dc/dc converterflybacksnubber
In many examples for DZ snubber zener voltage is calculated from maximum MOSFET rating and some arbitrary choosen voltage leakage (usually a percentage of Vin). They don't even use leakage inductance (available in some transformer datasheets).
It seems rather wrong, I would expect it to depend on Lp and Ip somehow. Or maybe, they use something close to maximum FET rating because it indeed provides best performance?
The design of this \$RCD\$ clamp requires the knowledge of the maximum peak current your controller can set up as well as the maximum voltage you tolerate across the MOSFET. I have presented all these equations in a seminar I taught at APEC in 2011 and entitled The Dark side of the Flyback Converter. The equations to determine the components values are there:
The principle of operations is to create a low-impedance voltage source hooked to the upper high-voltage rail which will clamp on the maximum excursion of the MOSFET at turn-off. However, be aware that the peak current can be much higher than the limit set by the controller considering the propagation time \$t_{prop}\$. That duration is the time needed by the controller to effectively turn the MOSFET off when the current sense pin has detected a maximum. Depending on the drive scheme, the MOSFET size, the input voltage and the primary inductance, there can be some significant overshoots destroying the MOSFET quite quickly. The capacitor value is more in the vicinity of 1 to 47 nF perhaps, as a very rough figure while the resistance cannot be too low considering the dissipated power.
People usually believe that the diode should be ultra-fast but it is little known that the turn-on time of a 1N4007 nicely competes with that of a MUR160 for instance. It is the turn-off (recovery) time that is much longer however but this lazy diode is often used in \$RCD\$ clamps of cheap adapters below 30 W because it nicely damps the oscillations at turn-off and reduces radiated EMI.
Also, surprisingly, the peak current going into the \$RCD\$ is often less than the power switch peak at the opening event. This is because part of the energy stored in the leakage inductance is used to charge the parasitic capacitance lumped at the drain until the diode conducts. By doing so, there is less current circulating in the \$RCD\$ network during the reset time. Adding a bit of capacitance across the drain-source of the MOSFET clearly helps on the \$RCD\$ power dissipation as long as the saved power is not lost in switching losses because of too big a capacitor. Typical values of 47-100 pF are often seen in commercial adapters and they also offer some snubbing advantage too.
the main cause of high leakage inductance is too much separation between the primary and secondary windings. But please provide more details, such as a picture of the cross section, what's the core material, is it gapped, etc.
Best Answer
Using your link and a specific circuit in the data sheet might help: -
The snubber zener is rated at 24 volts and the absolute maximum SW node chip voltage (from the DS) is 65 volts. Hence, with Vin at maximum (36 volts), the voltage that appears on SW is clamped to 36 volts + 24 volts + a bit more for luck + forward voltage of the schottky. I would say this will be no more than 62 volts and is inside the limit of 65 volts imposed by the data sheet.
Designers know that there may be 5% leakage inductance (worst case) and that there will be a back emf that needs to be quenched but knowing what the leakage inductance is is irrelevant to designing the snubber clamp voltage.
Sure there may be a peak current of (say) 1 amp in the primary when it gets open circuited and this (with a leakage inductance of 5% x 9 uH) is an energy storage of 225 nano joules that needs to be burnt by the snubber. It needs to burn it maybe 500,000 times per second and this results in a power dissipation of 112 mW.
So we do need to understand the leakage but only to determine the power of the snubber.