Highly Efficient Boost Converter

[kevyk]

Hi. I am looking to boost around 0.8v-1.3v to 3.7v in a highly efficient design. The ideal will be ~200ma of current available on the output even when input current is low. Any thoughts about ideal ways to implement with these specs? It could be a discreet component design or an off-the-shelf IC, just with super high efficiency at super low input voltages.

Thanks,
Kevin

[PCB.Wiz]

You are into the realm of energy harvesting, with that low input voltage, but the 200mA is high for those systems

Parts like ADP5091 have 500mV min (380mV typical), but they spec 150mA output.
Most are designed for 1-2 cells in, so do not deliver much power below 1.2V

[kevyk]

Yes, I had looked at Analogs 5091 and saw those specs. I am guessing that there may be a fundamental physics limitation limiting the output current given such a low input voltage. Any thoughts about that? Thanks!

[PCB.Wiz]

Yes, I had looked at Analogs 5091 and saw those specs. I am guessing that there may be a fundamental physics limitation limiting the output current given such a low input voltage. Any thoughts about that? Thanks!

You may need to over-rate the 3v3 side.

eg I find a TPS61021A, looks to spec about 500mA at 0.9V, but manages 1.5A at 1.6V 

[ArdWar]

You won't find one with high efficiency at that low voltage and that high current. Maybe you'll get about 50-60% efficiency at most realistically. 0.8V is barely enough for the internal circuitry to properly control the switching transistors, let alone fully turning them on.
Too much overrating might be counterproductive too, high current converters are almost always optimized for low Rds(on) which often also means higher gate drive voltage as a compromise. There may exist converter with internal charge pump, but that design kind of defeat the efficiency.

From a cursory search I found TPS61200 which lists I_O(max) vs V_in, but it only manages 60% at 200mA out from 0.9 V in. TPS61299 came short with specifying ~150mA Iout at 0.7Vin.
Other TPS612xx theoretically fits the specification, but none gives actual performance data. Not that I expect previous one to actually meet the advertised efficiency target in real application anyway.

[Someone]

You won't find one with high efficiency at that low voltage and that high current. Maybe you'll get about 50-60% efficiency at most realistically. 0.8V is barely enough for the internal circuitry to properly control the switching transistors, let alone fully turning them on.

Solutions from Webench provides some hints as to the possible efficiency even if TI doesnt offer a part with low enough turn on voltage. Analog, Maxim, Linear Tech, even Microchip or Renesas have multiple chips which meet the OP's requirements and push up into 80 or 90% efficiency.

[kevyk]

Thank you all for your input!

What about rolling my own - what about holding my tongue at the right angle and fine tuning specific parts in a Synchronous Boost topology (2 mosfets, ind., cap., timing, etc.). Of course, making sure the electrons are right side up so they don't all fall out (Dave is such a crack up...  ). I know a lot of smarter people than I have done this, but just wondering...

Thanks again!

[Doctorandus_P]

I am guessing that there may be a fundamental physics limitation limiting the output current given such a low input voltage. Any thoughts about that?

No. It is irellevant. Most MOSfets are enhancement types, and they need some applied voltage to open, but SMPS circuits for low voltages often have a built in charge pump to generate a high enough gate voltage for the FET.
(Are Depletion MOSfets even made these days?)

What about rolling my own

There are some very simple "joule thief" circuits but they rely on hand wound inductors, have to be tuned individually and work only reasonably well in a small power range.

Rolling your own can be a very good learning expericence, but if you will be able to make a well behaving circuit is not ensured. A starting voltage of 0.8V is a bit difficult to achieve. Adding a charge pump may be necessary (increases complexity), or if you can get it started, you can use the output voltage to have a high enough voltage to switch the FET's.

Controlling all the things is also a brain teaser. There are some microcontrollers that can work on 800mV I would probably start there. You can use it to generate PWM signals, put the control loop in it, and maybe a voltage divider on a digital pin is good enough to get a stable enough output voltage. You definitely want a synchronous converter with such a low input voltage, as even a single diode (Even schottky) will have a high enough voltage drop to get the efficiency below 50%.

Also keep in mind that if you want 3V7 @200mA at the output and your input is 800mV, then you need 925mA on average at the imput. Peak current during PWM is likely more then 2A. But SOT-23 FET's that can do 5A are readily and cheaply available these days.

[Marco]

0.8V is barely enough for the internal circuitry to properly control the switching transistors, let alone fully turning them on.

As long as it works somewhat, it can bootstrap itself.

[Doctorandus_P]

0.8V is barely enough for the internal circuitry to properly control the switching transistors, let alone fully turning them on.

As long as it works somewhat, it can bootstrap itself.

Yes, unless the load is to heavy, then you risk not starting it up at all. Building a small stepup such as the joule thief to power the circuit itself, and then building a bigger circuit for the high (and variable) current may be a good option. The joule thief can start from a low voltage, but it has a small power range and it's output is not very well defined, but good enough to power the rest of the SMPS circuit.

[Marco]

With luck the load has undervoltage lockout, otherwise one more MOSFET.

Just using two converters is probably easiest. One with low startup voltage to supply Vin on say TPS61022 or MP3421, after the high power converter is running you can turn off the low power one.

[kevyk]

Thanks again everyone for the input!

[T3sl4co1l]

You won't find one with high efficiency at that low voltage and that high current. Maybe you'll get about 50-60% efficiency at most realistically.

I don't know about that. Well, maybe at the low end -- at the higher end at least, I'd measured this thing,
https://www.seventransistorlabs.com/Images/JouleThief3.jpg
(with a schottky and capacitor instead of LED load) at ~60% efficiency.  Not bad for a dumb old blocking oscillator.   PBSS303NX and friends are really quite remarkable parts.

Doing it with a coupled inductor (like this) is pretty reasonable; it gets suddenly a lot harder if it needs to self-start while using a single inductor and nothing else (I mean, a chip, or couple transistors, that sort of thing).

At least there isn't much reason to prohibit coupled inductors -- they're modestly available nowadays, and a 1:1 ratio will do here.

To push "high efficiency" say 80% or better, you'll almost certainly need a MOSFET circuit, which needs a bias startup circuit (as hinted above) and a controller/driver.  Probably LT's low-input regulators would be something to look at, or I forget if they have a controller so you can scale it to MOSFETs of ~arbitrary rating.  "Energy harvesting" converters may also be of interest, for low voltage startup.

Tim

[Whales]

What about rolling my own

There are some very simple "joule thief" circuits but they rely on hand wound inductors, have to be tuned individually and work only reasonably well in a small power range.

A few other joule thief details:
 - can be very power efficient (are they zero-voltage switching?)
 - are magnetics cost inefficient (need bigger magnetics than an off-the-shelf boost controller)
 - do not have voltage regulation feedback
 - can be expanded to unipolar two-transistor inductor-fed variant (see The 8 Watt driver)

[PCB.Wiz]

What about rolling my own -

You could do that, but keep in mind lcsc show the TPS61021A I mentioned for 21c/100

or there is a Diodes inc AP72250CJ12-7 for ~31c, that starts at 1V and runs down to 570mV and curves show specs 500mA out at 1.2Vin


If you cannot find a single part, you could split the problem into two.
Find a low mA, low start voltage that can give gate drive voltages, and find another package that can do Boost when given a 3-4V bias voltage.
ie you need to find a part with separate VIN and GATE boost supplies.