Since some LoRaWAN sensor products use Li-SOCl2 (lithium thionyl chloride) and some hobbyists use these too: any experience out there which brand is best, for AA size (“14500”) cells?
I’ve mostly seen Saft LS14500 and Eve ER14505 being used, and I often see Tadiran in search results when looking for these batteries. I was hoping to find some first hand experience on which ones last longest in a LoRaWAN device, for example if one of them is better at dealing with the relatively high peak current draw during a LoRaWAN transmission, towards the end of the battery life.
We sell quite a lot of devices with Eve, Tadiran and recently Xeno. Most of the AA and 1/2 AA types use Eve, and we’ve always had good success with them.
There seem to be a couple of different designs, some are more suitable for higher currents but then have slightly reduced capacity. For devices with AA or smaller batteries I would recommend building in a large cap to cover your transmission pulses, it’s more gentle on the battery. This technique is used by the better sensors that I’ve seen.
IIRC from when I looked at LiThyCl batteries for metering and other apps a fair time back -2010/12? I believe Panasonic was another brand that came up. Haven’t looked lately to see if still doing these. This battery class is great for long standby apps where device in a deep sleep mode sipping gently and occasionally (heartbeat?) and where ultimately internal self discharge is the key issue/dominating factor…leaving charge available to be sipped gently or not at all until really needed far in the future. Again IIRC these where good for 15-20 years and potentially as long as 25 years. Problem was if used for a radio txt is the peak current can run to mA’s even many 100’s/100’s mA and this quickly ‘fatigues’ the battery & exhausts the chemistry and if tx period too often then there is insufficient recovery time leaving battery to deliver just a fraction of the advertised capacity :-(. The solution as Al suggests is to add a large cap to buffer…I would suggest a short c.r.c network to smooth out current pull not just a bulk capacitor. To that end large in this context should be thought of as e.g. a Supercap. Indeed I believe Tadiran whom you have already called out also offer batteries with SC’s built in. Such products can get chuffing expensive so be sure of real need and analyse performance carefully. Again from looking at a different requirement/project around 2014/15 it was realised that the peak pull and duty cycle was too high to benefit from the low internal discharge rate as the system would exhaust long before the life benefit of LIThyCl could be realised so paying over the odds for a theoretical benefit didn’t make sense…do a lifecycle cost benefit analysis and compare to other battery techs along side any duty cycle analysis.
I recently opened up a TBMS motion sensor. It contains 2 supercapacitors (in series I think) with a capacity of 1.2F each with a rated voltage of 2.7V. The battery is a 3.6V lithium thionyl chloride in 1/2AA form factor.
Attach a load resistor of a known value and see how long it takes for the battery to go flat ?
If the voltage of the battery is stable for most of its life, then to know how much capacity is remaining, you would need to measure how much current\power the circuit is actually drawing from the battery over time, which is not easy at all.
Yes, I am aware of that. But there are too many lorawan sensors on the market that work with li-socl2. I had the idea to make an estimation by counting the working time, but I’m not sure how successful it would be.
Seems as good a way as any. Get a device, power it any way you like, measure it’s consumption for as long as you can stand, figure out the average sleep, active & tx consumption, add the three, et voila. So if you end up with 50uA/h per cycle, count the cycles and you can make the rash assumption how much battery may be left.
And it only measures how many coulombs have been used, which is only of value if a system was started from a known point, i.e with a battery of know capacity…
I think I’ve read that the voltage remains virtually constant to the end of the life of the cell. At this point the discharge curve becomes very steep. A 3% drop in voltage indicates that the cell is near end of life and should be replaced.
Take a look at recommendations by Qoitech. And even then take exceptional circumstances into account. For instance V3 sends downlink requests that might cause additional transmissions which at higher SFs take a lot of power.
Anecdotal, recently I modified firmware to allow for acknowledged uplinks once every X packets and forced things so the node would not get acknowledgments. Where the node used 20 micro watts in an hour before it now started retransmitting at increasing spreading factors resulting in it burning 50 milliwatt in one hour. That is a huge increase in consumption and a huge decrease in potential battery life.
Thank you. I know these. I know about battery optimization, but measuring li-socl2 is very difficult. Maybe I need to switch to LiFeS2 type batteries. It has high power capacity and a linear voltage graph.