Multi-phase buck / boost - design pitfalls for the over zealous ...

We get a lot of inquiries regarding multi-phase buck or boost designs.

Some of these are from power electronics design engineers - but many are from people just starting out who - very unfortunately - are not aware of the many limitations of control chips, layout, choice of operating frequency, FCC ( EMC ) regulations, off the shelf chokes, and a host of the finer points that affect the make up of a good power converter that works the way you want it to.

There is a small coterie of what appear to be (on paper) some very fine multi-phase chips for controlling up to 6 phases and beyond, buck, boost, sometimes bi-directional. The downside is these chips are designed by people who appear to have never first built a similar converter and discovered the ills that arise from noise, heat, layout, cap ESR, mosfet limitations and choke behaviour.

If they had - the chips we see today would be somewhat different.

The crisis management we attend to comes mainly in 4 flavours:-

1. The layout is so "interesting" that the thing will never work,
2. The power hoped for is so far beyond what the chip was (perhaps) designed for that thermal limits of mosfets ( un-calculated ) will kick in before 1/3 of full power is ever seen,
3. The noise level from switching even at lower powers will completely swamp the Isense circuitry - especially when all phases are going -> bang!
4. It works sometimes but the RFI blots out all local radio stations.

Multi-phase designs need to see the current in each phase in order to keep the current in all phases about the same - this is good design practice.

Some what surprisingly ( but done to supposedly increase efficiency ) very low ohmic resistors are used on the "quiet" side of the inductors ( see image above ) to look at the switching currents to determine when to terminate the switch cycle. If you have a 1m-Ohm resistor and 15A, you get 15mV ( and 225mW - great for efficiency...! ) - however 15mV at full power is not a whole heap of signal, especially if your input or output voltage is above 12V say - the so called efficient ( but really aggressive hard switching ) of the mosfets will generate noise - and even more noise at higher Voperating. Also the noise from other phases can occur at very unfortunate times for the comparator trying to control its own phase inside the chip - the 15mV can easily be swamped - leading to - you guessed it -> bang..!

The choice of allowable switching frequencies from vendors is really "inspiring" too - often up to 600kHz, again, this may be doable from 12V to 1v2 - but not so flash up at 50V. When you calculate the losses in a mosfet just from discharging it's own Coss at 50V @ 600kHz - you can easily get 2 watts for a low ohm 80V fet, on a DPAK this don't leave a lot of room for other losses - as the R-thermal on a 1" square bit of 2 Oz pcb is 50 degC / watt - so your device is already at 100 degC plus ambient, before we start to switch any appreciable current.

At some point you may want to sell your converter in a competitive market - if you don't or can't pass FCC reg's then your competitor may be pretty quick to tell the FCC just that - and the FCC will lose little time in shutting down your sales until you do comply - completely new design cycle anyone ?

The first 3 rules of power electronics design are; layout, layout & layout - if you have never laid out a similar power converter ( high power - multi phase ) then my friend - you are on a steep, painful and expensive learning curve - that will take four times longer than the Gantt chart you submitted to management.

The Zeroth law of power electronics is never assume anything - so when calculating the power in a mosfet @ 600kHz super hard switching - be sure to include the losses in the internal gate R due to the rms gate currents ( higher than you think ) - and the extra turn on losses because you are commutating the other mosfet's diode pretty hard.

Also the heat from the nearby gate drive chip - often more than you anticipated.

We could go on about chokes not being what they say on the tin and other factors - but you get the idea. For designs that work - pwrtrnx.com