Is there a better way for a turn-off snubber for a MOSFET

[Electr0nicus]

Hey

I have a question concerning a MOSFET PWM stage i have designed. It switches 48V with 60A max on the LOW-Side. The prototypes are currently in permanent use (14h a day for 6 days a week) in a production environment for testing, without any problems with the electronics so far. We are currently updating and reworking all the schematics and PCB-layout for series manufacturing.
As I said, the PWM stage works fine, even with maximum load and no problem so far, but i wanted to deal with the turn-off ringing of the MOSFET, before finalizing the series production engineering.
Up to now, the MOSFETs are not equipped with any snubber circuit or someting of that sort. So i have a huge turn-off spike, which is clamped by the avalance breakdown of the Body diode of the MOSFET. After that inital pulse, there is a damped oscillation until the voltage stabilizes. As the MOSFETs are avalance rated this doesn't hurt them oviously.
I think I understand the mechanisms at play here, and therefore the ringing makes complete sense to me. But my paricular question concerns the design of a snubber circuit. I have made a LTSpice simulation ,which is really close to the actually measured ringing with the oscilloscope (unfortnately i'm at home right now and can't share the actual scope screenshots, but they are nearly identical with the simulation).

What i have tried:

• Dimensioning of a R-C snubber directly across the Drain/Source with the tutorial found here: https://www.digikey.com/en/articles/resistor-capacitor-rc-snubber-design-for-power-switches

This yields good results in dampening the ringing period of the turn-off Vds signal. The huge initial spike still remains

• Directly connecting a 10µF ceramic capacitor across Drain/Source

Works perfectly as the signal waveform is concerned, everything is smooth and flat. But the capacitor whines, that it isn't funny. Maybe that wouldn't be good for long term reliability

• Combining the RC- snubber from 1. with a parallel TVS diode

Works very well, as the TVS clamps the inital pulse and the R-C snubber deals with the ringing at the end.


So far so good, but I'm not really satisfied with that result. Is there a version of snubber which i can try, which can archieve a perfect flat response with the minimum of components or have i tried every possibiliy and I only have to decide which one to use? I've tried different gate resistors, but the big spike at the beginng only started to slowly vanish at 1kOhms and above and the MOSFETs got really hot. So that's no solution at all. I want to mention that the switching frequency is very low, only 15-16Hz.

I'm interested what you all think.

Cheers,
Gregor[/list]

[langwadt]

why not a flyback diode (or fet), do you need fast turn off?

[JohnG]

Here is a fairly good article on snubber types and operation: https://www.ti.com/seclit/an/slup100/slup100.pdf

Note that the fancy "non-dissipative" snubbers made a lot more sense when you were using TO-3 packages or even stud-mount transistor and twisted pair connections. With modern surface mount parts and PCB construction, it is really hard to make the inductance of the snubber less than that of the power commutation loops. The latter is a must for the snubber to do anything useful. On the other hand, modern layout techniques reduce the need for snubbers in the first place, unless you have a transformer with it's associated leakage inductance.

RCD snubbers are fairly nice, and sometimes you can even have the cap discharge slowly to the output and get a little of your snubber losses back. Again, the key to making them work in practice is to make sure the inductance of the snubber loop (the entire closed loop where the current flows when you are clamping the spike) is much less than the stray inductance causing the spike in the first place.

John

[T3sl4co1l]

See the conversation here:
https://www.eevblog.com/forum/projects/preferred-diode-type-for-high-current-flyback-diode-(1uh-120a)/msg4427359/#msg4427359

Don't worry about flatness, just get the power away from the MOSFET.  They're rated for avalanche once, maybe even a small amount [of energy] repetitively (did you check that your load has the same peak current or total energy as the device is rated for?), but most types as far as I know will eventually fail under repetitive avalanche.  TVS will not, its only concern is overheating.

Whether you want some rounding-off of the corners, or to dampen that ringing, is up to you.  An RC snubber will do fine for that.  Gate drive can also be slowed down, considerably (do you need a 10 ohm gate resistor here?), to similar effect.

The above explains why a clamp diode is suboptimal.

Tim

[jonpaul]

Depending on the topology, various techniques can absorb or regenerate the stored magnetic energy on turn off.

See Vincareli patent for Vicor fowards and double foward, using two switches and two diodes to erfectly clamp transients to the bus.

So called snubbers are inefficient and can be ineffective over wide load range.

Jon

[Marco]

The fastest non ringing lossless quench of the inductor current is to create an extra voltage rail slightly below avalanche voltage, with a Schottky diode from the drain to that rail. All the energy will be recovered to the second rail with the maximum voltage push back against the inductor current. Not a very realistic solution though, I must admit.

A two switch solution is lossless without a snubber and elegant . So a MOSFET above and below the load, with diodes to conduct current from ground to the positive rails when the switches turn off, similar to a two switch flyback. This has the advantage over a lossless snubber in that it's not dependent on say MOSFET capacitance or turn on time, which any lossless snubber certainly will be.

PS. I was wrong, the two switch solution actually quenches the current faster than using a flyback diode to a second rail in this case. The two switch pushes back with the supply voltage, while the second rail could only push back with avalanche voltage minus supply voltage, which is less.

[mag_therm]

Similar to methods mentioned by Marco and JohnB, I am using what I call a "Floating Snubber" which returns some energy per pulse, allowing smaller, lower inductance snubber components.

This snubber is in a hobby pwm push pull 300 Watt inverter, I am working on 2 versions which both have essentially the same snubber.
https://app.box.com/s/2orjo7boyqa74c0siq88yy8ydzspus4i

Snubber diodes are SB1H100, The converter link supply is 32 V DC.
The main IGBT's (obsolete IXGH32N60C) switch off with unimpeded rise to off-state until overshoot starts the snubbers.
The off rise time from about 12 Amp to peak overshoot is 70 ns and Collector Off-loss per cycle  is measured at 0.93 Watt.
Collector On loss to 12 A, again not impeded by snubber, measures 0.68 Watt per cycle.

The floating snubber capacitor voltage varies depending on collector current and duty cycle, and peak voltage is set by the value of the discharge resistor.
The discharge resistor is presently of 3 Watt rating.