SMPS: startup transformer inrush current destroying mosfet

Hi folks, I have been building and learning about smps , now I made a half bridge off line smps and it has worked with various different switching mosfets, at first I started with lower power ones then with higher.

for those who havent seen my idea and schematic I will attach one on thsi thread.

the last devices I used are some vishay power mosfets Sihg47n60e.they are rated for 600 volts dc and 47 amps max at room temp.
now at first i connected only one transformer anbd the resulting rectifier and smoothing caps at the secondary of that transformer and it worked , previously i have pushed up to 1.2kW through this transformer even with such old and comparably low power mosfets as the IRFP460.
although I once blew these too when i connected a load which had a cold starting resistance of about 10 ohms. the maximum load at which I got the 1.2kw running at 160v DC was about 20 ohms and 15 ohms cold.

this last time I had these 47 amp rated mosfets and as I said i started the first transformer and they worked with a moderate output load, then I connected a second transformer which is almoust identical to the first one also has a rectifier and smoothing and the secondary and had a small load attached , as I attempted to start this configuration the mosfets went short and dead.

so heres my question (not good at heavy maths) how powerful the devices should be for me to choose , since the inrush startup current of any transformer is if i remeber correctly atleast 7 times the normal idle current and especially if the secondary has a load attached to it.

maybe i should go for IGBT ?I have seen the IR2110 being used to drive thme too.before i plugged in my setup i checked the rive signal with a scope and it was ok.also given that the single transformer action was working i think it was overcurrent at the startup which caused the device failure.

- - - Updated - - -



the schematic

Could it be that the magnetic remanence in your cores is way to close to saturation? In theory the soft start circuit should take care of that. Can you increase the 33uF cap to 100uF and see if this helps?

Otherwise.....This is one of those instances where you need to borrow or acquire a current probe, to measure the switching current during startup.

firs of all do you think it's okay to feed two transformers from a single two switch half bridge ? given that the mosfets or IGBT's are powerful enough for the job?

i could increase the startup capacitor on the SG3525, though even with this one the starting up is slow enough for me to see the voltage rise gradually over a sec or two on a simple multimeter at the output,
though i must say there is this weird thing that before the voltage reaches its steady state there is a burst of overvoltage, say at idle im having at about 170v DC on the secondary side but at startup the voltage reaches up to 210 volts for a brief brief moment.
I applied varistor surge protectors to limit that , and when i tried startup with them attached the mosfets blew up , they managed a short moment of work since i saw a small flash of light from my attached load bulb and then BOOM.
pretty much the same thing that happened when i used too big of a load at the secondary.

you think a ferrite material is capable of having considerable amounts of leftover magnetism? I'm asking because quite frankly i dont know.

oH and also one important thing i want to ask, can I use IGBT's instead of mosfets for my switches? quite frankly the ones I have been looking at are higher amp rated and cost less than similar rated mosfets.seems like a gain.
Also the gate capacitance is either the same of smaller as compared to the mosfets I have used so I assume my IR2110 will be able to drive them properly correct?
for example.
Fairchild FGH80N60FDTU or
Fairchild FGH60N60SFTU

Although you could drive a pair of transformers from the same bridge, at that power level good engineering practice would dictate separate powertrains.

IGBTs may be used, but they won't switch as fast as Mosfets. For higher power levels there are particular gate drive considerations you must take into account. IR has good app notes explaining how to achieve good IGBT drive.

Wheter or not the core is saturating, you DO REQUIRE the proper measurement instrumentation, in this case a current probe. Or a low valued resistor and a diff probe

well , would you say 50khz is high ? i will google what the typical IGBT speed is these days.
sadly i dont have advanced instrumentation as it costs alot, but i will se what i can get.

also what harm could be done by driving two transformers from the same bridge , as long as there isn't any overvoltage spikes occuring or overcurrent ?
now each transformer has a 2.2uF capacitor at its other float end.before i attached the second transformer the mosfets were driving the first one.then i added the second and BOOM, but when i was driving the second transformer separately from some other previous lower rated mosfets it was also working fine.
now i understand that core saturation could lead to loss of inductance and the primary becoming a flat wire capable of huge momentary currents upon the capacitor charging up.
but is there any way of estimating how high a normal inductive startup current could have been given two transformers and each having a 2.2uF capacitor.?
it feels like the startup current was simply too big for the mosfets, but it's weird because the drain current maximum was rated 47 amps with pulse current some 80 amps if im correct.so the current must have been huge.

you need to connect about 10k to gate to source of fets
you need xonnect zener from gate to source of fets.
You need to read the post here which tells of the problems with bootstrap hi side drives.
Are you in current mode or voltage mode?, if current mode then you should not have the series caps in

https://www.edaboard.com/threads/231947/

Also, if you are in voltage mode then the resonant frequency of the series cap and the leakage L should be much lower than the switching frequency.

Ok, treez , happy for your input but I would like you to explain a bit more , to be honest I don't know in depth about current mode or voltage mode differences.

I can surely add gate to ground resistor and I assume you mention the zener in so that the gate voltage doesnt go over a safe level for the gate so that it doesnt break down.But in this case I said that it was working fine with one transformer but it blew with two much like it blew with lower power mosfets driving a single transformer with a heavy low ohm load on the secondary , and it always happens on startup.
heres what I have learned from building this high power smps ,gate drive signal and stability is cruicial to everything from the very beginning.at first I had overvoltage and it killed two fets, then I used a voltage regulator and now that thing is stable.

damn I was logged in writing this and it logged me out and i lost half of what i wrote , i'll just try to do some short points here.
i read the thread you gave and what i gathered is that one engineer said that he didint like bootstrap half bridge driver IC's the others said they dont agree and pointed out that every gate driving circuit has it's own imperfections and the consensus is then what ?

but i have checked the gate drive signal with a scope both without voltage applied to D-S and when working (in theory they should't change much since the gate is isolated) and the signal is fine and I've also run the smps for some periods of time with maximum loads and everything is ok.the only two problems i have had with it is , first I had too high gate voltage , added a built in to-220 style regulator now is fine, and the other problem is mosfet destruction upon startup which happened first with too high load on the secondary and now with two transformers one of them having a medium load on the secondary but since both were started at the same time i guess they boh together were a too high load for the fets and the fets showed what they think of that...

i guess if i dont have a decent primary overcurrent protection circuit then i need fets or IGBT's that have a large enough reserve in them to account for startup current surge.
I mean high power IGBT's are pretty cheap , even cheaper than the mosfets I use , if they could handle my 50khz frequency, why not ?

I cant see any primary current sense in your circuit, this is a big problem, even if in voltage mode, you need an overcurrent sense.

Here is stuff on current mode vs voltage mode. The sims are in ltspice, (free download)

- - - Updated - - -

Half Bridge SMPS

Quoting from page 19 of the LM5037 datasheet…
*Designers familiar with these topologies recognize that conventionally, current mode control is used for push-pull and full-bridge topologies while voltage mode control is required for the half-bridge topology*

LM5037 datasheet
**broken link removed**


IGBT for high power and low frequency, no good here.

thanks for the files , i went through quite some stuff this evening.well i dont know how good of a chip the SG3525 is or so i think there are better ones and the feedback for the error amplifier only made things worse in my case.
the way i have it now it's an unregulated smps half bridge, which mostly gives me he advantage of lighter weigth , less room and less voltage sag under heavy load.even the unregulated voltage sag is fine beacuse i was driving a heater with it giving out 1.2kw and the voltage dropped some 10 volts but i built it in an amplifier box and it will drive two self made 500w amplifiers so i think the reserve is there.

surely i will build more advanced things in the future as i go along and learn, but with this one i just want it to work and have a bit of fail safe in it.
do you really think the IGBT's the way they are made now would be that bad for 45-50khz? I was reading on IGBT's right now and basically it says that summarizing you only get higher losses with them at higher frequencies , but the losses are like only about 10to 20watts more than a similar mosfet for cirucits till 50khz, so doesnt sound much of a deal.
frankly speaking the reason i would like o go IGBT or atleast try is their higher robustness in current , and without a working and sophisticated current sense cirucit in my primary I would have to buy a MOSFET that is like 2 preferably 3 times the current I need at full load since i want to be sure it will take the startup current. and such a mosfet costs on average some 3 times what my IGBT will cost.
can you point out any other things that would really show the IGBT as a bad choice for what's im doing here ?
if im correct for todays IGBT's the gate charge is mostly similar to mosfets on average so the IR2110 can drive happily the mosfet and well as the igbt?
also i see people using them in tesla coils and induction heaters some of them being very crude by design , definately no current sensing and a frequency of about 30 to 40khz and they seem to do the job just fine.

now i'm not trying to be a wise guy here or anything i'm just laying out my odds so to speak of.

I here that igbt is more robust yes, but if I were you I would just put in a current sense transformer.
Maybe you could get round it with a soft start to some extent though.

- - - Updated - - -

read post #8 of this about igbts
https://www.edaboard.com/threads/347362/

well the only problems are at sartup under heavy load anything else is fine.I have been thinking about increasing the soft sart capacitor on the SG3525 as you noted here.
my only question is wouldn't that be dangerous for the switches? if the square wave rises too slowly instead of mosfet or igbt being fully open at beginning and undergoing current surge they will be half open and still undergo quite some current creating excessive heating ?

actually I reviewed your schem in top post and see that you don't have feedback from the output. This is surely bad.....I was going to say you could instigate a slow rise of output voltage by putting a RC network across your upper feedback divider resistor, but now I see that you don't have a feedback divider. I would put feedback in if I were you. It is very bad without it.

well i erased th feedback from the schematic since i'm not using it in real life , heres why, when i was having the feedback it worked like this, if the voltage was somehow under the zener voltage and the octocoupler didint do any work back to the error amplifier everything was alright , then when it started doing stuff back the output voltage started fluctuating , like ti went up and down and if i remeber correctly it once blew up the mosfets even under moderate load upon startup.so i'd say the feedback is crap on this one atleast the way it was implemented , maybe i;m wrong.

heres the origina; schematic.

oh and also is IGBT better off if I start such a crude smps with no load at the secondary ?I wonder what additional parts i would need if any if I use IGBT's here on the switches

the feedback in above post is set up wrong, that's why you have problems.

I was going to say that the ir2110 shouldn't be used when bridge node transitions at >50v/ns (see ir2110 datasheet), but that is the least of your problems.
diyaudio.com forum have a lot of schematic of the type that you present here, maybe you can search for one there and see the feedback loop

ir2110 (page 2)
http://www.irf.com/product-info/datasheets/data/ir2110.pdf

in any offline smps, the bridge node (if its a bridge converter eg full or half bridge) always transition at >50v/ns so those bootstrap high side drivers are really bad.

hmm , can you please explain in a bit more common english what is the bridge node? my intuition tells me that since it has to do with the bridge and both the IR2110 , it may be the pin 5 on the IR2110?

oh also , there was this thing that with one of he last pairs of mosfets i had in my smps i added a choke on the + and - rails on the output at secondary and then i added two varistors rated at 100 volts each one from rail to the middle ground., my rails are about 90volts at idle so 100 volts is the safety margin , when i powered up the smps the mosfets went KABOOM wih sparks and fireworks , the thing is as I analyzed at startup the rails go a bit over 100 volts for a brief moment and the varistor probably pretty much was a low resisatnce load for that moment and since it's all at startup the mosfets didint survive the current that was drawn through them,

well basically this is what I need a robust device with lots of amps for reserve and i will try to insert a larger cap for the slower startup maybe that would also take care of the secondary rising too high.since its a half bridge the transformer basically works of half the rectified DC which is about 325/2=162/165 volts DC, my primary and secondary is almoust 1:1, since the secondary is +-85/90.


I would like to try the IGBT but i'm not sure which one to use and how high an amperage should I choose.
I will definately make my next smps more advanced and such jsut that this one is already in the box so to speak of and to make large modifications now would be alot of job, so im hoping to get by just by using some big reserve switches and some other minor mods.basically since its an amplifier that will be powered it doesnt need a large startup surge current , because at startup it will draw only as much as the bias current and maybe some small extra current that goes into the amplifier and that;s it, the reason im testing this smps with these large loads both at running and startup is I want to make sure that before I close the box that the smps will be strong enough to serve me under light and also under heavier loads that might arise from time to time.
anything other than that this smps would have worked with an amp already the way it was , since if not for startup surge when it was running it could easily heat me some water putting out some 155v DC into a 20ohm load which averages some 1.2KW of power, i mena even the wall socket wire got a bit hot as to how much amps my smps was drawing and it worked fine , even the mosfets were only moderately warm.

what you need is a npn/pnp buffer on the 2110 o/p's with a 2k2 from the bases to gnd, this ensures the mosfets are OFF at start-up

doesnt the IR2110 have this kind of buffer built in ? but the mosfet destruction has never happened under lighter loads.

Under heavy loads the midpoint of the fets can go below ground forcing a destructive current in the attached pin of the 2110, adding buffers can largely eliminate this effect...

adding buffers sounds like a good solution...possibly a better solution to page 16 of the attached, which explains this problem...Salvador12..the "bridge" node is the high dv/dt node

- - - Updated - - -

Also, the ray ridley doc tells of the woes of the bootstrap technique.....those ir2110 chips, and all others like them, cannot handle it if the bridge node transitions at >50v/ns....which it always does in an smps......so bootstrap high side drivers are bad news.....as ridley confirms...can the 50v/ns situation be mitigated by the pnp/npn stage , I am not so sure