loading

SMPS H-Bridge trouble

Hi, 

I've built a 1kW SMPS designed to take a 350V unsmoothed DC input and convert it to 3kV @ 300mA. The circuit is basically Uzzors2k's design seen here minus the overcurrent protection and smoothing capacitor (for now), however I accidentally wired the GDT outputs backwards so the gate resistors (33ohm 2W) are on the source. I have used alternate MOSFETs the latest attempt with FDPF16N50T's.

I've had various problems with the circuit, primarily the MOSFET's exploding seemingly from overvoltage (evidence of flashover in some cases despite no evidence of secondary leakage (i dismantled the transformer to check)). During my last test I had a flashover between the drain and source pins of one of the FET's, in addition all of the gate resistors were smoking hot after, seems the gates had all been shorted (first time this happened, other occurrences were from case to ground). This happened despite the variac only being turned up to 40% which should be well well below the pin spacings maximum voltage tolerance.

I have checked the drive signals, they are quite nice square waves at about +-15V with some oscillation on their rise.

I have built SMPS's before, one is currently working with similar MOSFETs using a half bridge at 300W.

Does anyone have any suggestions as to why these troubles are occurring or have any suggestions to fix it or have any alternative circuits that may be less troublesome (is a h-bridge easier to manage?)?

Thanks,

Stephen


I​ just gave it another test with a H-bridge instead and the source of one of the FET's blew so it appears that there is just FET failure and maybe not arcing between drain and source.

sort by: active | newest | oldest
The MadScientist (author) 4 months ago

I've found the problem. The output rectifier on the transformer was incapable of high frequency hence was shorting due to it's long switching time.......................... .................................................................................................................................

Thanks for your help.

iceng4 months ago

Polarity needs changing for the main bridge, see the 1:1:1:1:1 xfmr orientation...

There is no provision of dead time between switch cross-over and no inverse parallel diodes to take over the intrinsic inverse diodes for reactive load continuation without stressing the power fets.

GP_inverter2.GIF
The MadScientist (author)  iceng4 months ago

The TL494 IC has a 5% dead time inbuilt so there is time for the fets to switch off before the others switch on (or does there need to be more?), and the fets i've used have inbuilt reverse diodes or are these inadequate?

And I dont quite know what you mean, what's wrong with the xfmr orientation?

Dead Time test, Disconnect the Top-Left leg at the positive Drain and run the inverter (might have to use a signal generator) now if the dead time is correct the Right leg top and bottom should be cool NOT EVEN WARM.. Repeat Right leg disconnected and see if the Left leg is cool.

--------------------------------------------------------------------

The intrinsic NFET diode ( purple ) is a very fast diode, But no guts for extended time reactive current. You put most any diode in parallel to take the heavy current away from the intrinsic NFET diode.

-----------------------------------------------------------------------

You didn't notice what I changed.. Look at the dots on the 11111 transformer and consider the dot a [+ voltage] for step_1 this turns On TOP-LEFT and BOT-RIGHT for a Right_load_current.

And no signal for step_2..

Now consider the dot a [- voltage] for step_3 this turns On Top-R and Bot-L for a Left_load_current...

No signal for step_4.... And repeat back to step_1.

GP_inverter2.GIF
The MadScientist (author)  iceng4 months ago

Hi,

I couldn't do your dead time test as at the moment I've switched to a H-bridge as i'm running out of fets, however i did use the feedback to tune down the %DC and I got the same results.

As for the GDT orientation I didnt notice the error on the original schematic but I didnt reproduce that error when building it.

So far I've tested it again and seem to be arriving to a problem in which the gates are seeing voltage from the mains supply, when I turn the variac up to as little as 30V i start to hear buzzing coming from the circuitry and then failure. Nearly all failures have resulted in damaged or blown gate resistors but I am uncertain as to whether this is happening during or post failure.

As for the internal diode problem, I am running another SMPS using a H-bridge with very similar FET's and I've never had an issue with this (300W), however I will test the theory when I get some proper diodes in.

Ok, I realised I have some power diodes so I rebuilt the H-bridge this time using HGTG27N120BN IGBT's and RHRG75120 diodes. This time I could turn the variac up a little higher but it still conked out, the IGBT's and caps got quite warm and the IGBT's failed whereas the diodes were still room temp.

I was also hearing quite a bit of 50Hz buzzing and the variac did internally arc a little so there seems to be a lot of current flowing (below 10A because my circuit is fused).

The MadScientist (author)  iceng4 months ago

I have just tested the circuit again and i took a photo of the drive signal at the FET gate here it is.

FET Gate.jpg
Downunder35m4 months ago

Everytime I produce such expensive smoke signal it comes down to how the gate is driven.
Mosfets require fast and strong signals.
Whenever possible I drive them directly but if I have to use a transformer for it than I make a few more turns for the primary to make sure the gate is getting enough amps and a stronger signal to work with.
It also depends on the frequencies in use, some gate drive solutions only work really well within a certain frequency range - if it drops too low or goes too high the driver fails and ususally one mosfet hangs.
This causes the other one switch and both go up in smoke.
The circuit you use seems to be very dependend on the right parts with little room for error, so a h-bridge in the normal design would certainly create less troubles finding a fault or source of problems.

The MadScientist (author)  Downunder35m4 months ago

I think I was having those problems initially and certainty have in the past but I've made some adjustments and corrected them, my gate signal now looks like this one @ 66kHz (very similar) http://www.loneoceans.com/labs/drsstc3/gdtoutput.j...

Now the problem is that I cant see how the drain and source flashed over at only 140V.

Possibly the intrinsic reactive diodes were damaged and when reactive current has no place to go, it causes a huge voltage to build up, arcing across your NFET pins...