LiFePO4 for small projects – crazy connections
mrburnette Sat Jul 11, 2015 5:18 pm
WARNING: DANGEROUS EXPLOSIONS HAVE BEEN REPORTED BY MANY WHEN CHARGING MORE COMMON LiON BATTERIES.
I have a couple of solar floodlights in the yard, one in the front to illuminate the Security Firm sign and one in my wife’s garden to illuminate her prized porcelain cat (do not ask.) These have worked flawlessly for a couple of years, but the garden unit failed to recover after the long, harsh winter. My options were to replace the unit in total ($19 USD) or purchase replacement batteries from China. I decided on the battery replacement idea and brought the unit to the lab to identify the battery type… I’m thinking NiCd… but no, it was LiFePO4.
The LiFePO4 cell was unknown to me, so I did some research. I went to eBay and got a great deal on 2x 5-pack of replacement batteries; about $20 including shipping. Eventually, the batteries came and I used one for the outside LED lamp. I’ve forgotten about the other 9 until this week. So, I read some more about the chemistry and the experiences others have had with this cell.
First, the common AA form factor (14500) battery is stenciled with 3.2V and 600mAh; older versions will likely say 400mAh. Now, this is nothing compared to the 3A+ of the 18650’s and therefore does not have internal protection as other hi-current lithium cells. That is a plus. The other thing I noted in the datasheets is that the cells can take a significant amount of abuse. But the most interesting concept that I read was that the cell can be charged to 80% to 90% capacity by directly float-charging on the 3.3V bus.
Let me repeat the meat of the above: LiFePO4 bridging the 3.3V bus… no resistors, no fuses, just the battery looking like a capacitor.
I experimented. Here is what I have experimentally concluded:
Yes, the battery can be float charged. But not at 3.3V, my results show a voltage closer to 3.45V is required to prevent the battery from discharging while the circuit is working (must be tested for every unique circuit.) As I am working with the ESP8266, that is acceptable since the specs are for 3.6V as the upper ceiling for the module. As I am on a breadboard, the interconnects do have some resistance, so the above voltage will need to be adjusted in an actual build-out.
The LiFePO4 cell acts like a huge capacitor! Monitoring the charging current, the lithium cell will often swing to a lower charging current when the ESP8266 current demands go up due to WiFi activity. I’m witnessing a 10mA deflection (analog Simpson meter) with the current from the lab supply set on constant-current. This corresponds to approximately what I observed without the LiFePO4 in the circuit and the lab supply set for constant voltage 0f 3.45 volts. My Simpson meter has a 2.5 internal shunt resistance for the 0 – 100mA scale. At a float current of 5mA on the cell, the actual voltage on the uC and OLED display is: 3.45V. When I pull the lab supply feed, the battery current goes to approximately 70mA with peaks of 80mA – again using an analog meter movement – absolutely no transit issues with the uC… everything just continued to work with the 3.21 voltage on the uC from the lithium backup battery (remember there is a 2.5 Ohm meter shunt resistance inline with the battery while this test is conducted.)
Where is all of this going?
My plan at the moment is to use an inexpensive bulk PSU fed from the 5V of a USB jack. The adjustable Vcc can be set with a variable resistor. The voltage will be adjusted to 3.45V. Once adjusted, a drop or finger-nail polish will hold the adjustment against accidental changes.
I will utilize a LiFePO4 as a supercapacitor; that is, directly wire the battery across the bulk-psu and then use a switch from this point – that is, if the unit is plugged into USB, the battery will charge and then eventually come to approximately 5mA float current.
This will provide me a means to charge the cell and to unplug from the USB and go out and about with approximately 0.8(600/80)hr = 6 hours of use before a recharge is required.
I am an hour into a 4 hour test cycle, the battery was left on charge (3.45V) until the current was near 0; then the lab supply was disconnected with the battery picking up instantly. The ‘8266 is scanning and displaying on the OLED every 10 seconds, so current peaks 6 times a minute from the nominal 70mA. In 3 hours, I will again connect the lab charger (to simulate connection to a USB 5.0 volt + buck switcher.)
Ray
