how to get +5V at 100mA and +6V at 2A from 7.4V or 11.1V battery supply

Li+ batteries are 'discharged' at 3V per cell, so two in series yields

6V (duh). In addition, their internal impedance rises significantly.

Given that, you would have to use a converter that can handle 100% duty cycle (not too tough to find), but I am not sure the batteries would deliver the pulse currents you want once close to discharge. Note that the output current is usually rated at 2C for most Li+.

I would likely go for 11.4V nominal (9V at discharge) which at least makes the pulses a little better for the 6V system (and slightly less at the battery pack connection).

To get above about 87% efficiency, you'll have to forego the buck converters with integrated switches (best I have had is 87% out of those), apart from the fact that they are usually rated at 1 ~ 1.5A max.

For small size, you want a small inductor, but that implies high switch current, so you'll also want to use nice FETs.

My suggestion (because I am doing almost exactly what you want) would be a LTC1735 controller, IRF6620 / IRF6623 FETs and a physically small (electrically, too - about 4.7uH) inductor. The specifics you'd have to work out - use SWCad from Linear tech (free download) to adjust the values and see the response. That would be the 6V supply.

There are a lot of controllers, of course, so you may have a different preference.

I'd then use a nice LDO for the 5V system. One I use that is good to

100mA with ultra low grund pin current (no load) is the LP2981 from National.

Hope that helps

Cheers

PeteS

Forgot to mention : I implemented a power supply like that (3.6V out, rated up to 4A load, load step 0 -3A causes

Hi pete - thanks for your post, you definitely have given me a good place to start from! A couple follow-up questions:

You mention that li-poly is discharged at 3V - I'm not quite sure what you mean by this. Do you mean that by the time the battery is drained it's voltage has gone down to 3V? This is the battery I plan on using:

(though I'd be open to other suggestions - my choice of this one is due to it having the highest energy density of any battery I could find) - anyways in the datasheet it says it has a "3.7V nominal working voltage" - though it also says it has a fully charged voltage of 4.2V. Am I understanding this correctly that this means that the battery drops from 4.2 to 3V as is goes from full to empty? This is something I hadn't realized about these batteries

You mention you were able to implement such a device in 8 square centimeters. I'm not sure I even have this much space to work with. It got me thinking, though - in the application there will be 6 very small seperate boards, all in need of 6V and 5V supplies. I had been thinking I'd have all 6 of these connected to the batteries (distances are about

5-10cm - and thus this step down circuit on each of the 6 boards. But if I were to use a central board I would no longer be so limited in space - 8 cm^2 would not be the slightest issue - even 20cm^2 wouldn't be a problem. So my question is this: would it be possible to implement a device with 6 times the output capacity? So, say, about 500ma at 5V and 6A at 6V (the reason those aren't 6x the values I initially gave is that only 3 boards will pull large amounts of current at any given time

- thus total current should never go above 6A - and in reality I expect it to be normally near about 2-3A.) Is such a circuit doable? Would the LTC1735 still be the best choice?

thanks for your time,

-Michael

Taking the points one at a time:

Li+ / LiPoly have a voltage range of 4.2V (max) per cell [and that *is* max - got to be careful charging them] to 3V nom at discharge, so yes, you have to design to take account of the input range. They are fine to use [and I use them in my commercial products] provided you take care about their charge and discharge characteristics.

The LTC1735 is easily capable of controlling that current. if the IR parts are a little difficult for you then you could use a more common leaded part, but to handle the current (pulse current in the range of

15A or so for a 6A load) you may end up with SO-8 devices. (If you didn't look at the IRF6620/6623 yet, I'll note that the pads for source and gate are underneath the device, but my local proto person has no problem putting them down with a hot air gun).

To split things, then I would still use the LTC1735 (I have implemented

15A supplies with it), although Linear makes a nice range of controllers - I am just very familiar with it and it does the job quietly. The LTC1735 has a wide range of outputs (16V in max). If you are going to use it, then get SWCadIII from Linear and also read their excellent app notes. I suggest AN76 in particular, as it specifically targets the loop control structure used in the 1735.

On the note of splitting things up, you may also want to use a switcher. If so, because the current is lower here, you could use a device with the switch built in (you want efficiency, but you quickly get to a point of diminishing returns). For this, I'd personally use the LT1767-5. Pretrimmed so there's no divider feedback network, although you'll still need an inductor. 4.7uH works well, but if you need lower current than about 1.2A, you can use a smaller one (which will slightly raise the efficiency due to lower copper loss in the inductor).

On the other hand, you'll get about 75% efficiency with a LDO rated at

0.5A (Pd @ 0.5A = 250mW due to load, plus internal power consumption) if you drive it from 6V - not 83% due to the fact the internal circuitry has to be powered.

The LTC1767-5 has a typical efficiency of about 88% - 90% at that load for 8V-11V in.

Hope that all helps.

Cheers

PeteS

This paragraph refers to your 5V supply options.

:: On the note of splitting things up, you may also want to use a switcher. If so, because the current is lower here, you could use a device with the switch built in (you want efficiency, but you quickly get to a point of diminishing returns). For this, I'd personally use the LT1767-5. Pretrimmed so there's no divider feedback network, although you'll still need an inductor. 4.7uH works well, but if you need lower current than about 1.2A, you can use a smaller one (which will slightly raise the efficiency due to lower copper loss in the inductor). ::

Cheers

PeteS

Yes.

Very similar in that regard to three NiMH in series.

Hi PeteS - thanks for your continued help. I think for now I'm going to just try to tackle the +6V supply. Once I get that figured out I'll go after the 5V supply. I will take your advice and try this with the LTC1735. I have downloaded the LTC1735 datasheet, the AN76 app note, and swcad iii. Looks like I have a lot of reading and tinkering ahead of me!

In the meantime, could you perhaps reccomend a mosfet for this application? I hadn't realized that the mosfets you mentioned had pads underneath them. I consider myself a fairly skilled solderer, having tackled even some terrible .5mm pitch parts - but at the moment I do not have any access to hot air equipment (though I'm currently investigating this) My analog knowledge is limited at best, so I wouldn't even know where to begin to look for a suitable replacement for this part. Ideally - could you reccomend one that is already in swcadiii? I'd really like to use one that is already in the swcad component library (unless a better one for the job exists).

One last question: Can you tell me if efficiency would be improved by using a higher or a lower input voltage? Also - can a LTC1735 circuit accept a wide range of voltage input? I'm asking these questions as I am not sure about the capacity that I will need, so it's possible I'll want to use more than 3 batteries, though I'm not sure of this just yet.

Thanks again,

-Michael

You are welcome for the help.

You will indeed have a lot of tinkering in front of you. One thing - do you have some basic knowledge of the amplitude and phase response of single pole RC filters? That's an absolute must for a switch mode controller.

The LTC1735 is a current mode controller, so it's response varies with output current, but is insensitive to input voltage variations - that's why I suggest SWCadIII. One of the experiments in there would be to set pulse loads from nominal (say a few hundred milliamps) to your various possible loads - you'll see if it becomes unstable. With that load variation, I would not use the default configuration for the LTC1735 circuit in SWCadIII, but the one in the datasheet (which matches AN76).

For your MOSFETS, you could probably use something like the Siliconix / Vishay Si7850 or 7852 for both control and sync. Both have reasonably low Rds(on) and fairly low Cgs and Coss. Those factors affect the efficiency - they are both in the SWCadIII default library. These are in SO8 packages, so they are easy to solder by hand.

For optimal performance, we usually choose low Rds(on) and ultra low Cgs for the control (top) FET, and ultra low Rds(on) for the sync FET - but that's when we're trying to squeeze 1 or 2 more percent efficiency out of the supply.

The efficiency of the circuit will get lower, typically, with a higher input voltage, so choose a voltage that gives you about 85% duty cycle or less on the Control FET (above 90%, a lot of controllers have trouble). The duty cycle is simply Vout/Vin. So at 11V in, 6V out, you get about 55% - a decent ratio. At your worst case input (batteries getting discharged), you get 67%, but with slightly higher efficiency (due to lower Vin). So my answer would be 3 in series (although you may want to parallel them - remember you have to stay within the 2C load current limit on Li+ batteries, usually).

It's perfectly feasible to get a supply with better than 92% efficiency when properly done (SWCadIII will actually give you a report on this).

Happy reading. I'll keep an eye on this thread over the weekend.

Cheers

Petes