Fuses and mosfets blowng up

[krish2487]

Hello yet again

i am trying to build a 36 to 400 v 100w push pull converter
i have posted my schematic here and confirmed that it should work if i dont physically mess things up

I have got the board made and assembled the components

However i keep blowing fuses and several times the mosfets too.



here is the data

The chip is a 3525 running at 100khz
the mosfets are IRF250

the primary of the trafo is a 18 center tapped
and the secondary has 121 turns in it

When i test the SMPS at 12 and 24 volts the Drain source waveforms are as they should usually be
except for ringing (which is anticipated because i first wanted to measure the ringing resonance freq before calculating the snubbers)

but when i run it from 36 v then i keep popping the fuses even before i can do anything

i tried increasing the fuse rating but then ended up blowing up the mosfets

i suspect that the transformer is saturating
but the transformer stays absolutely cool
though the mosfets become warm even at 12 v

the funniest (and sadly , even the saddest ) part is that for 100w at 36v the fuse rating should not exceed 3 A but i keep blowing even 6A fuses

Is it a design error or a PCB layout error

Please help

i am attaching the schematic for reference will upload the PCB layout shortly
along with pics of the assembled board

[krish2487]

PS forgot to specify
i took the Bm as 0.15T
a Kw factor of 0.4
current density of 3 A /cm^2
and wound the transformer on a EER28 core

my calculation show that the Area product EER28 core is three times more than what is actually required for my application so i assume that the core saturation due to Bm or Area product is not an issue here.

[Eva]

There are many things not shown in the schematic, including input filtering and the way in which a regulated supply voltage for the SG3525 is produced. The value of gate resistors is not specified either. Gate turn-off diodes are missing. In your circuit COMP pin should be tied to IN-. SYNC pin should be grounded.

MUR810 is a 100V diode, I suppose the ones actually being used are MUR8100.

Have you checked that the frequency is right?, that the SG3525 is not being disturbed?, that the regulation loop is not going unstable?

This is a quite difficult converter due to the high turns ratio. Parasitic winding capacitances and secondary side diode recovery may result in very high currents in the primary side. How is the transformer wound?

Please don't use JPEG images for schematics, they are quite unreadable, use GIF, PNG or even BMP...

btw: Glad to see someone doing calculations

[krish2487]

btw: Glad to see someone doing calculations

thank you thank you....



mea culpa

i just reposted an old schematic by mistake

i am not able to get a fairly good resolution screen capture of neither the schematic nor the layout
hence i am uploading the pdf files of both.

The gate resistors are 47 Ohms
and i was not able to procure MUR8100 so i went with MUR160

The sync pin is N/C
but the S/D pin goes to ground via a 100Ohms resistor (in the pdf it is shown as a 100K, my bad)

The gate drive signals are alright though not what you'd call good.
i ll uploading the pics shortly
the setup is at my workplace which i can capture and upload only tomorrow.

about the regulation loop
the output voltage with a supply of 12 and 24 v has not crossed 20-21 volts post the LC filter.
with 36 v i have not been able to check because i keep blowing up fuse after fuse.

though the output with 12 and 24 volts was expected as the transformer was designed keeping in mind a DC low line of 31 volts.

about the winding of the transformer
they say a pic says a thousand words
i d rather let the pictures only do the talking as i am afraid i will not do a good job of explaining the winding.



about the schematic

i cut off the drain pin and connected the transformer to the drain via the heatsink
i did not use a mica sheet or a silicon pad
i soldered the heatsink mounting pin to the squarish copper fill which is connected to the transformer start and end.

likewise with the RCD snubbers

Its not the perfect schematic nor the layout but i its my first attempt.

[Eva]

After a quick look to the layout:

- Input emi filter is required, at least two or three low ESR electrolytics >1000uF.

- You have made a huge ground loop on the layout of the SG3525, the circuit will easily disturb itself. The ground pin of the SG3525 needs a short direct connection to the sources of the MOSFET, bringing it through the voltage regulator first is a very bad idea.

- Turn off diodes very likely to be required.

- If 1kpf means 1nF, this is way too low a decoupling capacitance, particularly for regulator input (C7).

- 600V diodes are faster than 1000V ones, use them if they are enough, but there will probably be some voltage overshoot at the secondary side, unless the transformer is very good.

[TechGuy]

Originally Posted by Eva

After a quick look to the layout:

- Input emi filter is required, at least two or three low ESR electrolytics >1000uF.

- You have made a huge ground loop on the layout of the SG3525, the circuit will easily disturb itself. The ground pin of the SG3525 needs a short direct connection to the sources of the MOSFET, bringing it through the voltage regulator first is a very bad idea.

- Turn off diodes very likely to be required.

- If 1kpf means 1nF, this is way too low a decoupling capacitance, particularly for regulator input (C7).

- 600V diodes are faster than 1000V ones, use them if they are enough, but there will probably be some voltage overshoot at the secondary side, unless the transformer is very good.

I suspect you may run into problems with your linear regulator. The voltage drop from input 36V down to 12-15V is very high. and I believe it will run excessively hot and will probably fall in a short period. Consider using a simple switcher (ie Buck Regulator)

47 Ohms for your gate resistors seems excessfully high. This will increase switching losses as it takes much longer time for your mosfets to reach saturation (fully turn on). By decreasing the gate resistance you will reduce the heat load in your Mosfets.

[krish2487]

@eva

thank you
about the ground loop
back to the board again for me...

the transformer is decent enough
not much of a voltage overshoot to destroy the MUR160s

About the regulator section
i am not using the regulator section as of now for testing
i using the output from a RPS
and a huge tank capacitance at its output in parallel to a Ceramic disk cap

@techguy
i am actually giving the input to the regulator after using a 2:1 resistor divider network
most of the linear regulators have a 35 v maximum input voltage.

three batteries in series when fully charged will deliver about 13.8 * 3 =41.4 volts.


about the gate resistors.

thank you
i did not realise they were that high
i ll get them down and try.

i did have a doubt i have been meaning to post for a long time.
the datasheet of a mosfet gives us the input capacitance that the gate resistor has to charge right..

can i use the equation

i = C*dv/dt

with dv= Vc of the 3525
dt = the period of switching frequency (with some margin)

after calculating the current required

use V= I * R

where R = gate resistor
I= the current obtained from previous equation
V= Vc value

[luka]

you can always have start up circuit like in offline smps's, and one aux winding for sg, that should be easy!

[krish2487]

@luka

that was a mistake at the design level i made


thinking over now it seems to be the logical thing to do now..

[krish2487]

here are the pictures i have taken of the board and the waveforms

the PCB manufacturer gave me the board with the layout mirror imaged
which meant i had to mount the ICs on the copper side

here are the pics
Board Copperside


board topside


my setup

i used a 50 Ohms 5 A rheostat in series with the main power supply so as to limit the current and not blow out any more fuses
the rheo now reads 12 Ohms

Imageshack - picture006lq - Uploaded by krish2487

gate drives
Imageshack - picture007ws - Uploaded by krish2487

drain to source waveforms
Imageshack - picture008ey - Uploaded by krish2487
Imageshack - picture009oc - Uploaded by krish2487

the transformer output before the rectifiers

Imageshack - picture011x - Uploaded by krish2487

The gate waveforms were captured at 1:1 ration on the probe at 5v/div
and 5us/div

the Drain to source wareforms were at the same time base except the probe was in 10:1 mode at 5v/div mode

the transformer output was captured at 10vdiv at 10:1 mode

From the transformer i can see that both the mosfets are conducting at the same time

possibly leading to core saturation

When i tried to capture the drain to drain waveform
i heard a high frequency pitch from the mosfets which disappeared as soon as i removed the probe..

@eva
i modified the ckt as per your suggestion and u can see on baord copper side
i cut the ground track to the chip from the regulator and used a wire to connect the ground of the ship to the Mosfets source directly


Any suggestions....