Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

5kW Interleaved buck design mosfet blow up

Status
Not open for further replies.

yagizhanugur

Junior Member level 3
Joined
Oct 20, 2014
Messages
29
Helped
1
Reputation
2
Reaction score
1
Trophy points
3
Location
Turkey
Activity points
239
Hello,

I am developing a 2 phase interleaved buck converter.

Input Voltage : ~ 120VDC
Output Voltage : 1V to 50VDC
Output Current : 100A Max for any voltage(1to50V)

I attached my circuit diagram.

My problems starts when output current rises about 50A.

For below 50A everthings looks perfect. After 50Amps one of the channels mosfet got burned.


Note : I tested my circuit with 50ohms gate resistor. Because of the high switching ripples.
 

Attachments

  • d1222.PNG
    d1222.PNG
    78 KB · Views: 112

Lack of information. What is switching frequency? Are you really operating SiC MOSFET SCTW90N65 with 12V unipolar gate voltage? What was the actual output voltage when transistor failed?

I guess, it's either an overvoltage or overtemperature problem. Did you measure the Vds waveform with a differential probe, or Vd and Vs separately? Also Vgs maximum ratings might be exceeded with unsuitable driver circuit layout.

Do you have a thermal camera to observe the transistor case top temperature?

A layout photo would be interesting.
 

Lack of information. What is switching frequency? Are you really operating SiC MOSFET SCTW90N65 with 12V unipolar gate voltage? What was the actual output voltage when transistor failed?

I guess, it's either an overvoltage or overtemperature problem. Did you measure the Vds waveform with a differential probe, or Vd and Vs separately? Also Vgs maximum ratings might be exceeded with unsuitable driver circuit layout.

Do you have a thermal camera to observe the transistor case top temperature?

A layout photo would be interesting.


Design parameters are listed below

Vin : 120VDC
Vout : 1V to 50V Digitally Adjustable
Iout : 100A (for 1Vto50V)
Vout(Ripple) : %5
Efficiency : > %85
Frequency : ~40khz(actually it is 42khz due to the mcu clock limitations)
Cooling : Oil Cooled


I do not think it is an overvoltage issue. When i run circuit for 40Amps it is working great and i turned of my device for cooling.

I turned on device for 50V / 70Amps it is blows up in 2-3 seconds.

I already have doubts about my driving circuit. What you suggest.

I attached image(Heatsink removed)

Thank you.
 

Attachments

  • 868c528e-62b2-4f9e-a3d9-fc2a53c70c50.jpg
    868c528e-62b2-4f9e-a3d9-fc2a53c70c50.jpg
    246 KB · Views: 166
  • bottom.PNG
    bottom.PNG
    234.3 KB · Views: 129
  • TOp.PNG
    TOp.PNG
    335.2 KB · Views: 130

did the buck diodes fail? most likely is insufficient heatsinking - or volt spikes at turn off - as no snubbers ...!!!!
 

did the buck diodes fail? most likely is insufficient heatsinking - or volt spikes at turn off - as no snubbers ...!!!!

No always one of the mosfet fails.

I do not run it for long time.

My plan is run it for 30-40 seconds with an heatsink for beging sure that it is working. After that i will put it on oil. It will be oil cooled.

It burned always in 3-4 seconds when current goes up 50Amps.(Only 1 channel SiC Mosfet)
 

Hi,

Saturating inductors?
Please give a link to the datasheet.

What capacitors do you use? Are they high frequency, low ESR types? Datasheet?

Klaus
--- Updated ---

The PCB layout is far away from being ideal.
Did you follow/check the current loops:
* when FET is ON
* when FET is OFF

Klaus
 
Last edited:

snubbers on the fets and diodes will definitely help a lot in many regards - esp in reducing RFI getting into your controller
 

This is one of those instances where you do require to have the proper test equipment.
Namely, a high speed differential probe. Or a pair. Plus a current probe.

You really have to monitor all of the Mosfet’s operating characteristics as the load current increases.
 

As a minimum - look at the Vds on the 2 mosfets and the current in the Tx pri at the same time ...
--- Updated ---

Oooh - just one more thing - in the photograph - it appears you have no input or o/p caps on the board? is this the case? if so it answers your initial question ...
 
Last edited:

As a minimum - look at the Vds on the 2 mosfets and the current in the Tx pri at the same time ...
--- Updated ---

Oooh - just one more thing - in the photograph - it appears you have no input or o/p caps on the board? is this the case? if so it answers your initial question ...


I have caps, but not shown in picture.

3xInput : https://www.digikey.com/product-detail/en/nichicon/UCY2C471MHD6/493-6807-ND/3664406
3xOutput : https://www.digikey.com/product-det...nic-components/EEU-FS1J102/P123411-ND/8567300


Hi,

Saturating inductors?
Please give a link to the datasheet.

What capacitors do you use? Are they high frequency, low ESR types? Datasheet?

Klaus
--- Updated ---

The PCB layout is far away from being ideal.
Did you follow/check the current loops:
* when FET is ON
* when FET is OFF

Klaus


Saturating inductors probably not my problem. If they saturates it should limit current flow from mosfet. But in my case current goes up and burn my fet.





My problem probably false gate driving. I will change my SiC Mosfet with PowerFets( IXFX100N65X2 ) (https://cdn.ozdisan.com/ETicaret_Dosya/587086_8203867.PDF)


What you think about low switching like 20khz. Now it was 40khz




snubbers on the fets and diodes will definitely help a lot in many regards - esp in reducing RFI getting into your controller


Yes i ve added snubber it reduces EMI for output and controller side but it did not stop fets burning.
 

bigger snubbers - what do your heatsinks and insulators look like ?
--- Updated ---

you caps are way too small and too far away from the fets, also each fet is 10nF to drive + dv/dt current requirements, your gate drive will need to be something extra special - got any oscillographs? ...
--- Updated ---

40kHz is fine - MUCH bigger snubbers - can you show your calcs for knowing saturation is not an issue ...?
--- Updated ---

as the current goes up - the chokes get nearer to saturation - then the current REALLY spikes up - then bang ...
--- Updated ---

you currently have 50A per choke + switching ripple - your calcs will be instructive for Bpk @ 55 amp in a choke ...
--- Updated ---

you caps are way too small and far away from the fets, also each fet is 10nF to drive + dv/dt current requirements, your gate drive will need to be something extra special - got any oscillographs? ...
 
Last edited:

Hi,

. If they saturates it should limit current flow
Why limited? The current will rapidly increase...

Why are the capacitors not shown?are they soldered on the PCB or not?

Klaus
 

the fets are 30m-ohm @ 25 deg C, at 55A for half the time ( 50V out ) the dissipation is 45 watts + switching losses per fet, as they heat up to 100 deg C junction the power goes up x1.8 to 81 watts per fet - if you try and cool them in oil they will go bang quite quickly.

Your other heat-sinking will likely not be up to the job either - it will allow the fets to get to 100 deg C junc - and then you need to shift 90 watts per fet - unlikely they have this sort of performance ( 0.5 deg C / watt to air ) maybe with a high flow noisy 24V fan, 30W or so ...

with a 50 ohm GD resistor, the switching losses will be ~ 10 - 20 watts on top of the above ... so you are in trouble really .... thermally speaking ...

as well as on the de-coupling front ( caps near fets ) if you turn off 55 amps you need to soak it up somewhere, i/C = dv/dt, so 55A / 1000uF = 55V / mS - so for an overshoot of 5V on the input side ( the pulsed side ) you need 11 x 1000uf = 11,000uF of really high quality cap ( low ESR/ESL )

then if you have 100nH of wiring in the way ( between caps and fets ) V/L = di/dt and you turn off the fet in 50nS, ( 55A -> 0A ) V/L = 1100 volts per uH, x 0.1uH ( 100nH loop ) = 110V spike at least - more if you turn off faster - or have more than 100nH total wiring loop L ... - hence big snubbers needed - to avoid turn off spikes ....

say you can tolerate 50V at turn off ( on top of Vout ) 50 / 55A = 1 ohm say, i/C = dv/dt, 55 / C = 50V / 100nS gives C = 110nF

So put 4E7 and 22nF across each fet FIVE TIMES, i.e. 5 of each of these snubbers in parallel across each fet - the power in each resistor ( at 42kHz) = 2.31 watts each - so use 3W R's and 250V film foil caps, having 5 in // reduces ESL and makes the snubbers work properly.

Snubbers needed on the o/p diodes too - leave you to calc those ....

good luck
 
Last edited:
What you think about low switching like 20khz. Now it was 40khz

Power electronics issues *are never solved* by trial and error or by what one wishes. You really need to know what is going on, and the only way is to probe the proper circuit nodes with the proper instrumentation.
Any "fix" that you apply, its performance can be compared to the previous circuit iteration.
Lastly, you have to verify that the particular "fix" is not causing a problem somewhere else.

If you don't have the equipment to do so, please tell us.
 

in my experience as a consulting power electronics engineer - very many companies and people design things by trial and error - especially medical power supply manufacturers ... and inverter makers ...
 

Hi,

EASY, although I think you are right, it's not my style of designing circuits.
I'm with Schmitt Trigger. The better quality, the higher reliability, the less EMI ... you get when you know what you're doing.

"Time to market" for my designs may be higher, the development cost may be higher, too. But my products usually are no consumer products, they mainly are used in big chemical plants. Thus my customers know about hidden cost like long time fails in control circuit (causing their production batch to fail. Decreasing their production yield), maintenance time, repair time, calibration time....
In the plants ... one production batch takes about 10 working days. A single 10 second fail of the control loop would mean to lose 40.000 Euros and drop the yearly yield by 3%.
So in the end it's the customer's decision which way to go...higher development cost or much much higher long time cost....

Klaus
 

How is it that the picture shows that caps have never been soldered to the board ...?
 

What type of PCB technology did you use?
Is it Bus bar integrated into PCB or Thick copper PCB?
Curious to know, I am also designing a circuit which can handle 6KW power.
If you can share a contact of PCB manufacturer, it will be great.
Thanks.
 

Most pcb houses can pate up to 3 Oz, if you need more conductive capability simply bolt on copper or brass strips ... e.g. 1 x 8mm
 

    Abhishekabs

    Points: 2
    Helpful Answer Positive Rating
Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top