Help selecting switching power regulator

[mribble]

So I'm looking for a power regulator that can handle an input from 2 or 3 AA batteries or USB and convert it to 3.3V.  I thought this would be super common, but I'm having troubles finding something.

Here is what I've found so far:

• TPS6103x http://www.ti.com/lit/ds/symlink/tps61031.pdf I like this chip, but its only a boost converter so it can't step down the voltage like I need for USB to 3.3V.
• TPS6302x http://www.ti.com/lit/ds/symlink/tps63020.pdf This is a boost/buck converter so it should handle what I want from a voltage perpective, but it only comes in DSJ packaging.  This seems to be similar to QFN packaging which I really don't like.  I'm looking for a surface mount package, but it should have pins I can see (single ground plane in center like chip above is fine).
• Mc34063 There are lots of makers of this chip, but the data sheets only show it being used as a setup or a step down and not both at the same time.  I don't really know enough to know if can be configured to do both in the same circuit.  I also don't like the relatively slow switching time since I believe that generally results in more noise on the voltage rails.

Does anyone have suggestions?

[David Hess]

Any boost converter used in a SEPIC configuration will do what you want.

[Benta]

SEPIC is a nice choice, but for a qualified answer it would be nice to know the needed output current.

[mribble]

Thanks for giving me that sepic keyword.  I wasn't familiar with that.

I'm looking for > 0.5A output current (1A is idea, but 0.5 works).

I found this thread which had some recommendations, but so far nothing seems to work.  https://www.eevblog.com/forum/beginners/34063-switching-regulator-9-14vdc-in-12vdc-out/

I'm figuring minimum input voltage would be around 1.8V with 2AA batteries and that eliminates a lot of choices.

[jc101]

If you can go to 3xAA batteries the LTC3118 is a dual input buck/boost that looks like it may fit the bill as it comes in a TSSOP too.  They are expensive but work very well.

[CraigHB]

Low input voltages like that are a problem for most converter controllers, especially when you get into those higher currents.  One thing you can do is use a charge pump to provide the supply voltage for the controller.  I've done that before with low input voltages in order to use a controller with a higher minimum supply voltage.

The problem with buck/boost topologies like SEPIC and CUK is the inductor has to provide the whole of the output voltage compared to boost where there inductor only has to generate the difference between input and output.  It's a viable solution, but you get higher inductor currents and less efficiency.  The best way to go is with a four switch buck/boost that alternates between buck and boost mode as needed, but that's a pretty complicated regulator.  It's really quite a chore to get any buck/boost converter working well and there's not a wide array of controllers like there is for buck or boost.  It might just be easier to use four cells instead of two so all you need is a buck converter.

[mribble]

I kind of over simplified things since this was actually for 2 different projects.  I have too many things on my plate right now too so I actually posted a contract job to these forums.  If your interested in getting paid to help with this go check it out here: https://www.eevblog.com/forum/jobs/contract-work-2-power-supply-designs-and-motor-controller-circuits/

The LTC3118 is too expensive.  I'm going to be making hundreds of these.  I'm fine with the $1-3 range for this part.

3 batteries might be possible, but one of the designs is weight sensitive so I really want to go with 2 AAA batteries there.  If weight is still an issue I'll do the extra complexities of going to a lipo battery system, but I think 2 AAA batteries will work.

[mariush]

MC34063 or MC33063 or *3#063 generally run only from 3v and up, so that makes it unusable with 2 AA batteries (as rechargeables could be as low as 1.1v x 2 = 2.2v)
In addition, the switching frequency is fairly low so at around 60-90kHz so that means they require big inductors and diodes and a lot of other components besides the chip itself.

SEPIC (buck and boost) circuits require more components so there's more losses. In my opinion it's not worth it... if you plan to use 2 AA batteries, I'd say use a boost (step-up) only design, as you'll have about 2.2v in worst case scenario or about 3.3v in the best case scenario (2 x 1.65v with alkaline batteries) ... well, user may put lithium aa batteries inside which may have a "fresh" voltage of around 1.8v but seriously, just make a not there and say don't use lithium batteries (they're pretty rare and expensive anyway.

IF you go with 3 batteries, you may as well go with a 18650 battery which is 3.7v to 4.2v and 2500+ mAh .. either way you're guaranteed to have at any time more than 3.3v at the input , so you should just go with a buck only (step down) regulator.

Linear makes some excellent chips, but they're usually more expensive. For a one off or a few prototypes, they're worth it though .. the datasheets are excellent, with lots of explanations and even include
examples of circuit layouts and multiple example circuits.

In the past, I've used LT1307 to boost 1v to 2.5v from a supercapacitor to 5.6v (less than 10mA) - I've replaced a 9v battery in a multimeter with a supercapacitor, and 5.5v was the point where the multimeter started to complain about low battery voltage. 
It's a 100mA max boost regulator, but I chose it because it was available in DIP package (so it was fast to prototype a working circuit), it worked from 1v and up, and I only needed about 5mA for the multimeter.  As it works at 600kHz it needs very small ceramic capacitors and small inductor to work.
The bigger brother LT1308 can do over 1A and still works with as little as 1v at the input.

When I chose to go with LT1307 I was also checking out LT1613 which is even nicer, runs at 1.4mhz so requires very small inductors and ceramic capacitors, but it can probably only output around 250-300mA at 3.3v from lower voltages.

[David Hess]

At higher voltages things are easier so I will just consider the worst case while recognizing that your input voltage range straddles the regulated output voltage.  There will be problems getting 5 volts at 0.5 to 1 amp out of 2 AA cells down to 1.8 volts.

1. 5 volts at 1 amp is 5 watts.  Operating down to a 1.8 volt input, which makes sense for a 0.9 volt end of discharge, makes for a 2.8 amp average input current excluding losses which is a hard load for AA cells never mind those approaching end of discharge.  NiCd and NiMH cells are better than alkaline cells here despite their lower capacity because of their low internal resistance.

2. Operating down to 1.8 volts will require bootstrapping the switching circuitry to operate from the 5 volt output which is a standard procedure.  The difficulty here is *starting* into a heavy load but that can be overcome also; if necessary a small auxiliary switching regulator can be used to power the main switching regulator.

I have not checked the latest integrated converters but better performance will be available using an external power switch at these current levels.  A buck-boost design requires 1 inductor but 2 power switches.  A SEPIC design requires 2 inductors but only 1 power switch and the power switch is conveniently ground referred.

Diodes may be replaced by synchronous rectifiers for better efficiency but this adds complexity.  With a 5 volt output, even a schottky diode causes about 10% loss just by itself.

Attention will need to be paid to ripple currents through the input and output capacitors.  It might be worthwhile to choose a topology based on minimizing the input or output ripple current to lower the capacitor requirements.

What kind of operating life is desired?  This is relevant to selection of the input and output capacitor type.

[mribble]

Thanks Mariush and David.

It sounds like perhaps I should reconsider my batteries and maybe requirements.  What I gave were safe requirements.  I can probably reduce them to make this an easier problem.

I haven't thought about operating life.  5,000 hours would be more than enough, but less than that would also probably be fine.  I'm not sure what a typical operating life for something like this would be.

[David Hess]

It sounds like perhaps I should reconsider my batteries and maybe requirements.  What I gave were safe requirements.  I can probably reduce them to make this an easier problem.

The requirements just place some constraints on the design which will add to the cost.

I haven't thought about operating life.  5,000 hours would be more than enough, but less than that would also probably be fine.  I'm not sure what a typical operating life for something like this would be.

Nothing exotic needs to be used for that kind of operating life.  Proper derating still needs to be used though.

[CraigHB]

3 batteries might be possible, but one of the designs is weight sensitive so I really want to go with 2 AAA batteries there.  If weight is still an issue I'll do the extra complexities of going to a lipo battery system, but I think 2 AAA batteries will work.

500mA to 1A is a tough draw for AAA batteries.  Plus with a boost converter you're pulling higher amperage on the input side and boost needs a lower supply impedance.  What makes things involved is the need for minimal weight and a higher supply current than most small devices like that require.  I would consider going to a rechargeable Li-Ion battery which will provide higher voltage, less weight, and much lower supply impedance.  Though there can be safety considerations with Li-Ion batteries so that could be an issue.  There's also the need for additional circuitry to provide charging and safety.

[mribble]

I've completely reworked how this will work and separated the two designs.  Here is my new proposal for the more difficult one.

So to summarize  I'm working on a future add-on for this Kickstarter I did (https://www.kickstarter.com/projects/1492884446/electric-eel-wheel-5).  If you watch that 3 minute video you'll see there is that center piece that spins (called a flyer).  There are some hooks on it that you need to move by hand.  I've done some prototyping of the mechanical system that moves these hooks with a small motor and am happy with those results.  This project is to help design the circuits and electronics that will move this hook.

Because this will be on a rotating shaft I want to design everything to be light weight.  The motor will pulse on to full power for half a second every minute.  For that half second it will run at around 100 mAH at 5V, but if you could smooth out that current draw it would only be about 1 mAH at 5V.  So what I think we should do is slowly charge a super capacitor for 1 minute and then discharge it to run the motor for half a second.  With this kind of a slow charging circuit we should be able to use 2 coin cell batteries like CR2032.  My estimates is that 2 of those should last around 200 hours (1 mAh discharge and 225 mAh batteries) which is more than enough.

Here are the major components I'll need to design into the PCB:
1) Bidirectional motor controller
2) ATtiny45 microcontroller and programming header
3) On/Off switch
4) Shunt resistor attached to an analog input.  Software will detect when the motor starts to stall and switch motor direction.
5) Battery holders
6) Power regulation (5V for both micro and motors is probably fine)
7) Super capacitor to run the motor (I calculate around 50 mF) and a circuit to charge it from the batteries.

I will write software for the ATtiny and plan to put it into a deep sleep mode for 1 minute while the motor controller isn't running to save power.  I would expect the power usage from this micro to be small compared to the motor.

Here is the motor I plan to prototype with: https://www.pololu.com/product/2365