What is the reason for MOSFET burning?

[--BawA--]

hi
i will come to conclusion of "why mosfets are burning", if i get the answer of following questions.

1)have u connected the LC filter in the output stage of hbridge,? if yes, have u ever checked your hbridge circuit at 400v dc link and with no filter connected?
2)which scheme have u used to drive the mosfet of h bridge i.e. bipolar spwm or unipolar spwm?

 

[atom1477]

And tell how you get software deadtime, because i think it is imposible (when PWM is generated hardwarely).

 

[ratheeshvaram@yahoo.com]

hi
i will come to conclusion of "why mosfets are burning", if i get the answer of following questions.

1)have u connected the LC filter in the output stage of hbridge,? if yes, have u ever checked your hbridge circuit at 400v dc link and with no filter connected?
2)which scheme have u used to drive the mosfet of h bridge i.e. bipolar spwm or unipolar spwm?

hi
i will come to conclusion of "why mosfets are burning", if i get the answer of following questions.

1)have u connected the LC filter in the output stage of hbridge,? if yes, have u ever checked your hbridge circuit at 400v dc link and with no filter connected?
2)which scheme have u used to drive the mosfet of h bridge i.e. bipolar spwm or unipolar spwm?

-no i didnt connect LC filter
i had checked h bridge with 400v dc, and that time total supply shorts anf mofet burns
- i am using unipolar independeny scheme..
i.e. for 10 ms 30khz applied is applied to Q3, Q6 is full on and after 10ms Q3,Q6 off then dead time, (1ms. for testing) then for next 10ms 30khz applies to Q5,Q4 is full on and after 10ms Q5, Q4 off then dead time.......

 

[atom1477]

- i am using unipolar independeny scheme..
i.e. for 10 ms 30khz applied is applied to Q3, Q6 is full on and after 10ms Q3,Q6 off then dead time, (1ms. for testing) then for next 10ms 30khz applies to Q5,Q4 is full on and after 10ms Q5, Q4 off then dead time.......

I thought that You using bipolar PWM. And then the software deadtime was impossible because it would have to be included in each PWM cycle.
In this case (unipolar PWM) is what you described is basically not a deadtime but it should fulfill its role.
You must have an error in the control scheme.
Join the bulbs (2 in series, as low power as you find, 10W/230V for example) as I drew and see what happens (I think their blinks strongly).

 

[--BawA--]

you have said that your circuit is working properly at low voltage , but at high voltage your mosfet get damaged. u are using 1k resistance as a load.

i presume that you have provided sufficient dead time, so the problem might be in your bootstrap circuit. try to use some lower value of resistance like 4.7k from gate to source and slightly higher value of bootstrap capacitor, selection of capacitor should be done in such a manner that it can provide a voltage greater than 8v for entire 10ms,, as this capacitor is discharged by the resistance that u r using from gate to source,,so to decide the value of capacitor ,u can use capacitor discharging equation .

also your scheme that u have witnessed is not unipolar, its modified version of bipolar. so try to use UNIPOLAR spwm, to avoid body diode conduction.

 

[ratheeshvaram@yahoo.com]

I thought that You using bipolar PWM. And then the software deadtime was impossible because it would have to be included in each PWM cycle.
In this case (unipolar PWM) is what you described is basically not a deadtime but it should fulfill its role.
You must have an error in the control scheme.
Join the bulbs (2 in series, as low power as you find, 10W/230V for example) as I drew and see what happens (I think their blinks strongly).

ok atom,
i will try...

- - - Updated - - -

you have said that your circuit is working properly at low voltage , but at high voltage your mosfet get damaged. u are using 1k resistance as a load.

i presume that you have provided sufficient dead time, so the problem might be in your bootstrap circuit. try to use some lower value of resistance like 4.7k from gate to source and slightly higher value of bootstrap capacitor, selection of capacitor should be done in such a manner that it can provide a voltage greater than 8v for entire 10ms,, as this capacitor is discharged by the resistance that u r using from gate to source,,so to decide the value of capacitor ,u can use capacitor discharging equation .

also your scheme that u have witnessed is not unipolar, its modified version of bipolar. so try to use UNIPOLAR spwm, to avoid body diode conduction.

ok bawa,
thank u. and i will make all these changes and inform...

 

[ratheeshvaram@yahoo.com]

I thought that You using bipolar PWM. And then the software deadtime was impossible because it would have to be included in each PWM cycle.
In this case (unipolar PWM) is what you described is basically not a deadtime but it should fulfill its role.
You must have an error in the control scheme.
Join the bulbs (2 in series, as low power as you find, 10W/230V for example) as I drew and see what happens (I think their blinks strongly).

i connected 2 5w bulb in series with Mvolt...
then it blinks strongly...

after repeating the testing with defferent values of CB and Rgs.. I got many mistakes.. but atlast when I removed the CRO from testing the blinking stoped and I placed 47uF/35V as CB and 1K as Rgs.. but I am getting only ~12VAC... across 1K load... what may be the reason please help....

Ratheesh

 

[ratheeshvaram@yahoo.com]

So try my solution for preventing H-bridges burning while testing:
View attachment 98504
Connecting bulb before bus capacitor (for example in the rectifier input in AC powered H-bridges) not prevent MOSFETS.
This prevent.

If bulb (bulbs in Yours example) blinks You have strong Cross-Conduction.
But often bulb not blinks although cross-conduction occurs. This is becouse cross-conduction take short part of time. So measure voltage on Q3 and Q5 Drains by oscilloscope. Voltage should not falls to much. If falls, You have cross-conduction.

PS. You have very low-power DC-DC converter so maybe use low power bulbs, for example 10W.

atom,
I connected two bulbs as u drawn (both are 5W 230V), and one 5W bulb across the load instead of 1k resistor.
-I connected the multimeter across load
-CRO from Q3, Q5 drain to ground
when the DC-DC converter begins (with out Hbridge), nearest 400V is getting on CRO..
after 10s Hbridge starts and showing ~60VAC on multimeter ( load bulb is on with low light, also some voltage is across the series bulb)
On the CRO the voltage is reduced to ~200V and it is a pulse similar to the gate of MOSFETs...
Now the switching freq is 4KHz, it is applied on high side mosfet and the sine freq is 50Hz (it is applied on lowside)..
the dead time is 200uS
CB=47uF, Rgs=4.7K, all other are same as the schematic...
- when I increased the carrier freq to 16KHz or above, the series bulb is blinkig...
-and when I am connecting the same 50Hz to high side and 4KHz to low side, the series bulb is blinking....

what may be the reason....
why pulse on drain
why output is 60VAC...

please help................

 

[atom1477]

On the CRO the voltage is reduced to ~200V and it is a pulse similar to the gate of MOSFETs...

Strange. Are you sure that your DC/DC converter still capable 400VDC (you said that is only 10W power)?

I connected two bulbs as u drawn (both are 5W 230V), and one 5W bulb across the load instead of 1k resistor.

You should not collect any load during this test.
You never get full power on 5W load bulb if You have two another same-type bulbs in series with H-bridge.
Discoonect load bulb and 1k resistor too. And we will se if the voltage on Q3 and Q5 drains will continue to decline.

after 10s Hbridge starts and showing ~60VAC on multimeter ( load bulb is on with low light, also some voltage is across the series bulb)

I doubt that the multimeter correctly measured this signal. Nevertheless, if the bulb is on with low light this means that the inverter will work with. You only have a problem with cross-conduction since the voltage at the drains of Q3 and Q5 has the shape similar to the pulse of the gate of MOSFETs.
But I quite not understand how it is look like.
Is is oscillates from 0 to 200VDC? Or from 200 to 400VDC?
I would like to see photos from the oscilloscope screen.
How many channels Your's oscilloscope have?
I would like to see:
Signals on HIN and LIN on the one of the IR2113. If Your's oscilloscope is Dual-Trace, both signals on one photo.
Signal on the drains of Q3 and Q5. Tag on the picture where the 0V level is.
Signal on Mvolt.

 

[iop95]

"n high side I am giving 30KHz and low side 40-70Hz
all ground are connected together.."

I hope you reffer at diagonal switchs.

 

[atom1477]

In general this whole control:

Now the switching freq is 4KHz, it is applied on high side mosfet and the sine freq is 50Hz (it is applied on lowside)..

is some strange.
These 50Hz should not be simple 50Hz, it should be 16kHz PWM modulated by 50Hz frequency (one half of the 50Hz period).
But suppose it is.
In that case, as it is supplied to the lower MOSFET, the upper (of course at diagonal) should be still attached. Not get PWM.

I think the easier (at least for testing) is driving both sides (I mean left and right, not low and high) of the H-bridge by waveforms in antiphase. So one (left) side by 16kHz PWM modulated by 50Hz at 0° phase, and the second (right) the same 16kHz PWM modulated by 50Hz but at 180° phase (I mean 16kHz PWM is in phase for both sides. In antiphase is 50Hz modulating frequency).
In this case modulating 50Hz frequency is full 50Hz wave, not only half of period.
In my opinion it is simplier to generate, and giving results simplier to measure.
But this required hardwarely, cycle-by-cycle, deadtime.