the bus voltage doesn't get out of hand, that the AC output voltage, frequency, and phase are just right, that the maximum total and utility charge current limits are not exceeded, monitoring temperatures and controlling fans, and much more. Then there is all the co-ordination of paralleled and/or 3-phase machines for some models. And handling over a hundred different commands, and updating the display (though in many models that's a separate processor now).

we  never been in situation  of  some thing that giving and taking at the same time in electrobnics

that why it so hard to  grasp

 

we  acustomer to  input that  taking 

or out that giving  thanks  for  explanations

 

 

but i am embarking on the task  of  creating   forum  brief  https://forums.aeva.asn.a

 

 ...some thing like  guide lines

because i read  and  forget  all you say 

and have to back to the  piece of info  that i  forget then i  find my self shocked

that i dont  know  where i read it

so  i  start page by page  and combed and filtered    keep  some  briefs  that will help me

construct  ideas  during trouble shooting time

 

after all  mr  colomb    its  easier to search a lake  than to  search an ocean

 i  combed  forum page  by page    and  eventually  reach  a   post   where you explained  

how cpu   shut down the main smps.....i wanted to documet the right information

to me  it is  accptable if the  cpu  shutdown the inverter  while the main smps works  by solar   or ac 

the cpu can shutdown  lcd and the dc to dc    but as  you explained that cpu can shutdown the main smps

by  u8 copler  but   the moment it did  so  that will cut  the power for  cpu itself  and  prevent it from  making  an orderly shutdown >???

 

or maybe output caps  have enough  power  to keep cpu work till  the inevitable   "black out"??

 

cant see  the  on off  switch on your schematics ?  if it is  really  exists  can you show me how  actually

cpu  senses  the off on  status  and make  a  judgemnt of    black  out or not to black out 

 

with solar power or ac present it can  be accptable  as  those are ored to the main sps  and  

they provide nice power  for main sps primary   till  the cput  sense  the switch  and  decided to

black it out

 

but with only  battery  power exist  and from what i know the swicth  brings  power to main sps

here  we have situation   that cpu uses u8 to shutdown  some thing that is  already being shutdown

if  cpu make it first  he  sleeps  nicely?

if  power  goes  first  the cpu may be  sleep  with  courrpted  firmware? due to harsh shutdown?

thanks 

 

Edited September 7, 20232 yr by wael_fathe

7 hours ago, wael_fathe said:

but i am embarking on the task  of  creating   forum  brief  https://forums.aeva.asn.au

That's the idea behind the index in the first post. But it's hard to know what to put into the index, and what just clutters it up.

7 hours ago, wael_fathe said:

but as  you explained that cpu can shutdown the main smps

by  u8 coupler  but   the moment it did  so  that will cut  the power for  cpu itself  and  prevent it from  making  an orderly shutdown >???

The CPU shuts down the power supply as the very last thing, after an orderly shutdown. There will be a hundred or so milliseconds before the 5 V power supply finally collapses.

7 hours ago, wael_fathe said:

cant see  the  on off  switch on your schematics ?  if it is  really  exists  can you show me how  actually

cpu  senses  the off on  status  and make  a  judgemnt of    black  out or not to black out 

The power switch is the one marked "AC start". Opto U13 communicates the state of the switch to the DSP (CPU).

Note that turning the switch on only has an effect on the power supply momentarily; as soon as capacitor C7 charges, there is no effect. But by then the power supply will have started, and the 15 V power supply will be supplying U10, the main power supply chip.

Note that in this model, the SCC (Solar Charge Controller) has a separate switch (actually an opto coupler) that allows  the SCC to start the main power supply even though the switch is off.

8 hours ago, wael_fathe said:

but with only  battery  power exist  and from what i know the swicth  brings  power to main sps

here  we have situation   that cpu uses u8 to shutdown  some thing that is  already being shutdown

if  cpu make it first  he  sleeps  nicely?

if  power  goes  first  the cpu may be  sleep  with  courrpted  firmware? due to harsh shutdown?

As above, once the power supply is on, it keeps itself on unless or until the DSP turns it off with U8. So there isn't a situation where there is a disorderly shutdown, unless the battery fuse blows or the like. Even then, the DSP likely has time to shut things down, as it can execute tens of millions of instructions per second.

 

On 2023/09/07 at 2:01 AM, Coulomb said:

Any chance of verifying that hypothesis? Looking for noise on these rails?

 

On 2023/09/07 at 8:29 AM, JustinSchoeman said:

D32 bypasses the gate resistor for discharge (turn off) - so turn off is faster and less damped. You need to manage dI/dt during turn on, which is why you have R91. Don't have the same problem during turn-off, so bypass R91 with D32 for turn off.

@Coulomb , @JustinSchoeman   :  ok  on Justin's suggestion that it could be caused by the gate discharge 0R resistor via steering diode , I took the liberty of removing the the 0R resistor , and ,yes, that's the cause of the 'spike' .

Below is pulse capture with the standard circuit :

 

 

And here is the capture of the 0R resistor removed, leaving the gate driver with only the main 22R resistor

 

 

Now its pretty symmetrical.

I have learned via google search that IGBT has typically a slower turn off time in relation to turn on , and in relation to mosfets. Probably the reason for the heavier gate discharge needed when turning off , and hence the steered 0R resistor.

 

EDIT : I have finally decided that if I want to maintain the repairing 'journey' , then a SMD rework station is a must . So this week I purchased one as an early Christmas present ! Its a generic Chinese model , but it just works for me .  It has a fan for blowing hot air  , instead of a piston pump as per expensive models , but still very impressed. I have removed the 0R resistor literally within seconds , and replaced it in seconds . Should have done this when I started repairing. But so we learn.

Edited September 8, 20232 yr by BritishRacingGreen

14 minutes ago, BritishRacingGreen said:

 

@Coulomb , @JustinSchoeman   :  ok  on Justin's suggestion that it could be caused by the gate discharge 0R resistor via steering diode , I took the liberty of removing the the 0R resistor , and ,yes, that's the cause of the 'spike' .

Below is pulse capture with the standard circuit :

 

 

And here is the capture of the 0R resistor removed, leaving the gate driver with only the main 22R resistor

 

 

Now its pretty symmetrical.

I have learned via google search that IGBT has typically a slower turn off time in relation to turn on , and in relation to mosfets. Probably the reason for the heavier gate discharge needed when turning off , and hence the steered 0R resistor.

 

EDIT : I have finally decided that if I want to maintain the repairing 'journey' , then a SMD rework station is a must . So this week I purchased one as an early Christmas present ! Its a generic Chinese model , but it just works for me .  It has a fan for blowing hot air  , instead of a piston pump as per expensive models , but still very impressed. I have removed the 0R resistor literally within seconds , and replaced it in seconds . Should have done this when I started repairing. But so we learn.

Thats impressive, repacing 0R resistor solved the issue. 

Edited September 8, 20232 yr by tanveerhabib

@Coulomb @BritishRacingGreenHello can i use STGW60V60DF instead of STGW60H65DFB in 5kva inverter? 

Edited September 8, 20232 yr by tanveerhabib

10 hours ago, tanveerhabib said:

Thats impressive, repacing 0R resistor solved the issue. 

Hi @tanveerhabib , thanks , but we have not provided a fix here per se, I have only proved why there is a a short period of damped oscillation on the negative going switching of the driver circuit. But the OR resistor has been restored , as its primary function is actually to improve the slew rate of the pulse when the driver is switched off.

10 hours ago, tanveerhabib said:

@Coulomb @BritishRacingGreenHello can i use STGW60V60DF instead of STGW60H65DFB in 5kva inverter? 

I have not yet compared the 2 devices yet per datasheet  , they appear to have similarities ,  one of the major differences is 600V vs 650V . The 600V is going to work , but whether that extra extra 50V is  important in the long term , I just don't know. Maybe @Coulomb can shed some light here.