ALL switched-inductor topologies must be current mode.
You can use a voltage mode control with exceptions, but by the time you've hacked enough cases on to make it reliable under operating extremes and transients, you've made a shitty, overly complicated current mode control.
Better to do it right from the start.
I haven't had problems with that kind of dynamic range on any supplies I've made. Offset at the current sense amp manifests as a dead band near zero if positive, and a minimum current if negative. The one isn't so much an error as a preferred behavior (you can actually set it to zero), so just bias the current sense slightly to guarantee offset in that direction.
Don't use opamps that are prone to RF rectification, or if you do, be damn careful about filtering inputs, outputs and supply (notch ground planes if necessary to isolate loops).
And there, that's it. You've got at least 20dB useful range if the ripple fraction is very large, and probably more like 40dB (more than enough) with typical values. It may even be fine out to 60dB, but the error amp will be operating hysteretic by then so it's harder to say (effectively, the output ripple frequency drops).
My flashlight is a good illustration of this:
formatting link
formatting link
formatting link
When turned down to the narest sliver of setpoint, it does indeed get dimmer and dimmer, until the blinking is obvious. About 10Hz is the lowest I can get it.
(If you prefer not to have this behavior, but just a sharp cutoff instead, you can apply a little positive feedback around the error amp. Be careful that it doesn't dominate over normal loop function and turn it unstable. It's like Phil's favorite trick of making a PLL auto-sweep with positive feedback, but then when the phase detector kicks in, it more than overrides the high resistance path.)
Incidentally, this circuit is a very nice and general current mode controller; I build a Cuk variant for 18V 20A output:
formatting link
(packed like sardines into its box, I ran out of space for enough filter caps even; on the upside, it's only intended for intermittent operation, so I guess it doesn't matter much that everything gets hot really quickly. :o :^) )
Multiphase is definitely your friend here. You can't commutate impedances that low (i.e., ~0.5 ohm) without having a mess of loop inductance, even with low profile (DFN) transistors. At least, not quickly, and I'm guessing you want to go as fast as you can to keep dissipation and size down.
Tim