DC-DC like TPS62743 but with a (much) higher Vin max?

[GeorgeOfTheJungle]

Does it exist?

-Ultra low quiescent current.
-High efficiency at light loads: Most of the time my circuit is sleeping @ ~ 150 µA
-3.3V, need no more than 200..250 mA max when not sleeping.
-Vin max of 30v would be great so I can power the thing with 7 li-ion in series: 3.7v*2.5A*7= 64Wh => enough juice to run on batteries for years.

¿?

Edit: LM5166 looks good, is there anything better?
Edit 2: maybe it's not a good idea to put the cells in series?
Edit 3: Using this chart, my energy spreadsheet says 2s3p would be much better than 7s:

[GeorgeOfTheJungle]

Is TI now better at analog than Linear Technology/Analog Devices?

[GeorgeOfTheJungle]

Anybody knows what's this "IAGCD" chip?

https://www.aliexpress.com/item/10Pcs-Mini-DC-DC-12-24V-To-5V-3A-Step-Down-Power-Supply-1-8V-2/32848669016.html?spm=a2g0s.9042311.0.0.52e74c4d1YbPTh

[nctnico]

If you only need 3.3V I'd use several batteries in parallel. It may sound odd but if you do the math a linear regulator may be more efficient compared to a switching one because switching regulators use a lot of power. You really have to do the math and compare carefully.

[GeorgeOfTheJungle]

But if I use a single cell I can't extract all its juice because I can't go all the way down to the lower limit of a li-ion cell (2.5..3V). Then it should better be at least two in series, no?

1s range: 4..4.2V fully charged .. 2.5..3V discharged
2s range: 8..8.4V fully charged .. 5..6V discharged

¿?

[nctnico]

You shouldn't discharge a Li-ion too far and there is not much energy in the region below 3.3V anyway.

[Siwastaja]

I assume your "mostly sleeping" circuit consumes little average power, and you intend the li-ion cell to discharge slowly (tens of hours or more), i.e., the discharge rate would be somewhere below C/20.

Now, while you can and do discharge the cells down to 2.5V (or 2.8V or whatever), this is done in high-current applications, because the ESR of the cell is fairly high near zero State-of-Charge - especially in cold environment.

For you, such an issue isn't true - this means, there is little resistive drop over the Open-Circuit Voltage of the cell. Now, for a typical li-ion cell, the OCV of an empty cell is around...

3.4 V!

So, if you employ an almost zero-dropout LDO, the chances are, you are already extracting 100% (or very close) of the charge, and won't need to run down to 2.5V. Running down there wouldn't give you much extra energy, and would risk damaging the cell. I mean, if you read the cell datasheets carefully, they tend to specify the end-of-discharge condition like this:
2.5V @ 0.5C discharge current.

If you are not using this exact current, you are exceeding the specification, and going below their definition of 0% State-of-Charge. This may sound unintuitive, since you are using lower current than specified.

I.e., what nctnico said is exactly true for low-current applications.

For really high-power stuff (running near the maximum ratings of the cells), especially in cold weather, there can easily be 20-30%, even closer to 50% charge left at 3.3V.

[T3sl4co1l]

Do you really need 3.3V?  Can you not use 3.0, or 2.5V minimum?

The simplest, most efficient solution is no regulator at all!  You will need a 5V-tolerant MCU for this, and if you have peripherals that are more picky, they won't work too well of course.  It may be worthwhile to find peripherals that are tolerant of wide supply voltages!

Otherwise, a single chip buck-boost or SEPIC controller would seem a fine start, or two cells and a buck regulator.

Going to high voltage is definitely not the way: you waste a ton of power just moving the switching node back and forth.  Duty cycle is also low, which costs efficiency and other possible limitations for the regulator (like minimum on-time, therefore setting maximum frequency, potentially requiring larger inductors than expected).

Tim

[GeorgeOfTheJungle]

@nctnico Do you know what is the dropout of the best LDO money can buy? A problem I foresee that worries me is that when using one cell at low SoCs the full load current peaks may well make Vout go too low and hang the µC, because I have to enter full load mode to measure SoC, it's a sort of catch 22 situation, isn't it? With two or more cells in series that can't happen.

@Siwastaja I intend to deduce the SoC from the battery voltage (not a gas gauge), but will only look at it when both µCs are on => when the load is the most, around 200 mA. That'd the best way to do it, I think, no?

@T3sl4co1l Yep, I ought to look/study more carefully the Vin min specs of everything, good idea! But the Vin max is less than 4V, that I know for sure.

Hey, thank you all!

[nctnico]

I don't know which uC you are using but if you use one which has a wide operating range then it shouldn't be a problem. The dropout on an LDO can be less than 100mV.

[Siwastaja]

IMO there is no need to generate bigger-than-normal loads to measure the cell voltage for SoC estimation. Use a capacitor for dealing with short peaks. Note that the cells have considerable capacitance, and some capacitance-like chemical effects as well that work to some degree even in range of hundreds of milliseconds. So even near 0%, it doesn't drop below 3.3V if you have sub-millisecond peaks. I would definitely just test this out because it may well be the simplest case!

But do test it at the lowest intended operating temperature, and leave some leeway because the cell DCR rises as the cell ages.

Linear reg dropout is IIRC a parameter in the Digikey search, so you can find the lowest.

Also as Tim said, consider using no regulator.

Also, it's possible that your MCU / other devices may be just OK with lower than 3.3V voltage, even if your max limit would be at 3.6Vish. With most modern LDO's, there's absolutely nothing wrong in driving them in "dropout" or saturated mode; they just pass the input with dropout and don't regulate, but will step down the overvoltage.

[T3sl4co1l]

If the limit is 3.6 or 4.0V or something like that, then an LDO to limit it to 3.3 or so would be fine, and yes, there are plenty of LDOs that have low dropout, so you can get very nearly the full battery voltage when it's low but also limiting the voltage to 3.3V or whatever when it's too high.  And at an idle current of 10s uA.

Trying to squeeze that last bit of charge out of a battery, especially a Li Ion, is Doing It Wrong.  It's okay for alkalines, where you might have say 4.5V nominal (3 cells) but because of how they discharge, it's worthwhile maintaining operation down to 2.25V or whatever.  Not that you'll be able to draw a pulse load down there, regardless.

You simply get however many pulses of that heavy load as you get, and that's that.  If it's not enough, add more cells!

The only purpose for the MCU (if there is one) at low supply voltage, is to know the battery is too far gone to draw another pulse load, and tell the user that "It's Dead, Jim!"

Tim

[SiliconWizard]

How about this? http://www.analog.com/en/products/lt8646s.html

[GeorgeOfTheJungle]

That looks good, very good indeed, thanks!

[GeorgeOfTheJungle]

MP2315

Thanks!


I think I'm going to use 4*1.5V alkalines in series, size C or D, because the self discharge rate of li-ion is about 2% per month, but for alkalines it's only ~ 3% PER YEAR, and because 4(D cell)*1.5V*12Ah= 72Wh and that's more than plenty enough to make it run for ~5 years.

To measure SoC I'll look at the batt voltage when it's sending via WiFi (peak current of ~200mA), and if it's close to 4*1V then it's time to change the batteries.

I can power it in sleep mode (120µA@3.3V) from a 470 mF supercapacitor with the smps dc-dc converter off (quiescent current a few µA) and will only turn on the smps when the µC awakes, to refill the SC (and power the µC while it's awake).

The energy I need for the longest sleep period is only 1/64 the energy of a 470mF SC charged @ 3.3V. I might be wrong, this may be silly, but it sounds so cool! :-)

The smsp efficiency is a disaster at light loads (at sleep) but it's close to 90% for loads >= tens of mA (when awake), so in principle the SC plan sounds good to me. I still have to see/check better all the numbers in a spreadsheet, though...

I have not made the maths for other options using (good) LDOs yet, mainly because I don't think I like the idea of running the invention close to Vin min, I'm afraid of brownouts and the best recipe against them is higher Vins + switching regulators, not LDOs.

Please comment below, say what you think, if you like the idea, or insult me if you don't :-)

Thanks!

[Siwastaja]

I think I'm going to use 4*1.5V alkalines in series, size C or D, because the self discharge rate of li-ion is about 2% per month, but for alkalines it's only ~ 3% PER YEAR

Nooooo! Don't buy this classical myth which originated from the "Battery university" fake science site. 2%/month is considered a faulty cell. Li-ion leaks much less, unless at very high temperatures; it's one of the least leaky things you'll get. See some actual results here:
https://www.eevblog.com/forum/projects/low-curent-pulse-charging-a-sla-battery/msg1868660/#msg1868660

At around 3%/year typical, note that even the highest self-dicharges I recorded, about 6%/year, only apply to a full cell. Li-ion basically stops leaking below around 50-70% which makes it great for power harvesting applications, or long term storage; you just derate the capacity by deliberately not worrying about it self-discharging somewhat down from 100%. If you can afford doubling the capacity you actually need and only run between 50% and 0%, you are easily down to well below 1%/year.

Sony cells tend to be better; this is both hearsay within the industry, and my own measurements also proved this stance. Panasonic performed very well as well.

Severe overdischarge (i.e., lack of undervoltage cutoff protection) of the cells may permanently increase self discharge by copper ion dissolution in the electrolyte.

Li-ion is probably your best bet.

[Siwastaja]

Please comment below, say what you think, if you like the idea, or insult me if you don't :-)

Sounds extremely complex to me.

What's your intended battery capacity?

If your maximum peaks are at 200..300mA, and you decide to use, say, a 6000mAh cell (two 3000mAh 18650's in parallel), that would be C/20 maximum peak, which means, you'll extract at least about 90% (coming on the top of my head, but will be close) of the capacity by having your end-of-discharge at, say, 3.4V (with a 100mV dropout LDO). No need to go to 2.5V! Simplicity wins even if you might "lose" a few percent.

Note that a supercap stores about 2Wh/kg, while a li-ion cell stores 250Wh/kg. Supercaps are expensive, as well. Extra conversions stages waste energy and design resources.

By using your resources (money, space, weight) to just put a few more cells, you would build a super simple, quickly deployed and robust system with massively longer runtime between charges (who doesn't want this?), and the best energy density. You wouldn't need to think about self-discharge if you don't rely on having full 100% retained over the period of many years - maybe you can accept that after a decade, your pack has degraded to 60% of original capacity, if that original capacity is high enough? Extra capacity means less resistance - this gets important as the resistance rises through cell aging -  and less voltage drop, meaning the voltage-based SoC estimate would be almost spot-on with no compensation algorithms, just a simple table lookup. If the peak discharge is C/10ish let alone C/20ish, there's not much difference between the peak consumption and idle measurements.

You can first-order approximate the resistance of a (slightly aged, a bit cold) 3000mAh 18650 li-ion cell as 100mOhm, or near 0%, as 200-300 mOhm. That'll give you the idea of the voltage drops involved in your 200mA peaks.

The best way to get ultra low quiescent current for a mostly sleeping design is a linear regulator (or no regulator at all).

[TJ232]

If you need only 2-300mA you should give a try also to the MCP16311/2.
It's a high-efficiency,  fixed frequency, synchronous step-down DC-DC converter in that operates from   input   voltage   sources   up   to   30V.   Integrated features include a high-side and a low-side switch, fixed frequency    peak    current    mode    control,    internal
compensation, peak current limit and overtemperature protection.

I have mainly used it to replace hundreds of linear regulators in a poorly designed network equipment PSU. I like it so much that I'm using it almost anywhere where fits in requests.

You can find a lot of details, including schematics here:  30V Synchronous step-down DC-DC converter