Sonsivri
 
*
Welcome, Guest. Please login or register.
Did you miss your activation email?
December 04, 2016, 11:24:25 11:24


Login with username, password and session length


Pages: [1]
Print
Author Topic: Detemination of causes of Power IGBT failure and Safe operating area of IGBT  (Read 1873 times)
0 Members and 1 Guest are viewing this topic.
zab
Active Member
***
Offline Offline

Posts: 132

Thank You
-Given: 24
-Receive: 54


« on: October 25, 2014, 09:09:09 09:09 »

  IGBT failure is the major fault in power Electronics. There may be many reasons of its failure .
1 Being operated out of The safe operating area (SOA).
2 Driver circuit
3 lack of protection circuit
4 Cross conduction
5 Lack of Thermal conduction
are the major reason of its failure  I am confronting this problem in induction heating machine (Chinese make). The driver circuitry along with IGBT both got damaged. I am asking for help and guidance for analysis and remedial measures need to be taken.
Secondly The data sheet tells too many terms need to be taken into account for IGBT suitability for its operation.Determination of its operating frequency is also a difficult task.
« Last Edit: October 25, 2014, 01:19:05 13:19 by zab » Logged
fpgaguy
V.I.P
Junior Member
*****
Offline Offline

Posts: 89

Thank You
-Given: 98
-Receive: 86



« Reply #1 on: October 27, 2014, 05:40:52 17:40 »

flowchart for failure modes here:
www.fujielectric.com/products/semiconductor/technical/application/box/doc/REH984b/REH984b_04a.pdf

However a circuit , part numbers , and some scope traces and heat sink dimensions would be helpful otherwise
Logged
pickit2
Moderator
Hero Member
*****
Offline Offline

Posts: 3816

Thank You
-Given: 567
-Receive: 2049


There is no evidence that I muted SoNsIvRi


« Reply #2 on: October 27, 2014, 10:44:30 22:44 »

I done a report for a local microwave & oven repair shop.
As this was over 4 years ago, and don't have a copy of the work I done, I can only go by what I recommended to fix the faults they where having.
There were a number of makes, the main problems they were soon after a fix they were having returns for the same fault, this was they only replaced the IGBT, with the same type.
these were running very close to the limits of the device, added to this the driver chips and components (diodes and capacitors) in some designs needed replacing to.
Most successful result was to use a IGBT with higher max voltage and current load, than the ones fitted.

Noted: heat sink was another problem, some failed due to fan &/or duct obstruction, grease gets everywhere in a kitchen.
ever get silver finger when fixing microwave ovens. Sad

botton line most are built to a price not for quality of components used.
Logged

Note: If you have no posts other than, I want or reporting a dead link Then you can't complain If I remove your post So Stop Leeching
zab
Active Member
***
Offline Offline

Posts: 132

Thank You
-Given: 24
-Receive: 54


« Reply #3 on: October 28, 2014, 06:09:48 06:09 »

IGBTs used here are 1200v 300A FF300r12ks4 4 sets have been made each set have 2 double IGBT to make 200kw .The problem is, there IGBT anlong with its goes faulty simultaneously
the driver circuit has 2 4.7 ohm in parallel resistors  2  18v zener after these and 15k resistor at gate to emitter. the gate get damaged inspite of these zeners diode get short circuit. the question is why gate gets damaged when protection diode get short ?
Logged
pickit2
Moderator
Hero Member
*****
Offline Offline

Posts: 3816

Thank You
-Given: 567
-Receive: 2049


There is no evidence that I muted SoNsIvRi


« Reply #4 on: October 28, 2014, 09:43:46 09:43 »

I would suspect the problem is in the AC power supply, there is raw ac being fed into the driver circuit.
In some microwave ovens when there is a lamp or power fuse failure this can send a power surge into the low voltage side of the circuit, this wipes out the relay transistor and diode protection, both devices are found to be short circuit. this is just like what your seeing.
Logged

Note: If you have no posts other than, I want or reporting a dead link Then you can't complain If I remove your post So Stop Leeching
hate
Hero Member
*****
Offline Offline

Posts: 556

Thank You
-Given: 156
-Receive: 354


« Reply #5 on: October 28, 2014, 11:12:59 11:12 »

zab: A schematic of gate drive and related parts may do better in getting help. Regardless of that, what pickit said is worth a check. Also driving the gate in overrated frequencies may damage the gate but I don't think that's the case as the gate is probably driven at the resonance frequency. Again, a schematic may do better.
Logged

Regards...
fpgaguy
V.I.P
Junior Member
*****
Offline Offline

Posts: 89

Thank You
-Given: 98
-Receive: 86



« Reply #6 on: October 28, 2014, 06:04:14 18:04 »

yes - what pickit2 said -  put in a higher rating part for the part that's failing assuming you can identify it, repeat as necessary


be careful probing around there!

I remember one time when I was more a noob - I accidentally made my ground portion of the scope cable disappear in a puff of smoke and sparks

For switch mode circuits - one can get a good idea of their operating frequency by probing the "field" around the circuit either with a loop pickup coil for magnetic or
a probe tip for electric field - ie no physical contact with circuit desired

For most switch mode you can do this; however some quasi resonant controllers it's a bit harder to make sense of what one sees









« Last Edit: October 28, 2014, 06:08:06 18:08 by fpgaguy » Logged
nPn
Newbie
*
 Warned
Offline Offline

Posts: 10

Thank You
-Given: 6
-Receive: 2


« Reply #7 on: October 31, 2014, 07:01:49 07:01 »

Oh yes scope probes are notorious for either smoking boards or getting smoked. You said you have four sets, maybe you have some for high side switching? Take care for high side probing and mind where your ground clip goes!

Remember scope channels share common ground and are usually earthed with low resistance. If you put the ground clip any high side IGBT terminal for non-floating circuit, boom. :\

Connect ground to ground and use another scope channel for the floating ground you want to reference against, then use math mode to subtract that from what you're measuring.
Logged
zab
Active Member
***
Offline Offline

Posts: 132

Thank You
-Given: 24
-Receive: 54


« Reply #8 on: November 01, 2014, 07:54:24 07:54 »

A schematic of gate drive and related parts may do better in getting help.

The gate circuitry is attached for deference. I wonder when zener diode at gate got short due to any reason  then gate of the IGBT should not be got damaged. The connector is attached to gate drive circuit .Which provide +15v for switching IGBT ON and -5V for Switching it OFF . gate components have actual values used in circuit. while snubber circuit components values are not actual.

Logged
hate
Hero Member
*****
Offline Offline

Posts: 556

Thank You
-Given: 156
-Receive: 354


« Reply #9 on: November 01, 2014, 10:47:57 10:47 »

Do you measure a ~.7V between gate and emitter of the damaged IGBTs or do they just not work? Gate protection circuit seems capable but if the zeners get short, it may be because of a surge current resulting from a spike in power as pickit mentioned before. As an extra protection you may opt to go for a varistor or a suppressor diode instead of 2 reverse zeners in series. Either of them is much more faster than a normal zener in case of voltage spikes.
http://en.wikipedia.org/wiki/Varistor
http://en.wikipedia.org/wiki/Transient-voltage-suppression_diode

Also the parallel resistor resulting in 2.35Ohms seem to be a little weak to protect the gate from excess currents. Its value needs to be calculated taking gate drive frequency into consideration though.
Logged

Regards...
zab
Active Member
***
Offline Offline

Posts: 132

Thank You
-Given: 24
-Receive: 54


« Reply #10 on: November 01, 2014, 09:41:41 21:41 »

The Gate of IGBT acts like acapacitor, It can be check by anlogue mater. Secondly a practical way is charge gate and get continuity from emitter to collector ,and when the gate is charged negatively it disconnected from emitter to collector.beside functional trial.

Failure of zener diode to protect IGBT is really need to be explored , initially I was using 1w zener then shifted to 5w but no difference in result.
some fast medium may be required as suggested by hate.but their working capability is yet to be verified.

The frequency of the inverter is between 10khz to 50khz varies. on off curve will change with the change of these gate resistors,and may increase heat dissipation as well.
Logged
Ichan
Hero Member
*****
Offline Offline

Posts: 840

Thank You
-Given: 312
-Receive: 387



WWW
« Reply #11 on: November 02, 2014, 06:01:43 06:01 »

Which IGBT is/are damaged?

I am thinking about shoot through destruction, does the circuit before the driver has protection for it? Isolated? Enough dead band?

-ichan
Logged

There is Gray, not only Black or White.
hate
Hero Member
*****
Offline Offline

Posts: 556

Thank You
-Given: 156
-Receive: 354


« Reply #12 on: November 02, 2014, 10:32:52 10:32 »

The Gate of IGBT acts like acapacitor, It can be check by anlogue mater. Secondly a practical way is charge gate and get continuity from emitter to collector ,and when the gate is charged negatively it disconnected from emitter to collector.beside functional trial.
Are you talking in general or about the damaged IGBT? If the gate of an IGBT gets damaged, it will be a short circuit in 99.9% of the cases. So you will measure a current if you feed a voltage source between gate and emitter of a damaged IGBT. I'm trying to understand how the IGBTs failed as there is also the possibility that the snubber circuit is not enough for the back emf from the inductor of transformer.

The gate resistors are there to limit the maximum current into the gate at high frequencies where the gate cap can be considered as short circuit. So (gate drive voltage)/(gate series resistor) will give the maximum gate input current. If the maximum working frequency is 50Khz, then the maximum current will be (gate drive voltage)/(R+Zc) where R is the gate series resistor and Zc is the impedance of the gate capacitive effect at 50Khz.
« Last Edit: November 02, 2014, 10:41:02 10:41 by hate » Logged

Regards...
zab
Active Member
***
Offline Offline

Posts: 132

Thank You
-Given: 24
-Receive: 54


« Reply #13 on: November 02, 2014, 06:19:42 18:19 »

@Ichan   
No doubt dead band is very important to stop shoot through destruction IGBTs between input signals to IGBT driver.Here the in this circuit the dead band is fixed.
secondly If that happens top and bottom IGBT will be damaged simultaneously. That is not the case here. some time top side with the cross bottom IGBT fails.
The gate driver circuitry is not perfect in this case, there should be some intelligent drivers making it impossible for IGBT failure.ie short circuit protection dead band assurance , half conduction etc.

@hate
1.I was talking of IGBT testing out of circuit.
2.I have seen some drive circuits where if One IGBT got damaged others are shut off to minimize the damaged.
3. about the gate resistor value I know it is very important factor in drive circuitry. so is the snubber circuit. The value of gate resistor is High the IGBT can not work on High frequency. What should be the value in your opinion? here is the IGBT data sheet for ref. https://www.infineon.com/dgdl/Infineon-FF300R12KS4-DS-v03_02-en_de.pdf?folderId=db3a304412b407950112b4095b0601e3&fileId=db3a304412b407950112b433b35b5d41

It has many Characteristic Values in making decision of its operating frequency,current and volt. Calculation of operating frequency is a Jigsaw Puzzle as well.If some one can describe it.  He is most welcomed.

Logged
hate
Hero Member
*****
Offline Offline

Posts: 556

Thank You
-Given: 156
-Receive: 354


« Reply #14 on: November 02, 2014, 09:44:13 21:44 »

If that's an induction heating machine, the working frequency should be close to the resonant frequency of output cap and inductor. I'm guessing, the gate series resistance could be a little higher to limit the current to gate and protection circuit at high frequencies and voltage spikes, an impedance calculation should be performed to come up with a suitable value. Also check the surge current ratings for those resistors, use resistors with better surge current ratings if you can.
Logged

Regards...
kreutz
Junior Member
**
Offline Offline

Posts: 68

Thank You
-Given: 179
-Receive: 25


« Reply #15 on: November 05, 2014, 05:45:33 17:45 »

Could you post the schematic of the complete gate drive circuit? (including the ICs), and a photo of the components involved? As per the diagram above it seems ok. Gate drive resistors seem to have the correct values for the application but you must consider the dV/dt. Those resistors are also chosen to limit this parameter. Also the type and value of the capacitors used is very important (ESR, Gate drive's capacitor charge time/circuit for the Top IGBts...).

IGBT's get damaged when protection diode gets shorted because the limited gate charge received from the drive circuit makes the IGBT stay in the linear zone for too long, so it gets overheated. Protection diodes are only used to avoid gate over voltage due to Miller effect.

Note: Miller effect could also provoke IGBT turning on.
« Last Edit: November 09, 2014, 04:03:11 16:03 by kreutz » Logged
zab
Active Member
***
Offline Offline

Posts: 132

Thank You
-Given: 24
-Receive: 54


« Reply #16 on: November 06, 2014, 05:43:24 05:43 »

I have to make the schematic myself. It is not readily available.
My question was IGBT Gate voltage are far high than operated voltages . Once gate drive circuitry got problem and drive voltage changed . Now come the function of these clamping zener diodes if they got short circuit it means voltage across gate come down to zero. then gate should not got damaged. I think there may be two problems either these are slow to react (clamp) or in the process of clamping these got open and close , and gate is damaged due to excessive volt when zener is off , it is all thinking but really happens I do not ,This is what I am search of .
Logged
pickit2
Moderator
Hero Member
*****
Offline Offline

Posts: 3816

Thank You
-Given: 567
-Receive: 2049


There is no evidence that I muted SoNsIvRi


« Reply #17 on: November 06, 2014, 12:09:06 12:09 »

your driver circuit should never go over  about +12V to +15V. what is Max voltage of the gate? to be protected by back to back 18V zener pair? (protection voltage should be 18.6)
now if a power surge is wiping out your 18V zener. (are all diodes shorted? or only one of them, what about the others in the bridge? total of 8 zeners)
Then there is a power path your not seeing, this maybe in the driver circuit, or my thought is it via the pin 4 connector to your driver circuit shown on schematic you posted this is 600V AC.


Data Sheet says Gate voltage is 20V so is a value of 18.6V a save bet on protecting the gate.
« Last Edit: November 06, 2014, 12:20:48 12:20 by pickit2 » Logged

Note: If you have no posts other than, I want or reporting a dead link Then you can't complain If I remove your post So Stop Leeching
kreutz
Junior Member
**
Offline Offline

Posts: 68

Thank You
-Given: 179
-Receive: 25


« Reply #18 on: November 06, 2014, 03:31:23 15:31 »

The Miller-effect is nothing else than the feedback of the collector-emitter voltage VCE via the gate-collector capacitance CGC on the gate. This means "a change of VCE has the same effect as an internal current source into the gate bias circuit", where the current is given by the expression ig = CGC(Vce) x dVce/dt. Unfortunately CGC is not constant, but it changes its value with Vce. The strongest change of CGC results at small Vce.

Also, because of the Miller-effect, the gate current during turn-on or turn-off, first of all, is used for changing the charge of CGC. This is why charging up or down the gate is slowed down. This means that with a larger gate series resistor all events take a longer time, i.e. turning-on or turning-off last longer.

Protection diodes are used to limit the gate voltage increase due to the internal current source injecting charge into the input capacitance (CGE) when Vce changes, the rate of change of the collector-emitter voltage is also very important (dVce/dt) and is one of the potential rupture mechanisms of the device.

I assume that both the positive gate voltage, as well as the negative gate voltage power supplies are ok.




Posted on: November 06, 2014, 03:22:33 15:22 - Automerged

 Please, read also: http://www.microsemi.com/document-portal/doc_view/14698-optimizing-mosfet-and-igbt-gate-current-to-minimize-dv-dt-induced-failures-in-smps-circuits
« Last Edit: November 06, 2014, 03:34:42 15:34 by kreutz » Logged
zab
Active Member
***
Offline Offline

Posts: 132

Thank You
-Given: 24
-Receive: 54


« Reply #19 on: November 07, 2014, 02:42:33 14:42 »

The max gate volt is 20+- and here +15 v are used to switch it on _5 v are used to switch it off. while the zener at gate of 18.5v 18v+.5v zener volt +forward bias voltage. what do you suggest clamp voltage need to bring down to 17 volt 16.5+.5=17 volt .The 4 pin is used for desaturation,The  main voltage is blocked by 2 diodes in series not letting go high voltage to gate drive circuitry.


2. Longer shifting time also results in very high heating up IGBT. If you consult the data sheet you will come to know that max losses occur at switching time, not at the time when it is on or off. so delay in switching may cause more heating up . just think the middle point of switch from 300A to zero is 150 A .and at 600 volt Heat would be 150X600=90,000 watt 90kw.so dissipation of this heat would be also a problem. so is the solution , a middle path.
I put forward the problem so that we may understand all these aspects.
Logged
kreutz
Junior Member
**
Offline Offline

Posts: 68

Thank You
-Given: 179
-Receive: 25


« Reply #20 on: November 07, 2014, 10:55:29 22:55 »

Clamping voltage is ok as it was designed. What I want you to investigate and understand is that the gate circuit's protection diodes are most probably being destroyed by Miller effect and probably due to excessive dV/dt. IGBTs are able to withstand repetitive current overload for some micro-Seconds without beeing destroyed.

Without access to the equipment itself and/or its schematics we are limited to the amount of support we can give in this case. There is more information required, for example:

Is the problem often presented when powering up?
Is the problem more often presented when no load is applied to the equipment?
Is the problem presented after the equipment is being used for a while?
Does the equipment have soft start?
What is the DC voltage measured at the H bridge?
Is there a way to get oscillograms of gate voltage vs Igbt VCe? Do you have a high frequency current probe for your oscilloscope?
Does the problem present itself after tuning the tank circuit?
Can you use/(have access to) a 3-5 KVa variac for testing with a lower line voltage?
What does the manufacturer say?
Are the IGBTs the original OEM ones?

A lot more information would be also necessary that would be answered by having the service manual and/or schematics because different failure modes are associated with the specific circuit topology.





  
« Last Edit: November 08, 2014, 12:06:26 00:06 by kreutz » Logged
Pages: [1]
Print
Jump to:  


DISCLAIMER
WE DONT HOST ANY ILLEGAL FILES ON THE SERVER
USE CONTACT US TO REPORT ILLEGAL FILES
ADMINISTRATORS CANNOT BE HELD RESPONSIBLE FOR USERS POSTS AND LINKS

... Copyright 2003-2999 Sonsivri.to ...
Powered by SMF 1.1.18 | SMF © 2006-2009, Simple Machines LLC | HarzeM Dilber MC