DC voltage startup surge in my SMPS

[Salvador]

I recently built a rather crude SMPS using two very powerful IGBT's and a SG3525 and IR2110 driving circuitry.
It's a half bridge and even though it's unregulated the output voltage is quite stable , dropping only few volts as the power output goes up to 1KW and some more volts after that.but since i will be using it as a power supply for a power amplifier this voltage drop is fine with me because it was measured with a constant resistive load.

now here's the problem.two IGBT's in half bridge are driving two separate transformers with the primaries in parallel.
at the very powerup moment the secondary output voltage goes all the way up to 300 volts for a brief moment and only then goes to its intended 180v DC.
it happens on both secondaries.
the question is what could I use to limit this ?

if it were AC , i'd simply use a triac but I guess on DC voltage surges and spikes are harder to stop?
A varistor is built for voltage surges but i think it's not meant to be used as a regular voltage surge stopper , only for extreme cases and it also puts a lot of load onto the power supply.

basically the only real and simple way of getting rid of this for now to my mind would be to use a startup delay circuit with a relay which switches the power from the psu to the amplifier board with a delay.

 

[Windadct]

Possibly an oscillation - can you get a scope on it? Not much info to go on otherwise.

 

[jim hardy]

Is this a closed loop regulator that reaches back and varies pulse width ?
Sounds like it overshoots on startup..

What value soft start capacitor do you have on pin 8?

The datasheet seems sparse on how to use soft start
have you looked for a ST application note ?

 

[Salvador]

Ok I have attached the schematic , the way I have it built as it differs from the original.I made the differences because with mosfets i couldn't get enough power out for a cheap price and with these fast IGBT's I have pulled out over 1KW from each of the two transformers which is very good for me.Also I eliminated the feedback octocoupler because with that the output voltage wasnt stable and the overshoot was still there.

basically it's a simple schematic and once its started up it wokrs just good but as I said the startup voltage is a killer.I think it actually has to do with the transformers themselves because one of the secondaries jumps to 600v on my multimetter for a brief moment the other one hits 300v DC.and then they both fall to 177-180v DC and keep that way as long as the power is plugged in the socket.

I measured with scope the gate driving waveform , and it's pretty good.it's about 15v DC since I have an 18volt regulator on the IC's power supply.I used 15v egulator previously with mosfets but now with IGBT's i kinda though going a bit higher would be good as I read they have bit higher gate voltage for the device to be fully open.
the gate drive square wave goes something like this.at first it sets up its peak value rather quickly but it's duty cycle is very short , then after the peak value has risen the duty cycle gets wider but it happens slower than the rise time to the peak value.at weekend when I will have a normal camera I could try to film it.

When I was building this monster I was running another thread here and the folks there weren't quite happy with my way of doing things hopefully we could avoid that here.
Also I have to admit that that transformers were rather salvage types than bought specially for the purpose of this.but the core material is 3c85.

there is one thing that some folks over on the net told me which could maybe cause problems.one of the transformers is from an old philips TV, the one with the 3c85 marking on it and even though it's a cube box shape and doesn't resemble the typical tv high voltage tube transformer they said it could have an airgap in the middle leg which cannot be seen because the bobbin is on it glued to it.

 

[jim hardy]

i like your IGBT's , have seen industrial induction heaters made with bipolars.

a couple get acquainted questions and some speculations -

are you over-voltage-ing those 63 volt output capacitors ?

is your dmm autoranging ?
How long does that high voltage condition persist ? Do you see the high voltage with an analog meter ?

That your voltage settles around 180 with presumably max duty cycle suggests that's what the transformer turns ratio dictates for the supply voltage you have
which suggests something is causing a short lived voltage gain in your power circuit
and whatever it is isn't quite the same for the two output circuits
So it begs the question
how can a transformer change its turns ratio ?
Well, it can't

a couple possibilties for thought ?

a transformer's inductance changes when it saturates
but with series capacitance it's not immediately clear how that could happen
...
series resonant LC circuit has voltage gain, and L of your transformers is maybe unknown ? And not identical ?
Does that pwm modulator sweep its frequency during startup?

Might you try changing the value of that series 1uf capacitor beween transformer and "float" ? That's just an experiment to see if the circuit is choosy about that part...interesting project ... i was unfamiliar with those parts - it can't be that easy...

there exist application notes about choosing inductors for smps's , Coilcraft and Mag-inc come to mind

We learn fastest by doing .
Keep poking at it and it'll give up a clue
.

old jim

 

[Salvador]

ups sorry i didint change all of the values in the circuit as its not mine but i have modyfied it.
the secondary smoothing capacitors are actually rated 200v and they are across each end of the secondary tied to middle , so across the whole secondary the caps can stand up to 400v which is plenty for a 180v DC output , I think.

as for the IGBT's I feel sad I didint start using them from the very beginning, as these days they are fast and have literally the same voltage drop as comparable mosfets so them being more robust and far cheaper I find them very good for high power power supplies.
I mean these cost roughly 3 usd a piece, similar power mosfets would cost me 12 and upwards.

Well I don't have an analog meter at the moment but the digital one rises to 300/600v for each one of the transformers at startup but it only happens for like a split second the numbers then jump back to 180 and about 170 under medium up to 1kw load.
Also the primary series capacitor is actually 2.2uF for both transformers.

I'm not sure what could cause the voltage rise at startup beacuse judging by the IGBT gate waveform there should be no such rise also considering this is half bridge whihc means that the primary voltage is only half that of mains rectified which is 325v getting 600v at secondary is quite a thing.
Well as you said a transformer can't change its turns ratio but it can change its flux through those turns and maybe this is the way the core responds to a sudden rapid increase in its flux from about zero under no load conditions.

Maybe I should try some snubber circuitry across the secondary or some damping across it ?the only other option I see now is to get myself a startup delay circuit with a relay.

 

[jim hardy]

Well I don't have an analog meter at the moment but the digital one rises to 300/600v for each one of the transformers

why the difference ?

 

[Salvador]

well that I would love to find out , as much as I can tell I think it's because they are different , one transformer has one type core material the other one has another their similar but not the exact same , or maybe as some folks suggested one of the transformers was built for a specific device since it came from a old philips tv and maybe it has an airgap in its middle leg on which the bobbin sits although I'm not sure why would it have one.
anyway i have this voltage rise both with using just one transformers or two in parallel.

 

[jim hardy]

so both transformers do it...

i tried to post last night but system locked up...

Can you set your meter for AC volts
and measure voltage across a transformer primary during startup
then across that series capacitor between transformer and 'float' also during startup ?
That might give a clue about resonance between transformer's inductance and that capacitor

i found a blog someplace where a guy claims he corrected a similar problem by making Vref time constant longer than soft start time constant...
as i said i don't know that device - is its frequency steady during start ?

old jim

 

[Salvador]

well yeah one transformer jumps to 600v the other one reaches 300 at secondary at startup.

Ok I will try to measure tomorrow if i will have time.as for your question if I assume correctly that when you are referring to the SG3525AN Ic as the " device" then I can tell you that if it doesn't have anything attached to its error amplifier feedback then the startup frequency is fixed and its the running frequency so it doesn't change , the startup capacitor basically does what i saw in the scope screen, the square wave firstly rises to its peak value and then the duty cycle spreads until it reaches its operating duty cycle.

thats a nice idea to measure the voltages across primary and secondary maybe indeed there is some resonance , also because each transformers core material differs a bit that could explain the different voltage levels I get since the core materials parameters affect the resonance.
anyway I assume the primary voltage doesn't get as high because these IGBT's have a max CE voltage of 600v if I recall correctly and the previous mosfets I have used had 500 to 600v on average D-S so they would have been destroyed with voltage spikes appearing regulary at turn on.

 

[jim hardy]

if I assume correctly that when you are referring to the SG3525AN Ic as the " device"

yes, i was

One more thing to eliminate

i wonder what your startup sequence looks like...

i'd try temporarily powering the low voltage supply from upstream side of that NTC thermistor instead of downstream

that'll assure the pwm and driver are ready before high voltage appears

low voltage supply might not draw enough current to fully heat the thermistor
so when high voltage kicks on it may cause a dip in your 12 volt supply while thermistor heats the rest of way up
causing some sort of flukey restart

a short test moving the one wire as above would rule out that speculation

that's how projects go -
first you get it sort of working
then you get it working better
then you get it working great
and have learned a lot,
experience being the best teacher (even though it's painful)

that's a really intriguing chipset - thanks for the introduction.

old jim

 

[Salvador]

Hmm , I have to apologize to you Jim for one more thing , my input is not as shown in the schematic with a thermistor, its simply a mains conection with fuses and a choke in case some harmonics would flow backwards into the mains.
so both the small transformer for IC power supply and the large rectifier and smoothing caps have a common hard wired connection to mains so this is not the problem.
I will try to measure the voltages as you suggested today.

 

[Salvador]

So haven't been here for a while , I did some measurments just a few hours ago and something interesting happened.
At first when I plugged in the smps one channel spiked to a little over 1000v the other close to 300, then when i turned then smps on and off repeatedly , suddenly the voltages stopped spiking , everytime now I switched the power supply on the voltage reached 180v on one channel and with a slower climb the same level on the other one.

Also at first when the spikes still occurred i measured not across the whole secondary but across each end of the secondary to middle or center tap and there i didnt see anything unusual at all , the voltage climbed steady and evenly up to 90v DC.
I let the smps sit there for a while and repeated the same thing once again , one channel is perfectly fine and rises to only 180v DC with no voltage surges the other still wants to jump to 230v at first and then settles back to 180.

since I'm getting these changes couldn't it be that it has to do with the series primary capacitors somehow changing or adjusting themselves?
I tried to measure the primary voltage but my cheap multimetter doesn't show anything , I assume it can't read high frequency impulses and then interpret them as AC voltage.

 

[Tom.G]

1)For the initial voltage spike, probably the 318V supply is drooping. No load it should be about 340V. If the line frequency ripple is too high, increase the caps values. Look at it with a scope to see if it fully recharges to around 330V or so each half cycle. If not, you may want to reduce either the capacitor ESR or the supply/rectifier impedance, or both. I wouldn't be suprised to see 20 or 30 amp current peaks from the power line. The other option is to use voltage feedback regulation with the SG3525.

2) The spikes are now occurring on only one transformer
. 2a) therefore What's Different between them?
.. 2a1) the R's and C's on the primary and the rectifier and filter on the secondary

The snubber caps on the transformers, 470pf @400V seem way under rated for the voltage.
In ancient history when vacuum tube car radios were powered with a "vibrator" (essentially a relay set up to oscillate) feeding 12VDC to a transformer, the snubber capacitor on the 200V secondary was rated 1600V. Perhaps your snubbers are/have broken down. Try using some rated at a few kilovolts.

I'm also a little concerned about the 400V rating of the caps on the primaries. Rule-of-thumb says "100% safety factor", maybe 50% for hobbyist / first engineering prototype until the correct parts are delivered. When you take into account the possibility of 20% high line voltage, you are already at 400V for those 1uf caps and the 680uF caps in the 318V supply.

Another concern is the schematic shows only a diode connected to the IGBT gates. This does not allow the gate to charge during turn-on because the diode is back-biased. Ultimely this leads to very slow switching, high dissipation and failure. That's why the 4.7 ohm resistors were there. The gates, being effectively a 0.005uF cap, must have a DC path.

If your output rectifiers suffer from a short lifetime, try putting RC snubbers across the secondaries. Or a 0.001uF across each diode.

Careful about using a 'scope on the transformer primaries or secondaries and on the IGBT collectors. Use a high voltage probe. When a component fails the switching spikes could easily fry the scope!

(edit: added ref. to 680uF caps in paragraph 4)