1. Why the low pass filter , why delay shutdown ?

I don't know. A lot of that circuit is mysterious to me.

 

2. Why Q60 and Q61 on the IGBT  drives , but not on the MOSFET drives?

I don't know.

 

3. why is the Q60/61 clamp important.?

I've often wondered. Those MOSFETs seem pretty small, too.

 

4. why do some boards fits C15 and R108

I don't know, but it may be to suit the brand of SG3525 (see below).

 

5. why was this filter compatible with the SG3525  but not my new SG3525 one?

I've had conflicting reports on the AEVA repair and hardware upgrades topic on whether you can substitute brands of SG3525 or not. My guess is that a few of the brands are different enough that you need to populate different parts for reliable operation.

For example, see this post.

Did you replace with the same part number? I agree that an N suffix usually just denotes a plastic through-hole package. Was it a J suffix that back in the day denoted a ceramic package?

It seems like these chips are a pest for repair work.

 

Did you replace with the same part number?

yes  , new part is SG3525AN  , old part SG3525 .

Thanks , I had a look at your links. I  get the feeling that when the dsp inadvertently shuts down , it has to do with the sag in -12V due to the short caused by Q60/61 when the PWM is still providing output . The C15/R108 removal solves that problem. Note that  some main boards have that components removed as well. So I think its pretty much an ad hoc solution depending on the variation of 3525 types.

 

@Coulomb , @JustinSchoeman while on the subject of 3525 PWM based drivers , I have noticed that Voltronics has upgraded to a brand new design for Axpert MKS4 . Could be MKS3 as well but I cannot verify that.

So the new drive for the battery MOSFETS and the DC-DC IGBTS are based on an isolated based Texas Instruments chip called the  UCC21540 . See  https://www.ti.com/product/UCC21540 .  So in essence this is dual isolated  gate driver .  Because this is functionally only a driver  and the 3525 PWM controller does not exist on this board , I assume the DSP now generates the PWM phase outputs to the drivers. I have not traced it  but surely its under DSP control now . Maybe Coulomb can verify that in his firmware resources , if its possible to do so. Because it is isolated , the MKS4 board does away with the 4  pulse transformers altogether.

below is an image of the MKS4 pcb board showing the drivers :

2 chips , giving 4 isolator drivers for the battery full bridge mosfets:

and 2 chips , giving 4 isolator drivers for the dc-dc converter IGBTS's

 

Pretty cool I think.

Although tempted to trace this circuit , I do not have time , and will wait when I need to repair one before attempting to trace.

 

EDIT : In general the MKS4 board is very similar to MKS2 , with the dc-dc  drivers being the fundamental change. They also introduce TVS diodes for surge protection as you can see in the images above.

But the rest is very similar , the dc-dc drivers based on T350 remains stock standard.  There are of course subtle differenes , because the MKS4 can deliver 600 more watts(5.6kW) the caps are bigger and mosfets & igbts are rated higher.

Edited September 24, 20232 yr by BritishRacingGreen

 

@Coulomb , @JustinSchoeman while on the subject of 3525 PWM based drivers , I have noticed that Voltronics has upgraded to a brand new design for Axpert MKS4 . Could be MKS3 as well but I cannot verify that.

So the new drive for the battery MOSFETS and the DC-DC IGBTS are based on an isolated based Texas Instruments chip called the  UCC21540 . See  https://www.ti.com/product/UCC21540 .  So in essence this is dual isolated  gate driver .  Because this is functionally only a driver  and the 3525 PWM controller does not exist on this board , I assume the DSP now generates the PWM phase outputs to the drivers. I have not traced it  but surely its under DSP control now . Maybe Coulomb can verify that in his firmware resources , if its possible to do so. Because it is isolated , the MKS4 board does away with the 4  pulse transformers altogether.

below is an image of the MKS4 pcb board showing the drivers :

2 chips , giving 4 isolator drivers for the battery full bridge mosfets:

and 2 chips , giving 4 isolator drivers for the dc-dc converter IGBTS's

 

Pretty cool I think.

Although tempted to trace this circuit , I do not have time , and will wait when I need to repair one before attempting to trace.

 

EDIT : In general the MKS4 board is very similar to MKS2 , with the dc-dc  drivers being the fundamental change. They also introduce TVS diodes for surge protection as you can see in the images above.

But the rest is very similar , the dc-dc drivers based on T350 remains stock standard.  

This design also primes me to wonder now that the the dc-dc converter is under fine granular PWM control from the DSP , is it not possible to do away with the additional buck circuit altogether ? . So the DSP can now inherently buck the  battery output voltage , by chopping the IGBTS accordingly in line mode (charging) .  As I am not a magnetics guru , I wonder if the high frequency transformer can handle the narrower pulses.

Also with the 3525 based design , the IGBT's are still switched on and off when the inverter is in battery mode. Logic says that because the IGBT is a uni-directional  BJT , it servers no purpose to switch them in battery mode direction , and it can only introduce greater probability to fail when in battery mode. So one benefit of the new design is the DSP can keep the IGBT's off , while pulsing the MOSFET's in battery mode.

Edited September 24, 20232 yr by BritishRacingGreen

 

T350 ic output is incomplete too. Please check again. Thanks. (Maxo's 10g revision schematic is fine)

Thanks, I will check it out. 

inverter  come in with output igbts  blown   very  severly due to   customer  fed  mains  ac into  the output

i   replaced all  drive  componnets  that tests bad

after  that  the inverter  come on with error   06   i  rescanned the  area  and found  bad  zener  that  relflects   a  short  in one  copler tl350   with the zener  and  small opened  resistor  in serious with  zener   replaced

 

the inverter works and produce nice 230ac

 

now    can  tlp  250    replace  t350   it  actually  did  but  will it last??

second   question the  zener  in circled in red near the smps   is replaced  with 15vdc  zener  randomly  /....what is the  value of the   zener   i am afraid that  it is not  the right value  ? and igbt  will work stressed  since  it is   seems to be  across  the  vcc and ground of the copler  

 

 

by the  way i  even  loaded  the   3k inverter   with 100watt lamp  all igbt are not heat sinked it it work fine  igbt only   warm 
 

another  mystery  with  the  igbt  removed   i  turn  the  inverter  on i get  no  pulses  between  gate   emitter  at alll     ,,with the  igbts  installed   i  get 220  which   confirm the  presence  of the  pulses

 

can  i  calll  this  phenomenon   "phantom ..pulses""

 

reaally need  help  i did  repair ir  but  the  joy is  in understanding

Edited September 28, 20232 yr by wael_fathe

 

inverter  come in with output igbts  blown   very  severly due to   customer  fed  mains  ac into  the output

i   replaced all  drive  componnets  that tests bad

after  that  the inverter  come on with error   06   i  rescanned the  area  and found  bad  zener  that  relflects   a  short  in one  copler tl350   with the zener  and  small opened  resistor  in serious with  zener   replaced

 

the inverter works and produce nice 230ac

 

now    can  tlp  250    replace  t350   it  actually  did  but  will it last??

second   question the  zener  in circled in red near the smps   is replaced  with 15vdc  zener  randomly  /....what is the  value of the   zener   i am afraid that  it is not  the right value  ? and igbt  will work stressed  since  it is   seems to be  across  the  vcc and ground of the copler  

 

 

by the  way i  even  loaded  the   3k inverter   with 100watt lamp  all igbt are not heat sinked it it work fine  igbt only   warm 
 

another  mystery  with  the  igbt  removed   i  turn  the  inverter  on i get  no  pulses  between  gate   emitter  at alll     ,,with the  igbts  installed   i  get 220  which   confirm the  presence  of the  pulses

 

can  i  calll  this  phenomenon   "phantom ..pulses""

 

reaally need  help  i did  repair ir  but  the  joy is  in understanding

Unfortunately I cannot comment on the zener as its a 3kw machine, if you can maybe give us a closeup view around it so to identify the circuit. Can you relate this zener to an equivalent device onn my schematic in a couple of posts earlier, e.g. ZD7? 

As far as TLP250 is concerned it appears to be an older discontinued version of 350,but it may be good. The max frequency is 25khz as opposed to 50khz, which resolves to 40us vs 20us pulse width. Now I dont know what the max resolution is for the DSP creating the PWM for sine wave. Maybe @Coulomb knows. You can check the ac output on your scope and check if you have distortion.  Just make sure that your ground is not reference to neutral or earth  and use 10x probe. 

Also if you do have scope and you can test on IGBT gate and emitter, verify that your voltage swings are within specs, eg - 5.4v during pulse switch off, and 12-15v during pulse switch on period. 

 

Well done getting the machine running again, old chap 👌

Edited September 29, 20232 yr by BritishRacingGreen

 

another  mystery  with  the  igbt  removed   i  turn  the  inverter  on i get  no  pulses  between  gate   emitter  at alll     ,,with the  igbts  installed   i  get 220  which   confirm the  presence  of the  pulses

Remember if the DSP starts the fullbridge, but fail to get the right readings on the ac output, it will after a short while give up on it, but should raise an Output Short Circuit error. 

To capture the initial pulses you should start measuring at switchon, and set the scope to an appropriate trigger level to capture the gate pulses. 

Edited September 29, 20232 yr by BritishRacingGreen

 

Remember if the DSP starts the fullbridge, but fail to get the right readings on the ac output, it will after a short while give up on it, but should raise an Output Short Circuit error. 

To capture the initial pulses you should start measuring at switchon, and set the scope to an appropriate trigger level to capture the gate pulses. 

Just for my own interest, it appears there are only 4 igbt drivers, so I assume no Buck IGBT, which infers there must be an seperate AC charging section. Is this correct? 

 

Just for my own interest, it appears there are only 4 igbt drivers, so I assume no Buck IGBT, which infers there must be an seperate AC charging section. Is this correct? 

yes it has  a  seperate  ac  charger  .....i scoped between  emitter and gate ,  no wavform  at all ,

i did  it at the  power on moment and  get nothing ,  with all igbts  installed the device work and  provides 220ac

those unit have  mystery   ?

------------

u said

Also if you do have scope and you can test on IGBT gate and emitter, verify that your voltage swings are within specs, eg - 5.4v during pulse switch off, and 12-15v during pulse switch on period. 

 

------------------------

this unit do not  turn its   igbts  to -5.6  this discussion happen with  colomb at the other  forum

user  confirm that  near by the  igbts   there are  only 2 zeners one  both 16vdc  

 

 

so switch on   is  probably 15 volt  switch off 0vdc  

 

 

Edited September 29, 20232 yr by wael_fathe

 

The 1200 V rating is way more than you need (which is about 100 V).

you misunderstand me   i am  talking  about the h-bridge output igbts  i have   100  pces  40t60  they have no  diode

can i install  external one if i find   a  good way to  keep it cool?

 

The 1200 V rating is way more than you need (which is about 100 V).

 

u miss understood me  i am talking about the hbridge  igbts   i got 100  pces they have no  internal diodes

i ordered more 200  pces of  1200  30amp   diode  can i fit them  acorss   emiiter  collector of an igbt rated 40t60 

 

and I temp fixed the board by replacing the dc-dc IGBT's with 1N4007 (1000V / 1A)

so is that ever  possible  to remove  the  synchronised    igbts from dc to dc and install  general purpopse   diodes

i  know it is not right  for  normal  operation,  but can that work  for  testing  purposed  ?

 

so is that ever  possible  to remove  the  synchronised    igbts from dc to dc and install  general purpopse   diodes

Well thats what I have done as I stated, so yes it is of course possible. But its only appliccable in battery mode where flow is from battery to the high side dc bus. In this case the IGBT bridge performs the function of a full bridge rectfier via its body diodes. Remember unlike a mosfet, the IGBT is a BJT, so current cannot flow from emitter to collector other than via its non-parasitic  'body' diode, even if you bias the device to its on state. 

But dont use the 1n4007 unless you have precharge the bus capacitors gracefully. Use higher current power diodes instead, but still please precharge the caps or make sure you slowly ramped the battery voltage manually.

 

 

i  know it is not right  for  normal  operation,  but can that work  for  testing  purposed  ?

Yes

Edited September 30, 20232 yr by BritishRacingGreen

 

u miss understood me  i am talking about the hbridge  igbts   i got 100  pces they have no  internal diodes

i ordered more 200  pces of  1200  30amp   diode  can i fit them  acorss   emiiter  collector of an igbt rated 40t60 

This is a question for @Coulomb according to previous conversation.

Make  sure that the rating of the diode is at full bus current spec, because they not only there to protect IGBT reverse voltage breakdown as intended, they actually perform full load switching when in battery mode. 

Edited September 30, 20232 yr by BritishRacingGreen

Hi @Coulomb , I have bread-boarded a differential instrumentation amplifier  , as shown below , to read main board battery voltage from my Pi PICO  .  I want to point out that I can now confirm that when the -12V rail supply goes open circuit , or shorted to ground , then the op-amp outputs the maximum positive voltage that the it can drive , in my case it is  about .5 in access of 3v3 clamped by bleeding diodes . This is true for LF356 and a 741 and a TL074 as per DSP . So the CPU shows its maximum  reading and hence we see high values on DSP display , e.g. 120V for battery.  If the +12V rail collapses , the output is 0V.

So as we know by now this is typically due to the input filter capacitors  of the 7912 -12V regulator going dry.

 

Edited September 30, 20232 yr by BritishRacingGreen

 

Hi @Coulomb , I have bread-boarded a differential instrumentation amplifier  , as shown below , to read main board battery voltage from my Pi PICO  .  I want to point out that I can now confirm that when the -12V rail supply goes open circuit , or shorted to ground , then the op-amp outputs the maximum positive voltage that the it can drive , in my case it is  about .5 in access of 3v3 clamped by bleeding diodes . This is true for LF356 and a 741 and a TL074 as per DSP . So the CPU shows its maximum  reading and hence we see high values on DSP display , e.g. 120V for battery.  If the +12V rail collapses , the output is 0V.

So as we know by now this is typically due to the input filter capacitors  of the 7912 -12V regulator going dry.

 

 

I also found this schematic on your repair thread. The resistors R130 and R123 are 100k, so gain is 100k/(4 x 1M)., multiplied by 120V gives an output of 3V, which explains why the 5kW inverter shows  120V batt voltage when - 12V fails. 

 

This is a question for @Coulomb according to previous conversation.

Make  sure that the rating of the diode is at full bus current spec, because they not only there to protect IGBT reverse voltage breakdown as intended, they actually perform full load switching when in battery mode. 

then i  suppose  that  they will need  heat -sinking too  ?

 

This is a question for @Coulomb according to previous conversation.

Make  sure that the rating of the diode is at full bus current spec, because they not only there to protect IGBT reverse voltage breakdown as intended, they actually perform full load switching when in battery mode. 

then i  suppose  that  they will need  heat -sinking too  ?

 

you misunderstand me   i am  talking  about the h-bridge output igbts  i have   100  pces  40t60  they have no  diode. 

Duh! I don't know what I was thinking. Sigh. MOSFETs probably don't even come without diodes, it seems to be inherent. 

I have no expertise in this, but my guess is that it would be possible. I suspect that the diodes would need heatsinking, though they might not get very hot with unity power factor loads. I'd say when powering say a motor, they might conduct for about 20% of the AC sine wave. 

Edit: The above is for the main DC-AC converter. For the high end of the DC-DC converter, it may be that the diodes take most of the current, though I think that the saturated transistors should take the majority of it. 

Edited October 2, 20232 yr by Coulomb

 

Duh! I don't know what I was thinking. Sigh. MOSFETs probably don't even come without diodes, it seems to be inherent. 

I have no expertise in this, but my guess is that it would be possible. I suspect that the diodes would need heatsinking, though they might not get very hot with unity power factor loads. I'd say when powering say a motor, they might conduct for about 20% of the AC sine wave. 

Edit: The above is for the main DC-AC converter. For the high end of the DC-DC converter, it may be that the diodes take most of the current, though I think that the saturated transistors should take the majority of it. 

that  nice  and   promising asnwer i  ordered   1200vdc 30a  super  fast   diode  from china  i wil try  use them  in  my  next repair project  and  willl report back the results in this  forum  for the benfit  of  all of  us    whho  knows maybe  some  else get cheated  and  recive  igbts  with out  diodes   ....i bought the  part with the  d  at the  end   as indication of  diode  

but they only send  the part with no d at the  end  of part number

which  means  no diode installed   ,,

 

but they only send  the part with no d at the  end  of part number

It's possibly an easy mistake to make on their part. That tiny little letter "D" at the end... so close! But as you say it's a huge problem for you 😡!

Hopefully some good comes from this, as you say. Sorry to hear about your predicament.

Edited October 3, 20232 yr by Coulomb

On 2023/09/30 at 9:41 PM, BritishRacingGreen said:

as shown below , to read main board battery voltage from my Pi PICO 

sir you read  to  your pi pico   not  from  pi PICO    i think its  typo?

quite  a legendary work  proving all this for the  rest of us in  a  nutshell

measuremnt  problems  are  either 

resistors open

op amps haywire

voltages  unstable and  especially  -12vdc   as  it seems the  12+  willl  cause more  serious problems

if it  falls  segnifcantly  or  totally  collapse

 

nice  work  sir  

Edited October 6, 20232 yr by wael_fathe

7 hours ago, wael_fathe said:

sir you read  to  your pi pico   not  from  pi PICO    i think its  typo?

Thanks @wael_fathe, your inputs are always welcome on this thread. Yes, English not my first language , but I think you get the drift : the opamp output is connected to an ADC channel input of the Pico. 

7 hours ago, wael_fathe said:

voltages  unstable and  especially  -12vdc   as  it seems the  12+  willl  cause more  serious problems

Yes, the loss of - 12V  results in a maximum output of the  opamp regardless of any sense voltages from the resistor strings. That causes very high values being displayed on the pico or DSP controller. 

The loss of +12V will result in O volt output of the opamp. 

In addition, the +12v and - 12v are also used to generate the PWM mosfet and igbt gate drives of the DC-DC converter. The loss of +12V will result in the SG3525A chip not being  powered at all , so there will be no gate drives.  So I dont think the loss of +12v is catastrophical or serious. 

 

Edited October 7, 20232 yr by BritishRacingGreen

On 2023/08/01 at 6:13 PM, BritishRacingGreen said:

a Short Guide to Bring Up a 5kW Axpert Main Board in the Softest Possible Manner : Chapter 1 

The reasonable amount of success I enjoyed repairing the MAX range of inverter is because of using a methodology to power the system supplies of the board , and debug the critical sections like mosfet and igbt drivers , without powering the actual power chain . This means zero voltage on the battery bus , zero voltage on the high voltage DC bus , no grid , no pv. The max has actually an external system supply connector which makes this feasible. The 5kW boards does not have this option , but we have devised a non-intrusive way of  injecting raw dc for the system supply SMPS.

The reason why this bare minimum soft method of powering the system is important , is it allows you to debug all the Mosfet and IGBT drivers  without the risk of blowing those very  expensive Mosfets and IGBTs . You can now test the circuit in a very controlled and safe environment.

So the idea is that you remove the main board and remove ALL external interfaces . That includes the DSP controller board, the MPPT board , the fans , the comm modules . In the end you are left with just the bare main board.

Prerequisite 

The only initial condition required , is that the battery bus , the high voltage dc bus and the ac output bus have no short circuits . These conditions you can verify by checking with multimeter. if there are shorts you need to check the mosfets and igbts for failures by isolating them. You dont have to substitute them yet with new transistor , but you need to remove faulty ones. Refer to Voltronics service manual which will guide you thru resistive tests in order to verify the typical resistances required.  We are currently only interested that the gate to source are not shorted , because that will prevent us to test the gate driver signals later on the exercise. In the case of IGBTS this is gate to emitter.  

Also read the resistance of 12V,-12V and 5V rails to GND. with the 12v and 5v the capicitors will see to it that the resistance gradually increase . The 5V will easily reach >10k , the 12V >50k  , but the -12V will be reasonable steady at at about 450 ohms.

Once your board has no bus related short circuits , we are good to go in order to bring up  the main supply.

 

Powering the up the Main SMPS

So we going to inject external controlled power to the input of the SMPS . You will need a current limited 30V dc power supply , but you will need a minimum of 30V. The voltage may be as high as 55VDC , but we will stay with 30V for safety reasons. Limit the maximum current of the power supply to 200mA .  We going to draw about 100mA if the board powers up 'nicely'.

The SMPS is normally supplied by the battery input  . However we don't want to apply power to the battery terminals, so we inject our 30V supply as follows. Notice below that there is a 2-pin SCC connector . Next to it is a 2-pin AC-START connector . You must connect your 30V positive terminal on pin 2 of SCC connector. Doing so allow us to inject voltage on the SMPS input feed , but D53 will block the voltage reaching the battery positive , or CH+ as depicted in the schematic.

 

Now we are left with connecting the negative of our 30V supply to the negative CH- rail. Note that you cannot connect it onto the battery negative terminal , because  some boards have reverse polarity mosfets that will not be biased , so we need to connect to the CH- rail . A convenient way of doing so is to locate to upright metal spacers , which is labelled CHG+ and CHG- (or CH+/CH- as per schematic) . I have tied my power supply negative conveniently  to this metal spacer.

Once you have connected and powered on your power supply , you will notice that no current is drawn. That is because you must start the SMPS manually. This is what the AC-START 2-pn connector is there for. You can either connect the inverter on-off switch here , or you can merely temporarily short out the 2 terminal for 2 seconds or so. If all goes well you will notice the power supply is delivering about 100mA of current to the SMPS .

We can now check the 5V , 12V and -12V rail voltages . The 5V should be very close to 5.0V , 100mV either way is ok. 12V should be from 11.85V upwards but not exceeding 12.4V .  Same for -12V (-11.8 to -12.4)

I will post again as chapter 2 of this post .

.... to be continued ....

 

 

 

sir  your method is great but i think this method of colomb is easier 

1  rmeove all fets  and dc-dc igbt(though for igbts  its ok  to leave them if they not  shorted)
2  short u9 colpler to enable sg3525
3 remove cpu board
4  connect the inverter to dc power supply 48vdc 24vdc etc
5 hit the ac on switch on
6 scope for signals at mosfet and dc dc igbts

but i think your method is  a must  if

1  cpu is built in 
2  you have not removed the mosfets yet 

thogh shorted fet  or dc-dc igbt    will short its drive it seems  good to remove all fets  if we 
wanted to scope   for good signal   

am i right or there is  more hidden  benifits of your   method?

Edited October 10, 20232 yr by wael_fathe