Guru
Joined: 09/08/2007
Location: AustraliaPosts: 4406
| Posted: 02:07pm 07 Mar 2017 |
[quote]I think it keeps it down because in the first few hundred nano seconds, the inductor actually works on the fast rising wave front of the DC into the transformer primary, as though it feels like a multi megahertz wave front ( 100mhz?).... it is not?.
It is the leading edge of the rise of the pwm wave of the 20khz, and may be only <.1% of the 20khz time window... so it may be say 1000 x 20khz or less.. or more..... but whatever it is, it is very very fast, and it is related to the transformer primary impedance.. ( as was the time width of the wave from the fencers.... high impedance = wide waveform)
So if we can garner a few uh for a few nano seconds, it 's job is done. After that we don't need to care, and in the real world this seems to be the case.[/quote]
Yes, very nicely put Mr oZ.
Our mosfet has three distinct states. Off when it is completely open circuit, and on when it has a very low resistance (a few milliohms).
The third state is while it is in the process of turning either on or off, where it goes through a very rapid change of resistance. It cannot go instantly from off to on, or from on to off, there is some very short time interval where the resistance is actually changing.
Now that is a very big problem for us, because at some stage in the switching process both the voltage across the mosfet will be high, at the same time as the current is also very high. That causes a MASSIVE spike in internal power dissipation.
The faster you can make it switch, the shorter duration the heat spike will be. But there are obvious practical limits to how fast any circuit can be made to switch. You cannot make the heat spike smaller, only narrower by speeding things up.
So we need another way to get around this particular problem.
One solution is to put a very small inductor in series with the mosfet. What then happens is that the mosfet turns on taking perhaps 100nS to do so, the rate of current rise is limited by the new added inductance. What we want is for the inductance to be high enough for 100 nS (in this example) to hold back the current until the mosfet resistance has fallen down to the milliom region, then for the inductance to completely go away.
We can do that by winding a very few turns onto a ferrite core that has very high permiability (inductance per turn) but saturates at a low current. As long as the inductor can hold back the initial rise of current for a very short period, that is all it has to do.
The mosfet can switch on under low current conditions, held back by the inductor, and once its turned on really hard, the inductor saturates and the current can THEN really spike upwards, but the heat and dissipative power loss is no longer there.
So there is no real surprise that fitting an optimum turn on snubber inductance significantly lowers inverter no load switching loss.
There is a tradeoff here between the magnetic properties of the core, how many turns, and mosfet switching time. Too few turns and you don't get enough initial inductance. Too many turns and it will saturate too quickly before the mosfet has a chance to turn fully on. Its all a bit of a guess without being able to actually see and measure what is going on with appropriate equipment.
What works well in one set of operating conditions, may not be optimum for someone else. So its a case of trying a few things and tweaking by trial and error for minimum turn on switching loss. That should be reflected in reduced inverter idling current.
There is no real mystery in what it does. The magic is getting it to work exactly right for YOUR inverter, which is always going to be a bit different than someone else's, particularly with home brew projects.
Its a bit different for commercial production where all units are almost identical.
Cheers, Tony.