Switching Controller - Power dissipation calcuations

[norand]

Hi All,

I'm working on a design (will try to leave all the details out of this post for now), that requires a battery voltage of 7.2V nominal to be down converted to 3V and be able to provide load current of at least 4.5A (in case of motor stall).

When investigating which type of Power IC to use I understood the following:

• Linear Regulators get HOT. Therefore a SMPS is indeed the way to go.
• Switching Regulator - includes the MOSFET internal to the IC, and sometimes extra parts inside the IC so that typically all you have to provide are the inductor & decoupling caps. However you have to take care of the thermal management for the chip, and typically you can't get much current out of them, but they are optimised to be as efficient as possible with little external components. 
• Switching Controller - Idea of these is majority of the conversion is taking place outside of the IC, so therefore more design required and more for your BOM. However these can provide much more current as you can provide better thermal management.
  

Now to my question: How do you calculate the Power dissipation of just the Power Management IC of a Switching Controller?

Typically for a Switching Regulator, you calculate it as the following: (Output Power/Efficiency) - Output Power

 e.g. for 80% efficiency, (15W/80%) - 15W = 3.75W. Some of the switching controller ICs I have looked at such as the LM5642X (http://www.ti.com/lit/ds/symlink/lm5642x.pdf) can only handle 1.1W without an exposed pad (which I am trying to avoid so I can hand-solder) or 3.4W with an EP.

However, I believe I am right in stating that the Switching controller basically acts as a "smart" FET driver due to the resistor feedback, and modulates the higher voltage through the inductor, and therefore the IC should take none of this heat as it is all done external to the device. In this case how do I therefore calculate the controller ICs power dissipation? As I can only assume its based on the FET gate drive?

Thanks in advance for any help - and please correct me if I am wrong!   

[T3sl4co1l]

Controllers dissipate power from:
1. Operating current for the control itself
2. Driving the transistor gate
3. Ancillary functions if applicable

Some controllers have an internal LDO set for, usually somewhere 3-15V.  In this case, the supply current is independent of supply voltage (when above the nominal LDO setting), and so the power dissipation can be racked up quite quickly in this way.

LM25119 is one that I've used, which works this way.  It has four gate drivers, so the current consumption can get up there pretty quickly.

Consumption due to #1 isn't usually significant.  Indeed, some controllers exist which can run direct from 500V (TNYxxx come to mind).  (These do tend to be simpler ones, which helps to reduce current consumption.)  Probably the most you can find is old bipolar controllers, like UC3842 and TL494 (~10mA).

#2 is usually the dominant load.  Current draw is simply Qg(tot) * Vg(on) * Fsw.  Switching speed doesn't matter, and current spikes are (or should be) smoothed out by nearby bypass caps.

#3 isn't common, but can be useful in special cases.  Sometimes you need a local supply, which you can consider leeching off the VCC from the regulator (or VREF, or..).  Typically these supplies are not rated for much to begin with (a few mA?), aren't well protected (they aren't intended as general purpose supplies), and overloading will cause malfunction of the chip (since you're browning out the driver/controller, or shifting VREF, or..), so they need to be used cautiously.

Some chips take it even further.  LTC3810 has a regulator controller inside it too!  Add external transistor and you can dissipate as much power as you've rated it for.

Note that a bootstrap approach is very helpful, if you can arrange it.  This is how UC3842 and such start up -- supply charges through a big resistor, controller starts up, transformer kicks out an aux supply to run itself.  Most buck regs have a similar option, where if your output is 3-15V, whatever the controller can use, you can use a startup circuit then self-power it from the same output (no extra windings needed).  Other voltages, it may be possible to use a tapped or dual inductor to generate an aux supply, but this can get harder to arrange, as inductors of various ratios and values aren't very common.  At that point, a completely separate supply gets attractive, if annoying.

Tim