Coulomb counter IC for coin cell products.

[Psi]

Has anyone come across any tiny coulomb counter ICs for coin cell products.
The sort of thing you could put in a CR2032 powered device that normally lasts around a year and get reliable battery capacity stats from.
Maybe in SOT32-6 or 3x3 or 2x2mm DFN package. Something tiny.
Would need to be very low power, maybe +3uA. and probably need to be under 50c in volume.
Max Current maybe 80mA or so for bursts.

I'm looking myself but have not found anything suited to this.

[SiliconWizard]

The BQ35100 may do what you need, and a bit more.
https://www.ti.com/product/BQ35100

[Psi]

Needs to be smaller.  Think wearable with tiny PCB.   So like SOT32 / 3x3mm DFN package

[Smokey]

BQ27220YZFR ?
https://www.digikey.com/en/products/detail/texas-instruments/BQ27220YZFR/6024316
https://www.ti.com/lit/ds/symlink/bq27220.pdf

If you can't use that little BGA, maybe  BQ27441DRZR-G1A 4mm x 2.5mm?
https://www.digikey.com/en/products/detail/texas-instruments/BQ27441DRZR-G1A/4626408
https://www.ti.com/lit/ds/symlink/bq27441-g1.pdf

[ArdWar]

Just go up and down on TI's parametric table. You'll eventually find one in DSBGA/WCSLP package if you really need miniaturization.

[inse]

The bq27220 fuel gauge will consume a considerable amount of the CR2032s capacity.
Without monitoring, the battery lasts about a year, with monitoring you need to replace it after nine months ;-)

[Smokey]

The bq27220 fuel gauge will consume a considerable amount of the CR2032s capacity.
Without monitoring, the battery lasts about a year, with monitoring you need to replace it after nine months ;-)

Ya, but hopefully the 9 month replacement won't sneak up on you

[Psi]

Yeah, current consumption is a bit of an issue.  They usually draw too much.

[Doctorandus_P]

What about simply using a clock / calendar and replacing the battery at fixed intervals.

You can combine it with some dummy devices, that are rigged to have a slightly higher power consumption and that are only used to measure battery life. (Probably not very accurate because of temperature differences, spread in battery capacity (at least use the same brand / batch, and probably other factors).

[Peabody]

Is just measuring the cell voltage out of the question?  From the curves, it looks like a 2.8V reading tells you it's almost out of juice.

[Smokey]

Is just measuring the cell voltage out of the question?  From the curves, it looks like a 2.8V reading tells you it's almost out of juice.

I would imagine at such low currents you are essentially measuring no-load voltage which doesn't tell you much about remaining capacity. 

[Peabody]

Well I guess it depends on what you're going to do with the information, and how much of the cell's capacity you're willing to expend in getting the information.

[thm_w]

Is just measuring the cell voltage out of the question?  From the curves, it looks like a 2.8V reading tells you it's almost out of juice.

I would imagine at such low currents you are essentially measuring no-load voltage which doesn't tell you much about remaining capacity.

Nothing would stop you from loading the cell and measuring the voltage if you wanted it to be more accurate right?
I don't know what the minimum measurement time would be, some ms I guess.

[SiliconWizard]

Needs to be smaller.  Think wearable with tiny PCB.   So like SOT32 / 3x3mm DFN package

Yeah, this only comes in one package that will take about 5x7mm on PCB pads included. It's not huge either. Unfortunately, so far I haven't found anything similar in a smaller package.

Beware that this one is specifically made to monitor primary litihum cells. The BQ27220 mentioned by someone else, and many other such ICs, are made for Li-ion batteries only. This is not the same estimation algorithms at all. Maybe you can still use the latter if you set it up properly or only use a subset of its features, but that'll require reading datasheets very carefully.

The one reference I had otherwise in mind, which is *only* a "coulomb counter" and so would work pretty much with anything that's in the right voltage range is the LTC4150 that i've used before. But it draws too much current for your requirements, and is definitely not very accurate for small current draws.

[inse]

If the battery monitor is ten times the price of a battery (BQ35100), I doubt it makes sense economically…

[Berni]

It doesn't make sense to use a fuel gauge IC for this.

Those chips are just MCUs that keep measuring the battery current and voltage using an ADC and then plugging that into some fancy algorithm to come up with a SOC in %

Whatever you are powering has to make extensive use of sleep modes to last this long, so just build the coulomb counting into that MCU that runs the show. They consume a pretty predictable amount of current when running and they can time themselves with a timer, multiply the two together and you got the energy used. If you have extra loads being run by the battery you could use the MCUs own ADC to measure the current, but only doing it while you are actually running. When in sleep assume the consumption is 0 by making your sleep mode consume <1uA

Once you get into the last 1/8th of the battery life you can switch to using voltage as a battery indicator (since even fully accurate counting wont be precisely accurate for SOC because not all coin cells are made equal and differ slightly in capacity)

[Psi]

hm.. I wonder how small one could build a circuit that reads the voltage difference across a current sense resistor and translates that to rate of charge into a capacitor. The more current through resistor the faster the cap charges. Then have it trigger an interrupt on the MCU when it reaches a maximum cap voltage. MCU wakes up, increments a counter and discharges the cap back to zero then goes back to sleep.  An advantage doing it this way would be that it's all done as analog circuity so it should handle pulsed loads like BLE very well.

Just a thought, probably not very practical, accurate or low power.

[Berni]

Yeah the precision analog might be more power hungry than the device. Things like BLE can be incredibly low power.

If you wanted to go that way it might make more sense to charge up a capacitor and then run from it until woken up by a undervoltage, turn on a FET to charge it up and repeat. The downside is that you expose the coin cell to short high peak current draw pulses, while they prefer to have a slow constant draw for longest life.

But if you use something like a Nordic nRF, you have the same CPU handling bluetooth as can run user code, so you likely can hook into the wakeups for BLE and time them to count the used charge.

[tszaboo]

and probably need to be under 50c in volume.
Max Current maybe 80mA or so for bursts.

I'm looking myself but have not found anything suited to this.

I think coulomb counters are considered a premium feature and TI/AD seems to charge a lot for them.
But if you find one I'm also interested.

[bookaboo]

We measure the voltage of a CR2032 to give a single point "low battery" status in an alarm product. Granted there's no indication of actual remaining capacity (which perhaps is a a requirement)  just a "good" and "low" status but it's working perfectly in a very mature high volume product.

The measurement is done just after the ISM band transmits at ~10mA, so that helps give it some load.
A pretty simple hysteresis algorithm is all that was needed, we chose a level that we figured gave the customer ~1-2 months to change the battery (~3-5 year expected battery life).
Yes it's crude, but the engineering was simple, it's zero component cost and negligible power cost. 

[inse]

Another thing to consider: as the battery is replaceable, how would you know the exact initial capacity?
As different brands could be used or the battery is not factory fresh anymore.
Counting the coulombs would not make sense then..
It all falls back to measuring the cell voltage and making best use of the information.