Power rail design from Li-Polymer battery chip output

Question

Case:

I am building a battery driven device and so far it seems like I will be using the MCP73871 to provide my device with battery power and charging via USB. (Datasheet )

I need to regulate the output from the MCP73871 to 5V and 3.3V. The 5V is to power the digital part of the circuit (microcontroller mostly) and the 3.3V is to power an accurate analog measurement circuit.

The Li-Polymer battery will have a voltage between 2.5 - 4.2 V so must be able to convert from this range to 5V and 3.3V.

I am looking for design recommendations.

Some questions that arises:

1) Is it possible/does it exist a non-switching step-up converter?

2) If I need to use a switching regulator to get 5V, should I use one boost converter to get 5V and one boost/buck to get 3.3V? (As the input voltage might be higher and lower than 3.3V)

3) If I use a switching regulator (boost) to get 5V (for the less sensitive part of the device). Do I still benefit from low noise on the power rails by using a LDO or non-switching regulator to obtain 3.3V if I connect it on the regulator outputting 5V? Is this a better solution with respect to noise on power lines?

4) 5V and 3.3V are very common used voltage levels and I assume there must be some type of chip that can have an input of 2V+ and output both 5V and 3.3V simultaneously? If this is the case, an example of such a chip would be appreciated, but do they have/exist with low noise features?

The questions above reflects my thoughts on how I see this might be solved. Any other ways are very welcome!

EDIT:

I am using an ARM Cortex M-3 running at 72 MHz. It will for the most time be in sleep mode and when running it will be clocked at 72 MHz. Is also has bluetooth communication draining some battery.

It is possible to be running the digital part at 3.3 V as well.

Answer 1

IMO the two main question would be :

1) What microcontroller do you use? You might be able to use it at a lower voltage, maybe by decreasing the clock frequency. This would simplify your problem a gread deal.

For example if you want to use an ATMega, you may power it with as low as 1.8V if you keep the clock frequency at or below 4MHz. It will also be happy at 8MHz with much less than 3.3V. Then you can power it straight from the battery/power manager chip; not using a regulator at all will give you some significant battery gains.

2) What is the 3V3 part's lowest rated voltage? Depending on that, and the possible gains you'd make by not regulating the mcu voltage, you may be able to just discard the last drops of juice from your lipo for the same battery life (there is not that much energy left when it drops below 3.3V anyway... plus, that'd be healthier for the battery), and get rid of the step-up to 3.3V altogether. Now a LDO regulator might to the job. No switching regulator means less noise ; less parts, less power usage, less volume/weight and less price also might mean a possible bigger battery...

In order to make the right choices you'd need to specify the specifications of the parts you'd use. But using an unregulated voltage for most of the circuit, including MCU, and regulated voltage for only the 3.3V parts has many advantages.