Coin cell performance in low temperatures

[electr_peter]

I made some detailed measurements of remote control activity during transmission. Setup: external PSU set to 3.2V at 70mA. Currents is measured with scopemeter in series (mA range, ~1.6 Ohm, insignificant voltage drop).

Remote transmits (also beeps and lights LEDs) for ~975ms at 48mA with 3.2V supply. See screenshots 1, 2, 3 below. This is 150mW power and 0.042 mWh or 0.013 mAh energy. CR2032 capacity is ~200 mAh, so in theory it could last for 15k times (excluding idle drain).

When measured with CR2032, voltage on idle was 3.04V and 2.64V during transmission. See screenshot 4. Voltage drop is ~0.4V, which gives 8.3 Ohm internal resistance (assuming it reaches 48mA with partially discharged CR2032). This is a ballpark of CR2032 IR according to datasheets (~10 Ohm IR).

Conclusion: remote control requires ~50mA short current pulses and it really pushes CR2032 possibilities. This likely creates problems with any temperature variations. LEDs and buzzer on board are pretty weak, RF and MCU are most likely power hogs.

[richard.cs]

Yes, this is a lot of power for a CR2032, and too much to buffer with a capacitor. Most new, good-quality CR2032s will manage it at room temperature, but part-discharged ones with higher internal resistance may not, and even new ones will struggle in the cold.

It's not really a good design. I wonder how much power you'd save disabling the LEDs, or if a larger cell could be made to fit.

[electr_peter]

Yes, this is a lot of power for a CR2032, and too much to buffer with a capacitor. Most new, good-quality CR2032s will manage it at room temperature, but part-discharged ones with higher internal resistance may not, and even new ones will struggle in the cold.

Larger cap would do nothing with such long pulse duration. Extra battery could improve things if there was enough space.

It's not really a good design. I wonder how much power you'd save disabling the LEDs, or if a larger cell could be made to fit.

LEDs and buzzer were not measured. I do not see how one little SMD LED would consume much current, maybe buzzer is too powerful? I will look into it.
Battery is held in soldered cage, case limits battery size as is (or would require significant modifications).

[richard.cs]

It wouldn't surprise me if the LED were driven at 10-15 mA if it is meant to be visible outside the case in strong sunlight. A 20% load reduction would be significant.

[Siwastaja]

Remote transmits (also beeps and lights LEDs) for ~975ms at 48mA with 3.2V supply. See screenshots 1, 2, 3 below. This is 150mW power and 0.042 mWh or 0.013 mAh energy.

Because F = As / V, if you allow 0.5V voltage drop during the peak, required capacitor would be
0.013mAh * 3600 s/h / 0.5V = 93600µF.

Not going to happen.

Standard electrolytic, tantalum and high energy density MLCC capacitors are good for peaks up to some 1ms.

Now if this was 1ms instead of a full second, a 100µF 4V part would work. This would still be smaller than the battery itself.

This is why many communication devices really require low-impedance battery sources, unless the packets are really short. Power density of capacitors is great, but energy density is like 5-6 orders of magnitude behind batteries.

Still people often solder capacitors as cargo cult attempts to solve these issues, without performing simple napkin calculation first. The result is obviously disappointing. For the same reason, you may see unpopulated capacitor footprints on the PCBs.

[electr_peter]

I isolated power usage from LED and buzzer on PCB. LED draws <1 mA. Buzzer (piezoelectric disk element) draws 4-6 mA max, so there is no point in disabling it.
System architecture is pretty complex with 2 main chips onboard - CC110L (TI Sub-1 GHz wireless transceiver) and STM8L151G6 (ST 8 bit MCU, 16 MHz).
Only options to increase battery performance are to add 2nd battery or upgrade CR2032 to something bigger.

[Siwastaja]

Redesign of the firmware would likely be the right thing to do, why does it have to transmit non-stop for a second, is this transferring tens of kilobytes or what? But obviously, this isn't a realistic option.

Software is everything in low-power wireless. Do the minimum required, in minimum time, and go to sleep.

[cybermaus]

- measuring a coin cell voltage with a DMM is pretty useless there’s some load because of the 10Meg input resistance. If you add a little discharge resistor in parallel with your DMM  (10k - 100k) you’ll get some more useful results

Yeah, that's why I have one of those cheap dedicated battery testers. They do put a small load on the battery while testing. Can't remember for sure, but I believe it was like 10K
Also more convenient to hold when you have only 2 hands like me (as opposed to holding two probes, a dmm and a battery)

Anyway, so far the 106 capacitor I added works. Already 2 frozen morning without problems.
Now I only hope the leakage of the capacitor leakage does not drain the battery too soon.