23 minutes ago, BritishRacingGreen said:

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.

Ok, i learnt alot from you and @Coulomb. Thanks for your expert advices here 

The STGW60V60DF seems to have a typical input capacitance of 8 000 pF, and the driver chips seem to have a maximum rating of 10 000 pF, so it would appear that these devices are OK in that aspect. [ Edit: Some 80 A parts are known to have too much capacitance for the driver chips used. ]

My guess is that 600 V is OK, as I suspect I've seen 600 V parts used before in Axperts.

Other than that, I don't know; this is not something I know much about.

Edited September 9, 20232 yr by Coulomb

Before I kick off with part 2 of chapter 3 of the soft way of  bringing up the inverter , I would like to mention an important thing I have picked up during the week while repairing the battery side mosfet drivers and mosfet of a 5kW Axpert. This will be part of part 2 coming.

The battery side had a short circuit being a result of 3 of the 4 mosfet strings being short circuited. So I  isolated the shorted mosfet and removed them. Some strings had only one mosfet shorted , the other 3 mosfets in the string seemed ok.

So I only replaced the faulty ones with an exact type. Done all the basic tests , and finally the machine inverted in battery mode . Nice. But the driver circuit became rather hot after a while , including the -12V regulator on the main SMPS. So my first assessment was that of partially failed driver components. I isolated the 22R driver resistors on one string and that made a big difference in power consumed.  So this led to the fact that the mosfet itself was drawing more gate power than it really should be.  So I decided to replace ALL 16 mosfets from the same batch. Not new , but from a donor main board that was working on the battery side of things.

Problem solved . The main board was drawing about 255mA from 50VDC , which is sort of my benchmark when the 3525 PWM  becomes enabled, and about nothing else. Also the drivers got warm as i am use too but not hot . Same with the -12V regulator . Tested in battery mode at 2.5kW ac load , and charged the battery at 30A in line mode. Happiness.

So I think when the first  mosfet goes short circuit , one of its failure modes includes the gate shorted to drain as well , and this can deliver nasty 50V onto the driver , and that is spread towards the other 3 mosfet siblings as well. This is of course way beyond the 20V max rating of gate voltage. It appears it damaged the gate silicon to a level where high gate currents are in effect , although the device actually works! , which is not good news from a quality point of view .

So , although this is not news for experienced members , it is of utmost importance to replace all mosfets or igbts in parallel in a string with matched and device from the same batch.  I am not too sure if its really important to match between strings as well. I personally dont think so , but I would like to learn your thoughts.

Below is a measurement I took  over the gate drive resistor of one good  mosfet.  This is typically a 22R value . So although the measurement is voltage , it actually represents the gate current over a 22R resistor :

The positive current pulse is  representative of the switch-on transition of the gate . The negative current pulse is  representative of the switch-off  transition of the gate . This is due to the parasitic capacitance on the gate with respect to mosfet source. So these pulses really represent the gate charge and discharge characteristic.

 Also noteworth is the observation that when the gate is steady on , or steady off , there is no current being drawn , due to the high DC impedance of the  gate. 

Ths is an important test , so we can verify the gate drive . A simple measurement with the true RMS multimeter reveals a current of about 57mV AC over 22R , in relation to the diagram above. if this reading becomes in the order of 140mV , that mosfet is faulty. I see that value from 50mV to 70mV is good. Also all mosfets within the string should have the same reading , within 1mV or so.

I had one working mosfet pulling 155mV gate current which is bad . I have not made oscilloscope test with that faulty device , but my gut feeling is that the mosfet was drawing dc current while in steady state , but this is a shot in the dark.

Also noteworthy is that I have learned so far that the gate charge and discharge times for the mosfets and igbts I have been working with is around 800 to 1000 nS . This is also something I test as a benchmark while repairing. Note this slew rate might be different on the DC-AC devices, they use a different driver, and I am yet to measure. 

 

 

 

Edited September 10, 20232 yr by BritishRacingGreen

Hello @BritishRacingGreen can you share waveform of primary side and secondary side of gate drive transformer tx5,tx8,tx10 and Tx11 in dc side and buck side igbt showing peak to peak voltage. Facing some ringing in igbt side. 

 

it is of utmost importance to replace all mosfets or igbts in parallel in a string with matched and device from the same batch.  I am not too sure if its really important to match between strings as well.

My policy and strong suggestion is: any time a MOSFET or IGBT has failed short circuit, replace all devices in that string (i.e all upper and all lower devices on that side of the transformer or load). It seems that the other string should be OK if it didn't shoot-through.

As you have found, it is possible to "non-fatally wound" a power transistor, and this will cause trouble down the line, perhaps not far down the line.

 

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.

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.

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.

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.

 

oh  yes  got it ....it boostap  circuit....capacitor shorts  baise one 2 transistor   they work power  pin 7 

after that cap  charges  and  open  circuit  both transisors  are off again but at that time auxiliary winding brings

the  juice  again  and  ka3842  works again  amazing   stuff  that capacitor is  failure point  for  sure with out it  inverter may not work

 

thanks  colomb  

 

have  little  comment    for  the  bidrriectional  converter  

battery and  bus are locked   by transfer  ratio  the inverter power  dirction is most natural

 

need  and  demand   

 

however  beyond capctor  of 500vdc   the  thoery  is little   diffrent

the cpu  sense   the need and  demand  and  adjust  the phse  to  allow  sine wave  created  by cpu to get out of it

or  the one   at input  to get in     

the cpu  dont  force  the battery to accpet  power or   to deliver power

the power  dirction is natural   however the cpu complete this  power  dirction once it reaches to  500vdc  capactor 

from there its    export power   if  he  "felt "  that  power  dirction  from battery to line

and   import  power  if  he  feels  that   power  dirction into the battery

 

 

Before I kick off with part 2 of chapter 3 of the soft way of  bringing up the inverter , I would like to mention an important thing I have picked up during the week while repairing the battery side mosfet drivers and mosfet of a 5kW Axpert. This will be part of part 2 coming.

The battery side had a short circuit being a result of 3 of the 4 mosfet strings being short circuited. So I  isolated the shorted mosfet and removed them. Some strings had only one mosfet shorted , the other 3 mosfets in the string seemed ok.

So I only replaced the faulty ones with an exact type. Done all the basic tests , and finally the machine inverted in battery mode . Nice. But the driver circuit became rather hot after a while , including the -12V regulator on the main SMPS. So my first assessment was that of partially failed driver components. I isolated the 22R driver resistors on one string and that made a big difference in power consumed.  So this led to the fact that the mosfet itself was drawing more gate power than it really should be.  So I decided to replace ALL 16 mosfets from the same batch. Not new , but from a donor main board that was working on the battery side of things.

Problem solved . The main board was drawing about 255mA from 50VDC , which is sort of my benchmark when the 3525 PWM  becomes enabled, and about nothing else. Also the drivers got warm as i am use too but not hot . Same with the -12V regulator . Tested in battery mode at 2.5kW ac load , and charged the battery at 30A in line mode. Happiness.

So I think when the first  mosfet goes short circuit , one of its failure modes includes the gate shorted to drain as well , and this can deliver nasty 50V onto the driver , and that is spread towards the other 3 mosfet siblings as well. This is of course way beyond the 20V max rating of gate voltage. It appears it damaged the gate silicon to a level where high gate currents are in effect , although the device actually works! , which is not good news from a quality point of view .

So , although this is not news for experienced members , it is of utmost importance to replace all mosfets or igbts in parallel in a string with matched and device from the same batch.  I am not too sure if its really important to match between strings as well. I personally dont think so , but I would like to learn your thoughts.

Below is a measurement I took  over the gate drive resistor of one good  mosfet.  This is typically a 22R value . So although the measurement is voltage , it actually represents the gate current over a 22R resistor :

The positive current pulse is  representative of the switch-on transition of the gate . The negative current pulse is  representative of the switch-off  transition of the gate . This is due to the parasitic capacitance on the gate with respect to mosfet source. So these pulses really represent the gate charge and discharge characteristic.

 Also noteworth is the observation that when the gate is steady on , or steady off , there is no current being drawn , due to the high DC impedance of the  gate. 

Ths is an important test , so we can verify the gate drive . A simple measurement with the true RMS multimeter reveals a current of about 57mV AC over 22R , in relation to the diagram above. if this reading becomes in the order of 140mV , that mosfet is faulty. I see that value from 50mV to 70mV is good. Also all mosfets within the string should have the same reading , within 1mV or so.

I had one working mosfet pulling 155mV gate current which is bad . I have not made oscilloscope test with that faulty device , but my gut feeling is that the mosfet was drawing dc current while in steady state , but this is a shot in the dark.

Also noteworthy is that I have learned so far that the gate charge and discharge times for the mosfets and igbts I have been working with is around 800 to 1000 nS . This is also something I test as a benchmark while repairing. Note this slew rate might be different on the DC-AC devices, they use a different driver, and I am yet to measure. 

 

 

 

i think this is not  a  universal   measurment that applied on   all   inverters  since  there could  be  a  difrrence in   resistance  value  and  the  gate  capacitance  of the mosfet ?

 

i think this is not  a  universal   measurement that applied on   all   inverters  since  there could  be  a  difference in   resistance  value  and  the  gate  capacitance  of the mosfet ?

Yes , you are quite correct , but this also my goal with my testing and procedures in order to match drivers to a particular IGBT/MOSFET . So my goal is to build up a reference set of parameters within the requirement of the AXPERT 5kW machines , whether old MKS2 or MKS4 variations.  So  in the end I will have some good reference waveform specifications in order to verify that a particular driver arrangement will match a particular type of IGBT. This includes the both the AC and DC characteristics of the IGBT , as well as the drive current profile  as I have mentioned on previous post.  For instance , I can  then decide if I need the anti-paralleling resistor/diode circuit or not  in the gate drive for example if my replacement IGBT is of type 60N65  by measuring the drive pulses. If the IGBT gate switch off time is too far off from my reference data , I install the anti-paralleling components and check whether I have an order of magnitude better switch-off time than before.

 

 

 

 

from there its    export power   if  he  "felt "  that  power  direction  from battery to line

and   import  power  if  he  feels  that   power  direction into the battery

If you mean the inverter will push power into the grid if necessary to keep the bus voltage under 500 V, then I agree.

Certainly it will import power from the AC input any time that the situation demands.

All this of course only applies when the relay connecting AC-in to AC-out is on.

@BritishRacingGreen,

I came across an interesting post on the AEVA forum today. The poster had an unusual fault where his battery capacitors were literally exploding from excess heat. He wondered about MOSFET current, so he devised a way of using a MOSFET's source lead as a crude current shunt, probing very close to the device's epoxy. The results were quite interesting.

He has asked for comparisons with other models. His is a 12 V model, but I think that similar waveforms would be present in other models. I thought it was such an interesting testing technique that I'd mention it to you for possible inclusion in a suggested repair methodology. And perhaps you could provide waveforms for normal operation.

The post is here: https://forums.aeva.asn.au/viewtopic.php?p=97939#p97939

BTW, his current waveforms look very wrong: no current for half the cycle, and in the cycle with current, it reverses direction and has a little glitch in the middle!

Edited September 15, 20232 yr by Coulomb

I get the feeling he is only seeing ground path inductance in those plots.  They look pretty much like what you would expect if measuring over a small inductor in series with the source.

 

@BritishRacingGreen,

I came across an interesting post on the AEVA forum today. The poster had an unusual fault where his battery capacitors were literally exploding from excess heat. He wondered about MOSFET current, so he devised a way of using a MOSFET's source lead as a crude current shunt, probing very close to the device's epoxy. The results were quite interesting.

He has asked for comparisons with other models. His is a 12 V model, but I think that similar waveforms would be present in other models. I thought it was such an interesting testing technique that I'd mention it to you for possible inclusion in a suggested repair methodology. And perhaps you could provide waveforms for normal operation.

The post is here: https://forums.aeva.asn.au/viewtopic.php?p=97939#p97939

BTW, his current waveforms look very wrong: no current for half the cycle, and in the cycle with current, it reverses direction and has a little glitch in the middle!

Very interesting , a non-intrusive way like this is great to test current thru the MOSFETS .  I will for sure check this out , but unfortunately not on the main boards I have repaired already as they are assembled .  I am busy repairing a donor main board which will be my own testing jig for all future tests . Unfortunately I have declared myself as clumsy when it comes to post repairing testing on 'live' system with proper 400V dc on the main bus using oscilloscope. I need to be honest by declaring I have destroyed some silicon due to bad testing methods.  Especially the battery side MOSFETS. I have inadvertently excited some MOSFETS while hooking on the scope probes  , slipping leads etc., and the end results are very unforgiving. Hence my procedures to do low voltage low energy verification first , which is proving to be affective. But when I scale up to 400VDC/220VAC , everything becomes out bounds in term of PCB testing, apart from temperature tests.

Having said this , as I mentioned I am busy bringing up my own main board test jig and I will keep this proposal of current in mind. The scope traces I cannot relate to . The only place where you can really perform a 2 channel trace is on the two strings that have have their MOSFET sources to BAT- ( the 2 lower part of the full bridge ) . And obviously I can vote that the 3525 phases are alternate as per the two channels. Also the one channel of the member's trace  shows a negative current in relation to the other channel. That's confusing , if we are flowing power   from the battery towards the DC bus then all current of the four full bridge strings on the battery side flows from drain to source . Conversely if we charging , then current flows from source to drain for all 4 strings . I would like to know where he hooked up his two channels .

 

 

In some cases with some boards with heavy copper fills, I have previously resorted to using a soldering iron from both sides, before trading one for a compressed air gun to blow out the solder. It works really well on stubborn vias, but be prepared for a cleanup.

Having proper equipment is really worth the investment, though!

I use Stainless Steel(solder can't/won't tin SSteel) Electric fence wire holding solder iron on opposite side of pad to to heat up solder then push through towards the solder iron. Works very well, after struggling as much with "solder balls" jammed inside pads/vias....

 

I would like to know where he hooked up his two channels .

I don't know for sure, but this is my guess:

That's for one side of the transformer, the other channel is obviously on the opposite side (one of the 4 MOSFETs at the other end of the row of MOSFETs).

It occurs to me that the references are not independent. It would be best to use a 4-channel scope, and subtract the different earths (references). That might be why the current appears to change sign part way through. So maybe this technique only really works for one side at a time, unless you have a 4-channel scope.

Bear in mind that his is a 12 V in 120 V out Xantrax, that bears a remarkable similarity to Axperts, but may or may not be made in a Voltronics factory.

Edited September 18, 20232 yr by Coulomb

 

Since my last post , I had been involved in creating the schematic page for the dc-ac igbt and buck igbt driver section.  In the process I also updated the power chain schematic. The two schematics below refers:

 

First the update power chain :

 

and the igbt driver section schematic :

 

I struggled initially to understand the 3 isolated power supply feeds . So I added a little context diagram in the bottom left hand of the page. An igbt drive signal must be referenced to its associated emitter .  So from the power chain drawing and the context diagram , one will see that we need  three separate isolated supplies ,one each referenced to INV L  , INV N and BUS - .  The half wave rectified dc output of each  secondary winding  has a RMS value of about 20.5 V . This feeds a 5v6 zener diode via a 18k resistor . In practice there is about 5.3V across the 5v6 zener and 14.8v across the resistor , and of course there are filter capacitors in order to smooth the two voltages . The 18v zener has no function other than clamping overvoltage . If you take the top power supply as example ,  you'll notice that INV L is connected onto the midpoint of the zeners . Now because INV L is the emitter connection for QB2 transistor , the gate voltage can be set to either 14.8V or -5.3V , which is exactly a good turn on and turn off voltage level for the gate.

It must be noted that in contrast to the DC-DC converter on the battery side , the DC-AC converter and BUCK converter are under processor control. The DSP therefore will produce appropriate PWM  signals for the IGBT transistors .This is of course done by a driver interface  IC eg U1.  This ic differ between MAX7.2 and 5KW models , but the basic function remains the same. Interesting to note is this one is not an opto-coupler , but an eDiode (emulated diode , see datasheet) . So light is actually not used .

The processor drive electronics confused me somewhat because of its complexity. I understand the circuit must be fail-safe but I  still don't get the rationale behind this one. Basically R180,R179 and R170 will bias the driver diode into on state . But transistor Q3 is also biased to on , and so will shunt enough diode current away , so the drive diode is off. The voltage across pin 1 and pin3 is 0.1V , the spec requires anything lower than 0.8V . Otherwise with the transistor off, the forward current is about 10mA which switches the ic on (about 2.1V across the diode). The  control pin CN11-1 is active low , so when the DSP wants to switch the IGBT on , then it pulls CN11-1 low , which lowers Q3 bias voltage, shunting less current away from the diode . When on , the pin 6 voltage (14.8v) is routed to the gate output , when off pin 4 (-5.3v)  voltage is routed to gate. 

I initially thought that because of the low resistances of R180,R179 and R170 across the 12V rail , that the faulty high 12V rail  voltage  would probably have  destroyed the driver. But it didn't , not one of them five circuits . The transistor probably have a high enough VCE rating , so it probably saved my but. So really no further damage experienced here except of course the driver transformer , which I rewound.

To test the  circuit , I checked the gate voltage on all igbt's while manually switching the control pins by shorting them. They all passed this DC characteristic test quite nicely.

But I soon discovered that I also need to perform AC characteristic tests , by applying suitable narrow PWM pulses on the pins , and checking the results on the gate pins. This I haven't done yet , because I realized that I will inevitably have to  introduce a controller that can help me out here . So the rest of spare time this week i am going to look at Raspberry Pi Pico and introduce some software control code to drive these signals. In the spirit of this being an open source hardware thread , I will also make the software available to whoever might be interested. This controller will also help me with the relay testing , the bus soft start and other sundry tests.  In the long term I have also a plan to introduce a rudimentary oscilloscope function with small display , in order to empower someone to check signal quality where an expensive scope is not available.

Apart from DC and AC characteristic tests , I will probably also need to verify that I have good isolation between the three sections , and that breakdown will not occur when I start switching the power chain.  How this can be accomplish I must still learn , because low voltage resistance tests will most probably be  not good enough.

I have spent a lot of time here , because the IGBT transistors are particularly expensive , and one would like to leave very little to the imagination when replacing these transistors.

i will come back to do the AC tests when i have done  the software , and bread boarded a rpi pico.

In the meanwhile , I am anxious to move onto the bus soft start section , which intrigues me to no ends . 

 

 

Hi   is  CN11-1  the connector that connects/plugs the DSP board into the main board...

I need to confirm where the signals come from to pulse the LED on U1 as on my board  only QB has sine wave pulses

 

Hi   is  CN11-1  the connector that connects/plugs the DSP board into the main board...

I need to confirm where the signals come from to pulse the LED on U1 as on my board  only QB has sine wave pulses

Hi @AndrewF there is three female connectors on the main board , they are CN9,CN10 and CN11. Well as far as the Axpert 5kW inverter is concerned . 

I am releasing the block diagram below  prematurely for your benefit :

 

So the first 5 pins on CN11 is for pulsing of QB , QA , QD , QC , and the buck igbt.  These pins are normally left open to switch the corresponding igbt off. By shorting the pin to GND , CN10-10 , the corresponding igbt is swiched on.

You can use a hard wire and switch them on individually as you like , or use a microcontroller with an open collector output to drive them .

 

 

Hi @AndrewF there is three female connectors on the main board , they are CN9,CN10 and CN11. Well as far as the Axpert 5kW inverter is concerned . 

I am releasing the block diagram below  prematurely for your benefit :

 

So the first 5 pins on CN11 is for pulsing of QB , QA , QD , QC , and the buck igbt.  These pins are normally left open to switch the corresponding igbt off. By shorting the pin to GND , CN10-10 , the corresponding igbt is swiched on.

You can use a hard wire and switch them on individually as you like , or use a microcontroller with an open collector output to drive them .

 

and here is associated schematic , also premature , needs additional checking.

 

Edited September 22, 20232 yr by BritishRacingGreen

Are you sure those diode-capacitor pairs between pin 8 and 7 are not between 8 and 5? As shown there, they will kill the switching speed and likely cook the drivers?

 

Are you sure those diode-capacitor pairs between pin 8 and 7 are not between 8 and 5? As shown there, they will kill the switching speed and likely cook the drivers?

Yes, it is wrong. Capacitor is connected between 8 and 5. Diode is fine.

 

Are you sure those diode-capacitor pairs between pin 8 and 7 are not between 8 and 5? As shown there, they will kill the switching speed and likely cook the drivers?

yes , thanks , indeed the cap is across T350 ic pins 5 and 8. Corrected it in above post as version 1xB.

Hi @Coulomb , I have experienced a very frustrating problem on a 5kW mainboard where I replaced a 3525 pwm controller and 7912 voltage regulator . The board had a SG3525A chip and I replaced it with a SG3525AN one. I have searched far and wide for the difference between the two , but only info that I have is that the N only denotes a plastic DIP footprint.

After I have performed all initial tests , I brought up the main board with its controller card . But as soon as it starts the DC-AC converter , the DSP will either switch off , or reset itself. After a number of retries , it blew 2 x dc-dc igbts and the buck IGBT.

I was really frustrated , and I temp fixed the board by replacing the dc-dc IGBT's with 1N4007 (1000V / 1A)  diodes , as well as the buck IGBT. So that is ok for  no-load  and for battery mode only when producing AC. of course to accommodate the low amps of 1n4007 , I removed the bulk bus capacitors  and replaced with small C value cap. (this is risky when there is heavy inrush to the caps , but dsp soft starts helps. And if I don't use dsp  I ramp the voltage manually by varying the input battery terminals from 2.5V upwards ). The 4007 can also handle a forward surge current of 30A.

I had no more silicon failures while testing , but the DSP would still fail exactly when the DC-AC starts firing. 

I then checked for SMPS supply rail failures , and  detected with the scope that the -12V dipped when the DSP instructed 3525 PWM to switch off. This occurs when the dc-dc is on and when the HV bus becomes too high , it switches off for short periods.

Long story short , I removed fet transistors Q60 and Q61   as you can see in the diagram below:

Problem solved , no more short circuits on the -12V. It turns out that when the DSP switches off the 3525, the 3525 disables its PWM outputs A and B via a positive voltage that is applied on pin 10 of the 3525. At the same time Q60 and Q61 , which is also controlled by the shutdown command , clamps the gate driver to -12V . Now my scope indicated a short delay before the 3525 outputs are actually removed. With the Q60 and Q61 immediately biased , there is actually a short between +12V and -12V for a finite period. This causes the short circuit on -12V regulator , it current limits and its output turns off for a short period.

I then investigated why the 3525 will not immediately shuts the PWM off , and realized that is because of the low pass filter around C15 and R108. So I started to measure passive components value and consulted another 5kW board to check its values, only to find that C15 and R108 are NOT fitted !

So I removed these two components from the board under test , replaced Q60 and Q61 , brought up the inverter , and the results can't get better !

 

This does bring up more questions than answers:

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

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

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

4. why do some boards fits C15 and R108

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

Anyway , this failure mode important to detect going forward when we debug the working of the dc-dc inverter. It is important to switch the shutdown command off and check that there are no rail current surges , and also to check that there is a graceful shutdown of drive pulses on the IGBT  and MOSFET drive pins , as opposed to stray pulses during the metastable period.

 

 

 

Edited September 23, 20232 yr by BritishRacingGreen

I think something else may be wrong here.

Worst case, the SG3525 outputs will shut down 0.5us after SD pin exceeds 1V.

R113/C57 already provides a delay function with an initial rise of 0.17V/us.  So as long as the threshold voltage for Q60/Q61 is above ~1.2V, they will not turn on at all until long after the SG3525 is shut down.

C15/R108 seems to have 2 functions: (1) make sure the device is NOT shut down at power up, and (2) increase the ramp time/delay even more.

Not sure why Q60/61 are there. They seem to only speed up the final decay of 1 Vbe drop.  If it was other side the capacitor (directly on the coil), it would be a very effective final clamp.

The DC blocking capacitor is quite a bit bigger than for the IGBT drive, which implies more current, and also more energy in the LC tank which would need to be damped at shutdown, which could explain why a clamp is included.

 

I think something else may be wrong here.

Worst case, the SG3525 outputs will shut down 0.5us after SD pin exceeds 1V.

R113/C57 already provides a delay function with an initial rise of 0.17V/us.  So as long as the threshold voltage for Q60/Q61 is above ~1.2V, they will not turn on at all until long after the SG3525 is shut down.

C15/R108 seems to have 2 functions: (1) make sure the device is NOT shut down at power up, and (2) increase the ramp time/delay even more.

Not sure why Q60/61 are there. They seem to only speed up the final decay of 1 Vbe drop.  If it was other side the capacitor (directly on the coil), it would be a very effective final clamp.

The DC blocking capacitor is quite a bit bigger than for the IGBT drive, which implies more current, and also more energy in the LC tank which would need to be damped at shutdown, which could explain why a clamp is included.

Hi Justin, I am busy working thru your comments, thanks. 

Only thing I can respond to so far is as per powerup, are you aware of the soft start feature of 3525? , so on face value this should keep the outputs at bay at startup, I think. 

Good point - that time constant is not even enough to make a dent on the soft-start cap charge.

The off time may just be a side effect of the need to discharge the cap.

I see one other side effect. Adding R108 reduces the SHDN voltage from ~12V to ~5V (relative to -12V).

Is it possible that they use alternative parts for Q60/Q61? If they use one with a low threshold and low Vgsmax, then adding this RC pair could make a difference.

 

yes , thanks , indeed the cap is across T350 ic pins 5 and 8. Corrected it in above post as version 1xB.

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