It's basically a boost: -3 is what you want. -5 is typically chosen for 1-2 switch forward, where the 50% is "wasted" as reset flux and you definitely don't want it trying too hard to push into that range.
Which, just to explain that a bit more:
In a half-wave forward converter, the transformer reset voltage is inverse with the reset time. Basic flux balance of course. If it's on for say 320V * 5us, it needs to be off for equal Vrst * t_rst. At 50%, times are equal, so the voltages will also be equal, and a clamp RCD snubber will charge up to this voltage (more or less). Thus the peak switch voltage is 640V (plus not much transient peak, since, well, it's clamped right there, isn't that handy?), so probably using an 800V MOSFET will do.
Note that, if load current is zero, the controller doesn't know much about what's going on with the primary magnetizing current. The transformer is high inductance (a transformer transformer, as opposed to the coupled inductors of a flyback), so the current is small, probably less than 10% of total load current. So whether the magnetizing current is high or low, relative to its own limit (saturation current), is not something the controller is very sensitive to.
So, if the controller happens to push over 50%, the reset voltage will rise, eventually quite sharply (it's hyperbolic with PWM%), thus dissipating more power (there's more voltage dropped across the snubber resistor), thus drawing more current (magnetizing current may transition into CCM). And once it's in CCM, the primary DC current is a free variable, balanced relatively slowly (over multiple cycles) by the snubber RC time constant. Eventually either enough current flows through the snubber to put the transformer into saturation, or MOSFET avalanche causes a hard limit and flux walks up very quickly. (Obviously, TVS clamped reset does the same thing, too.) Once the transformer is banging into saturation, the controller only notices something different when peak magnetizing current (the extra due to saturation) approaches load current, and more power cannot be delivered (i.e. it goes into current limiting, probably with a lot of extra core loss).
It's not an instantly destructive situation, if it's not avalanching the transistor that is -- the peak currents are of course on par with normal maximum load currents, and the primary winding has to be rated for that anyway, so it's not adding more winding or switching losses. And output power is reduced by way of saturation effectively stealing what would otherwise be load current.
(Remember that, in a peak current mode 1/2 wave forward converter, the output inductor's reflected current is what's being tested. So it works just the same as a flyback, as far as the controller is concerned. And, like any peak current mode control, the inductor needs to be relatively small, to work in DCM, or shallow CCM with slope compensation.)
The real trouble is that there's no good limit, it's nearly a free variable; the PWM is very squishy, and likely the snubber isn't designed to draw very much current (to drive the primary into saturation, where such a limit would occur) before the MOSFET avalanches. So, we'd rather just design the thing with a hard 50% limit to begin with, and never run the risk of all this weirdness. The 50% limit then is fairly arbitrary, but it's most reasonable by symmetry (equal and opposite conduction and reset pulses), and allows clamping at +V (either with a reset winding and catch diode, or by flipping around both ends of the same primary as in a 2-switch forward).
So that's what motivates that.
Anyways, such a limit isn't relevant with SEPIC -- higher PWM is higher Vout, nothing weird going on, it's tight and controlled.
Tim
Home
Help
Search
Login
Register
