Isues with 1kw SMPS at 125khz

[moro696]

Since a few days i came with an idea of using SG3525 and IR2110 as driver and some IRFP460 in half-bridge mode and using an ETD59 3F3 ferrite core rated at about 400-500khz.
Anyway, my operating frequency is somewhere at 127khz using an 5600 ohm value for SG`s RT and 1nf for CT, resulting somewhere at 250khz oscilating frequency for SG3525.
For primary and secondary turn ratio i tryied this deutch calculator http://schmidt-walter.eit.h-da.de/sm...gw_smps_e.html
My entry data there was:
Vin min 222Vdc
Vin max 345 Vdc
Output voltage 50V and output curent 20Amps, it showed me etd59 very suitable using 21 turns in primary and 10 turns in secondary.
My primary widing is composed of 10 litzwires, each one has 100x0.04sq mm.
The problem is that my power transistors keep heating very fast, last night i just blow to babyes
Here is the transfromer

This the test board with IR2110 and SG3525 mounted there


The big nasty thing is that i have spikes and i tryied diferent snuber values using a simple RC circuit paraled with a high voltage recovery diode such as PR1507 rated at about 700Volts. I tought it might cause a problem the primary widing with few or more turns, i tryied varying the turns from 18 turns to 27 turns, and stil no good results. I tryied using 220pf, 470pf, 560 pf ceramic capacitors @ 2kv in Drain-source coupled but stil nothing. And last night i tryied to couple a reaction in the mountage using some 50V zenner with an optocupler, with no reaction in the secondary i have about 75-80volts, with the reaction coupled i have 48-49volts constant. I tryied loading the source with two lightballs at 220Vac and 100W and it-s constant, but in primary the spikes are very nasty. I have tried searching google for some answers regarding snuber circuit design and found nothing mutch. Unfortunatly i dont have a scope at the moment, it was from my neighbour and last night he needed.... Unfortunatly i dont have mutch to do right now so that's why i am trying some advices from you

[luka]

my first ? would be, why so high freq. and so "bad" fets, when you have so big core, 1kw is nothing for this core, probably could use 20kHz to do it...

If fets are heating, I could think of.... cross conduction, too little dead time...
You need your RC across primary, R is what kills spikes, C stops low freq., switching freq.

Schematic could help a lot... also cap values that you use are probably useless, too small, and useless on fets...

And you won't be able to do much without scope

[moro696]

Originally Posted by luka

my first ? would be, why so high freq. and so "bad" fets, when you have so big core, 1kw is nothing for this core, probably could use 20kHz to do it...

Hi luka, i know the core can deliver nicely up to 3-4kw, in the future i want to make a half bridge using HGTG30N60A4D with that core, that's a 75amps rated transistor, that will be in the near future.

Sorry i forgot to post the schematic, it-s right here
http://webpix.ro/imagini/sursa.jpg
As you can see the snubers there i tryied diferent combinations of values
I even tryied coupling a 220,470,560 pf and 1nf and mounting a potentiometer, for me it was very quickly to set the R value. As i saw, on a higher value of R, the disipation increases.
Regarding the scope, don`t worry in a few days i will have`it back :P
I chose 125khz because i can achive a high randament... Somebody told me that IRFP460 is worse for that high frequence, i know it has more that 4nf gate capacity that is very nasty for rapid comutation times but i said to me "let`s start easy from the beging..." this is a continues experiment, i wanna start from zero and little by little to upgrade till i reach 3-4 kw using IGBT, but now i am stuck here with this heating.
What do you propose luka?

[moro696]

Ah, sorry for the second post, regarding cross conduction problem... i thinked about that and i put 10 ohms resistor in series with a 100ohm potentiometer between pin 7 and pin 5 on SG3525, that way i could vary the deadtime, when my fets blowed... tha value on that Deadtime resistor was maximum value... somewhere at 120ohms. I say that there was enogh big deadtime to prevent crossconduction a bit

[Eva]

Forget about point to point prototypes. A good PCB is a must.

Snubbers are usually intended for compensating the nasty effects of parasitic inductances (resonances), which in a point to point wired prototype are huge...

[moro696]

Originally Posted by Eva

Forget about point to point prototypes. A good PCB is a must.

Good point Eva, as i speak now, i am designing in protel a PCB. At 125khz it's a bit tricky to design the board, but there are many considerations that i have to count about... like the inductance traceings... i will use ground planes to "Embrace" the signal tracks and things like that.... I have reached this conclusion two last night when i saw thouse ugly ringins there.... I know there must be shot conections betwen the high frequency components... I will use smd a lot! But for know i need to make'it work as posible as i can....
As for my 1kw aplication... i think i will go for IFRP450 ... is mutch better than the 460's ones... I saw one one forum DjLeco sucking out from the K6 supply 2kw just using instead of irfp450 IRFP460... The last ones are quite hard to open "brutaly" due to the gate capacity in order to reach a optimum saturation and therebefore a "squarest" signal as posible...

[Jaac]

Also check the gate signals, if they are weak your Fet's are not in the saturation area and doing lots of heat dissipation.
But first things first post your schematic, and while doing that make your PCB like Eva mentioned.

[moro696]

Originally Posted by Jaac

Also check the gate signals, if they are weak your Fet's are not in the saturation area and doing lots of heat dissipation.

here is an oscilogram that shows the drive signal out from IR2110

Sorry for the clarity of the image, it`s shot with my nokia lol

[luka]

my , I don't think you'll need RC cross fets, also you could start at lower freq. since its much better to work there

C8 still seems to be very low, I think 10-22n should be better at start, you could use a bit higher voltage for IC's, 13-14.. but no need for now... and having even bigger capacitance for IR won't hurt too..

Apart from that, I think it looks fine, even how it is now... just loose point to point design, I made PCB and it workd first time powered on.

How big is your primary capacitance?

[wg_ski]

Originally Posted by moro696

here is an oscilogram that shows the drive signal out from IR2110

Just the drive signal coming out of the chip doesn't tell the whole story. If you get ringing across the source/emitter inductance, that voltage will subtract from the drive signal, causing desaturation. A PCB will fix a lot of this - but ultimately there is a limit to tolerable di/dt. IGBTs are normally rated for less current in hard switching than soft.