Inverter: to anti-parallel diode or not? / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Inverter: to anti-parallel diode or not?

1 2 next

Move Thread

LAN_403

hen918

Tue Sept 01 2015, 09:22PM

Registered Member #11591 Joined: Wed Mar 20 2013, 08:20PM
Location: UK
Posts: 556

I'm in the process of building a 3kW inverter based on a switched mode design, and have found a problem: On the output H-Bridge, which, I have decided, will be composed of 600/650V MOSFETs, are extra flyback diodes required or not?
Not including them means that if the drain diode goes into forward bias for whatever reason, a lot of energy will be expended recovering, but if they never will go forward biased, extra diodes would be a waste of space and money!
The output will be filtered with a low pass LC filter and that's pretty much it. The inverter will be expected to do anything including driving motors, etc...

Hazmatt_(The Underdog)

Wed Sept 02 2015, 12:13AM

Registered Member #135 Joined: Sat Feb 11 2006, 12:06AM
Location: Anywhere is fine
Posts: 1735

I'm designing a small 20W inverter and we have blown up several mos because the body diode conduction issue. The Mos reverse voltage ratings were in the 20-30V range, and we're operating with a supply of 12V, which is fine until you have a ringdown, which comprimises the diodes eventually.
So we bridged the mos with some ultrafast diodes to take care of any needed clamping, and we should be fine.

Check the reverse voltage specifications of the mos. If it is much greater then 2x Vin, you should be fine, if not, I would add the diodes, mos gets expensive after a while, and if you design without diodes and you find you need them, you're going to have to make space somewhere, and your design will be compromised.

Prototype it and see if it fails under load, if you have extra mos. You could also gradually increase the input voltage with a variac and see how much ringing you have on the drains on your mos. If it looks excessive, add a RCD clamp.

Russell Haley

Wed Sept 02 2015, 11:07AM

Registered Member #2478 Joined: Mon Nov 23 2009, 03:24AM
Location: Texas A&M University
Posts: 47

MOSFETs are used in reverse conduction mode for synchronous rectification. They can't block reverse current, but they can pass it through the channel with substantially less voltage drop than across a diode. Or the low voltage ones can, anyway. Those high voltage FETs tend to have high R_ds(on), so you should check to see if the voltage drop at your target operating current will be enough to bias the body diode into conduction.

To avoid using the diodes this way, you'd have to drive your bridge such that each leg always has one transistor turned on. This way, current forced into or out of a leg always goes through a FET channel instead of a diode (provided I*R_ds(on) is less than the diode forward voltage).

Sulaiman

Wed Sept 02 2015, 12:58PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

I would design-in the diodes,
the cost and space are not that great compared to overall cost and space,
and if you only save one set of mosfets (and commonly driver circuitry also) it is worth it.

Diodes will dissipate some power due to charging/discharging of diode junction capacitance
but RC snubbers will dissipate more.

hen918

Wed Sept 02 2015, 02:15PM

Registered Member #11591 Joined: Wed Mar 20 2013, 08:20PM
Location: UK
Posts: 556

I seem to have found the perfect MOSFET if my wallet can stand it

DerAlbi

Wed Sept 02 2015, 09:45PM

Registered Member #2906 Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 727

Awwww use them! <3
But pay attention: Gate Breakdown is 22V.. and Rsdon is specified at 18V.... smells a bit like it was pushed to the limits to show those values.

Btw: implementing extra Diodes sucks more than you might imagine right now. The layout becomes more complex and the additiontal current paths make it harder to implement a proper low inductance design...

BigBad

Thu Sept 03 2015, 12:55AM

Registered Member #2529 Joined: Thu Dec 10 2009, 02:43AM
Location:
Posts: 600

Well, I've actually built a 3kW inverter, so I have first hand experience, and I blew up several MOSFETs doing it and read lots of papers on how to not do that, and some of them even told me stuff that worked.

In my experience if the MOSFETs have body diodes then external diodes aren't necessary and probably do nothing very useful.

The most important thing by far is board layout; the switching transients and stray inductances and MOSFET capacitances within the board itself are the long pole in the design; either they will blow up your control circuit/MOSFET or they won't. If they don't then the circuit will almost certainly work. External diodes usually don't control switching transients at all adequately, and nor do the body diodes.

The way to control those is board layout (earth and power planes, and keeping everything around the MOSFET ridiculously short and planar) and put a resistor on the drain side of the lower MOSFET between it and the control circuit to help damp out the oscillations. The usefulness of gate resistors is wayyyy oversold, because they don't control the oscillations very well and reduce efficiency a lot, but probably should be used.

Hazmatt_(The Underdog)

Thu Sept 03 2015, 02:44AM

Registered Member #135 Joined: Sat Feb 11 2006, 12:06AM
Location: Anywhere is fine
Posts: 1735

It's up to you what you want to do, you're the designer after all. I only mention the diodes because we're blowing up 50A mos with a 20W inverter. The Gate-Source is clamped, and after a while, poof!

The only reason the mos could be dying is because the ringing is exceeding the voltage handling of the body diodes on our little 12V circuit.

With the UFS diodes in, we shouldn't have that problem ever again, and it wasn't that big of a deal to add them in.

BigBad

Thu Sept 03 2015, 04:07AM

Registered Member #2529 Joined: Thu Dec 10 2009, 02:43AM
Location:
Posts: 600

No, there's multiple failure modes when there's ringing. I lost quite a few MOSFETs and one of the big failure modes seemed to be high frequency ringing causing an undervoltage on the middle of the bridge which took out the controller and then that in turn blew the MOSFET by failing to control the gate voltage correctly.

Another failure mode is that the middle of the bridge can ring far above the power rails- the Q factor can be a few, so it goes up and can blow things that way (that one is rarer- the head room on most of the components is usually ample).

You'd think that diodes would be a big help, but the ringing can be at quite high frequency, around 20mhz or more, so that the diodes don't have a chance to switch off, so they're of little help really.

Adding very fast external diodes could potentially help, but for my circuit when I tried it, it didn't seem to help at all, it tends to just oscillate at a higher frequency instead.

The sovereign cure is damping- a ~10 ohm resistance on the ground side of the bridge to the controller, damps the ringing right out.

I also found that low voltage bridges are comparatively easy, it's what happens when you wind the voltage up that separates the men from the boys. My bridge is currently running at 300 odd volts successfully.

DerAlbi

Thu Sept 03 2015, 10:42AM

Registered Member #2906 Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 727

Yeah, driving big Mosfets is a bitch by its own.
There is a tradeof to be made with the gate resistors: on the one hand side it must dampen the strayinductance of the gate-source loop and on the other hand it must be as small as possible to not weaken the gatedrive so that drain transients and stuff influence the gate in a ridicules way through Cdg.

I think BigBads experience is one of the worse out here. The Gate-Ringing can partly be a measurement problem (you cant use the long ground lead on your probe for sure. you just reveive any shit with it, but not the actual gate waveform). The gate-soruce loop where the gate current goes through has to be as narrow as possible (to reduce L). Big packages cause a problem inherently through the spacing between gate and source pin alone.
As long as you dont measure incorrectly the placement of the gate resistor is not critical it can be done in the source or gate connection to the mosfet. Its just an LCR-(Series)Circuit in the end. Bring the L down and the R up and you control the quality factor.
What you also want to do is to have a seperate current path to the soruce for the gate drive. You should NOT reuse the high current path/plane leading to the soruce pin and connecting your driver through it. Idealy you route a crude differential pair to gate and source seperately, keeping power and drive well apart and just connect those directly at the source. you can even use the gate resistor as jumper and make that differential pair twisted. Ive done that (i put Rgate/2 in the gate line and Rgate/2 in the source line to have multiple jumpers) and had really good waveforms (measured without ground lead directly at the mosfet pins - anywhere else the waveform was crappy)
Yes, it crys for a 4 layer PCB. But thats more fun routing anyway. ( ireally recommend it)

Edit: i am not offending BigBad that he is incabable of measuring the waveforms.. its just such a common failure to see ghosts on the oscilloscope just because you used the ground lead and additionally mabe picked a bad location to hook it up to the board.

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob