Author Topic: DIY SMPS killing MOSFETS  (Read 2675 times)

0 Members and 1 Guest are viewing this topic.

Offline wlamb02

  • Contributor
  • Posts: 8
« on: July 03, 2017, 07:53:55 am »
I am experimenting with resonant circuits and and the LLC converter. I built a tl494 variable frequency driving a half bridge through a gate drive transformer. I am only experimenting with this circuit so i havent bought an LLC controller. The frequency is manually tuned to resonant frequency. The mosfets die close to and at resonant frequency where power is maximum (about 200w). I have tvs diodes between gate source and drain source but the mosfets keep blowing. The drain source voltage only reaches a max of 175V, the mosfets are rated for 600V. During operation near resonant frequency they warm up but are never hot on the heat sink. So my main question is what more protection do I need to keep the mosfets from dying.

This is the simplified circuit diagram

This is the gate source waveform on the mosfets. It is about 47% duty cycle.

Picture of gate drive transformer

Picture of half bridge, I blew all of my 180v tvs diodes so I used 2 90v diodes in series

Picture of resonant circuit. Transformer has blue core, resonant inductor has green wire and resonant capacitors are the blue boxes. The transformer has an electric stovetop burner as load.

« Last Edit: July 03, 2017, 08:18:12 am by wlamb02 »

Online blueskull

  • Supporter
  • ****
  • Posts: 14010
  • Country: cn
  • Power Electronics Guy
« Reply #1 on: July 03, 2017, 08:39:47 am »
Your inductor and xfmr all seem to be a bit small at your frequency. Are they saturating at high current near resonant frequency?

Offline fourtytwo42

  • Frequent Contributor
  • **
  • Posts: 875
  • Country: gb
  • Interested in all things green/ECO
« Reply #2 on: July 03, 2017, 08:51:45 am »
Have you checked your deadtime under all conditions ? You may be suffering shoot through destruction.
The following users thanked this post: HackedFridgeMagnet

Offline IanMacdonald

  • Frequent Contributor
  • **
  • Posts: 944
  • Country: gb
    • IWR Consultancy
« Reply #3 on: July 03, 2017, 09:41:20 am »

Gate overdrive might be a reason. You need some series resistance in the gate drive otherwise an excess voltage will hit the SA12CA with a huge current. The leads on the SA12CA are too long. The Vgs only needs to exceed 30v or so for a few microseconds (maybe less?) to  knacker the gate insulation. With this much lead inductance a spike could still get past them.

Excess dissipation could also be a reason. TO220 packages can handle about 40W on a decent H/S. (Don't believe manufacturers who claim ridiculous figures like 90W) If your circuit is reasonably efficient that should be more than adequate for 200W output, but if the output gets reflected back, for example due to inductor saturation, then the FETs will last only a few seconds. You won't notice this on the H/S temp either as the actual chip will reach failure temperature before the H/S has time to warm up.  Maybe use TO247 or similar devices for testing as they can take a lot more punishment.

Oh, and this type of inverter should be driven at a fixed frequency with a variable duty cycle that is controlled by feedback. If you drive it with 100% duty cycle that's maybe putting more energy into the secondary than the load can absorb. In which case it has to go somewhere. It will mostly return to the supply rail by backbiasing the internal source-drain diodes of the FETs. This will create a circulating current -possibly ten or more times the supply draw- that will rapidly heat things up.

BTW the source-drain zeners are largely superfluous as the mid point voltage is constrained within the supply voltage range by the internal diodes of the FETs. 

Offline Floyo

  • Regular Contributor
  • *
  • Posts: 90
  • Country: nl
« Reply #4 on: July 03, 2017, 09:44:08 am »
Ok, quite a few things I see here that could be wrong.

- As mentioned, the magnetics. I have no idea what the cores are, but it all seems a bit too small for their size, especially it 50odd kHz.

- Working straight of the mains, not a good idea ;)

- Add some current limiting, even if its just a latching to off state. The melted and charred croc clips suggest prolonged over current events have occurred in the past.

- Why are the diodes (the SA180 & P600M) there? The mosfet reverse diodes are fine for the operation of the LLC.

- Gate drive, check both channels to look for the proper deadtimes. An LLC needs a long enough dead time to discharge the mosfet output capacitances for the switching node to transition
from rail to rail. If this time is too short and you fail to do so you no longer achieve ZVS and the main point of the LLC goes out of the window, also you now have switching losses in your fets heating them up.

- From your scope shot it seems that you are driving the gates negative by a whole lot, I'm not sure they'll like that much. You can get nice gate drivers that are isolated and can be bootstrapped easily,
no need to mess with gate drive transformers.

- Heatsink+Fan, this suggests your fets are running hot. A properly designed LLC of 200Watts can run the fets without any heatsinking at all, so again this indicates something is awry.

- You don't have a rectifier on the secondary side, you need that for the current to behave properly in the resonant tank. The single resistor replacement trick only works for the FHA design technique, in a real circuit you need the rectifier + cap there for the converter to operate properly. See attached screenshots.

Legend of the simulation screenshots:
Red+Teal, Mosfet gate signals.
Blue: switching node voltage, notice how this voltage transitions from high to low and vice versa in the time between the two gate signals (both fets off), this ZVS and very important for an LLC converter.
Green: resonant current, note how its anything but sinusoidal when only a resistor is place on the secondary.

Offline oldway

  • Super Contributor
  • ***
  • !
  • Posts: 2174
« Reply #5 on: July 03, 2017, 10:54:27 am »
Use a variac with insulation transformer, start with low input voltage (something like 10Vac), check transients and all gate and drain currents and voltages of Mosfet, if all is OK, increase input voltage slowly.

Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo