Bluetooth low energy and coin cells II


I wanted to follow up on my earlier post "Bluetooth low energy and coin cells", as we have now finished our study on coin cells and published our results. The results can be found in our white paper "Coin cells and peak current draw", which is also accessible through the CC2540 home page. 


To summarize, we have run extensive lab tests on CR2032 coin cells (the most commonly used kind of coin cells for these applications) over the last 9 months, and our results indicate that there is no sigificant difference in effective capacity between a peak current draw of 15 mA and 30 mA. We have also tested batteries from several different manufacturers, and it is clear that there were some significant differences in effective capacity between brands.


Some background: many battery manufacturers do not specify a capacity for pulsed operation at all, just for constant current draw. This is usually provided for a very low current, a few mA at most. What we term the effective capacity is the actual capacity measured (assuming an end-of-life voltage of 2V) with a current profile resembling a worst-case BLE transaction. These tests were done without any form of capacitor or decopling, in order to measure solely the performance of the battery itself. The batteries from the best manufacturer achieved the same capacity in this scenario as the capacity rated for constant current (220 mAh), while the worst "no-name" batteries only achieved around 50% of the constant current capacity. No significant difference in capacity between 15 mA and 30 mA maximum peaks were found for any of the cells. Our tests were run on a relatively large number of batteries, as there is quite a lot of cell-to-cell variation (which is to be expected, there are large tolerances in capacitance values for large capacitors as well, for instance). The numbers we provide are averages.


What actually happens is that the internal resistance of the battery increases as it gets discharged. This results in a larger and larger voltage drop over time. Once this resistance becomes too big, the battery voltage drops under 2.0V under maximum load and we declare the battery to be dead as at that point it would trigger the brown-out detector on our IC. Adding a decoupling capacitor improves the performance of the poorer cells quite significantly, as it effectively decreases the battery impedance (just as in conventional decoupling use). A capacitor increases the effective capacity of the poorer-performing cells by up to 50%.


The white-paper also includes formulas to help you calculate the right size capacitor for your application.


The implication for users designing coin-cell powered RF devices (either BLE or other RF standards) is that you need to de-rate the battery capacity somewhat depending on whether you have full control of the battery selection (sealed/disposable devices) or not, and then you can calculate battery life using the average current. You don't need to worry about the peak current consumption (there may be a limit, but it is certainly bigger than 30 mA for CR2032). For average current and battery life estimates, we have put together an appnote and accompanying Excel spreadsheet to help you calculate battery life. This is based on actual measurements of the CC2540 (including power consumed by the BLE stack processing). We have seen some theoretical figures based on radio current only that are not representative of what is actually possible using BLE. Right now, the spreadsheet covers only connected operation, but we are planning to extend it to cover advertising as well later.