I have been recently working on my current-fed converter. Its only advantage is that it works, but not exactly as advertised. This circuit deserves a separate autopsy thread, but the scope for some reason doesn't like the pen drives I have at hand. So it would be pointless to start without some screen dumps indicating where the problems are.
I would like to evaluate it against some alternatives. A bag of independent bucks running from a shared Vin seems fair. For the purpose of this discussion ignore cross-regulation issues and assume the circuit is there to produce non-isolated 3.3V@3A from Vin in the range of
10..20V. So this makes this 8-MOSFET synchronous buck-fed full-bridge with synchronous rectification a bit Rube Goldberg machine for the task (hereinafter CFFB for short), but never mind. Its peak efficiency is ~86% -- disappointing.But the bucks are not shining either. A simple converter made of a TPS54302 is about 90% efficient, which is in line with what the datasheet says. But why just 90%? There are just two MOSFETs and an inductor. The CFFB has 4 switches in series with an inductor and a transformer and is still just 4 percent points worse. So I'd say the problematic device is the TPS, not the CFFB. For such a simple circuit I would expect efficiency in the 95..97% range. But it is just 89-91% peak. My first thought was "no problem, there are other synchronous buck chips". And I was very wrong. The crop of my search has brought either some "high efficiency controller, up to 90%" devices or "up to 95% efficiency" chips with some laughable max V_in of 5..7V. There are two exceptions: the MP2315, which -- according to its specs -- should have ~92% @12V V_in and ~90% @19V. But it looks so similar to the TPS that I find this (unacceptably low) figure of merit coming from the PMPO fantasy land. The second exception is the LTM4607 from LTC. This is a buck-boost, but they claim up to 98% efficiency. No graph for my requirements was provided, but I assume they can have 94%, they are the LTC.
Now, this is insulting. I call your expertise to the rescue, there must be some way to build a 98%-efficient step-down 10W converter. V_in=10..20, V_out=3.3V, I_out=3A, non-isolated. No other constraints: any topology, any parts including GaN, no light load efficiency requirements. This is my variant of the Little Box Challenge, except the million dollar prize. ;-) I refuse to accept that my CFFB is a success.
Best regards, Piotr