Optimizing the shelf life of battery-powered devices

Background

For a few years now I've been placing small lithium batteries in a lot of devices, including a digital camera, a multi-meter, an RC car remote controller, a table vacuum cleaner, a few UPS etc.

Unexpectedly, one of the devices that required the most adjustments was the voltmeter:

I carry it everywhere in my computer bag and don't use it very often. As usual with lithium battery powered devices, it's difficult to have a gauge because the battery is connected to a regulated DC-DC converter so you discover that the battery is depleted when you need to use it. I drilled a hole to plug a micro-USB connector to recharge it but it's quite annoying to have to wait before using it:

So it was about time to investigate how to improve this.

Opening the assembly again

The voltmeter displays the "batt low" warning once the voltage goes below 7.00V, which never happens here. I thought about ways to improve that so that the output voltage starts to progressively drop when the battery is getting near its end, but that's not the point for now, as I've been pretty sure that the battery should last longer. Let's have a look inside. We find my good old battery with the DC-DC converter + charger I found back then, I don't even remember the module's name nor reference but similar ones are easy to find on Aliexpress or eBay. That one produces 9V from a single lithium battery and can be charged over micro-USB:

These are two separate chips, a classical TP4056 for the charge, and a B6287q for the conversion:

 

We already see a voltage divider made of a 10k resistor at the top (the one marked "1002") and a 140k at the bottom left (the one marked "1403"). That indeed gives 0.6V from 9V output (9*10/(10+140)). But... wait a minute, 150k serial on 9V should drain quite a bit of power, no ? 9/150k = 60 microamps, or 540 microwatts. Considering 80% conversion efficiency, that's roughly 675 microwatts drained from the 3.7V battery, or 182 microamps! That's sufficient to drain this entire 240mAh battery in 1300 hours or about 50 days. That's no much more than what I'm observing, it's drained in one month.

Let's verify by placing the ammeter in series between the battery and the module:

Bingo, 254 microamps drained from the battery! That's 40 days, roughly in line with my obvservations.

Solution

The solution sounds simple: let's just change the resistors. But these are already high values, it's not certain that the voltage regulator will operate correctly with larger ones. Also many voltage regulators still have a quiescent current that's a bit too high for such a use case.

 

After some research, I found the perfect chip: TI's TPS61040 and TPS61041 have a voltage input range from 1.8 to 6V that's compatible with single-cell lithium batteries, an output from Vin to 28V that's quite sufficient, a switching current of 400mA (or 250mA for the -41 variant), a very low quiescent current of 28µA and a feedback current of only 1µA at 1.3V which allows to use even larger resistors. And the best, it has the same pinout as the existing chip, it just requires to solder the top two pins together, which is not too hard :-)

At such very low current values it requires a bit of experimentation though, because the IFB isn't exactly 1.00µA but around 1µA. In practice I found that using a 2Meg resistor at the top and a 330k at the bottom of the divider were sufficient to provide approx 9V on output, thus very close to the theoretical value (the resistor drains 3.9µA which is still above the theoretical 1µA from the chip).

Tests

Let's assemble this and test again:

Bingo! The current was divided by 11! One will note that the measured value is even lower than the datasheet's quiescent current now.

Measuring the output voltage shows irregularities when the voltmeter is turned off, the output oscillates between 8 and 10V, with 10v peaks every few seconds. This seems to be caused by the low switching frequency: the pulses, despite being short, are sufficient to significantly refill the output capacitor. This does not happen when the voltmeter is turned on, so I don't care. The most important in fact is that it remains slightly above 7V so that the multi-meter always operates in the recommended range without needlessly draining current through the bleeding resistors.

After about two months, I found the voltmeter discharged again but this time it wouldn't appear to take the charge. I found that the regulator's output wouldn't work anymore, it was trying to emit 30V in open circuit and much less when connected, and the battery was totally depleted. I think I finally understood what happened. The battery was an unprotected one because the previous chip likely had an under-voltage protection. But this new chip does not, so it continues to drain current even when the battery is almost dead, and it very likely failed when entering very low voltages which draw more current.

I replaced the chip and used a protected battery instead. Now it might have been 3 months, I've used the multi-meter a few times and it's still operating fine this time.

Thus this time it looks like a solved issue!

Improvements

It would be nice to design a small circuit that would bias the feedback voltage so that a lower battery voltage increases the voltage on the FB pin and lowers the output. But when I tried this in the past it used to always drain significant current. It would be better done directly inside the chips, the battery-focused ones could instead modulate their internal reference in parallel to the battery voltage so that the output varies with the battery voltage, but varies less. On a DC-DC regulator with an external Vref pin it might be sufficient to connect a 200k-to-2M resistor between Vin and Vref to adjust the output voltage. This was the same problem with the digital camera and other devices. This would allow the multi-meter to report a battery low condition. Maybe these things will finally appear over time.

Conclusion

In conclusion, such a mod to replace a 9V battery is quite easy with these boards, it remains easy to recharge the device, and assuming the willingness to replace the SOT23-6 chip and two SMD resistors, the shelf-life after the conversion becomes basically as good as with the original battery. Even without the chip change, the results were very good (sufficient for other use cases).

Here it should last around 10k hours or more than a year without a recharge, not counting leakage, which is why it's not dramatic not to have a "battery low" indicator anymore. And I feel like I'm no longer annoyed with replacing batteries in such not frequently used devices.