I have invented a method to achieve interleaved phases for transition mode boost converters, while also ensuring it is possible to drop phases if one of them fails, and when that phase drops out the others will automatically compensate.
I have a 4 phase prototype below.
The two phase prototype works, but it has a glitch during startup when at the start of every 120hz trough, both phases turn on at the same time. So to ensure this doesn't happen the input capacitor has to be big enough that the phases don't drop out, which will limit the PF to about 99% at full load, reduced a bit at low load.
The input capacitor also has to be large enough to keep the input voltage ripple (when fed from a current source) below about 25%
also, there's another glitch were on the downslope of the 120hz rectifier output, the phases don't spread themselves out properly, they don't overlap, but they do run together, one turning on immediately after the other, you could call that a 90 degree interleaving. (which is actually a known way to reduce the required size of the emi filter.) I think this can be prevented by proper ratios on the current sense resistor.
(each boost converter has a current sense resistor, but at the midpoint they are all connected together)
specifically: the method by which it forces interleaving is in the current sense resistor. they are all 'on top of' one resistor (actually a bunch of discrete ones in parallel) such that the master current ripple is "on top of" 'all' of the current sense resistors.
another method i have thought up is to resistively sum the total current ripple on top of the individual phase currents, doing away with the ultra low impedance board design needed otherwise.
photo of the 4 phase version
large one here
http://johansense.com/bulk/boost/4phase_v1.JPGmore info here (the last two posts)
http://www.edaboard.com/thread285741.htmlanyhow, i think this is worth patenting, because the method i've used to achieve natural interleaving across any reasonable number of phases also works if an emulated current ramp is used instead of current sense resistors, which is what i'm thinking about building from discrete components.
this method of interleaving any arbitrary number of phases would also work for transition mode buck converters.