Li Ion Battery Charging - Initial Results for Float Charging New LiIon Cell

Background and Method

The initial testing I undertook revealed where the shortcomings in my approach were. To bring more rigor to the testing I have started a trial as follows:

Take a pair of brand new batteries. One, let's denote it as x1, will be subject to the CC/CV regime at around 3.9V. The second, say x2, will serve as a reference. The capacity of x1 and x2 will be initially tested fully charged. 

I then put x1 in my CC/CV charger and left there until the capacity is tested, about fortnightly. After testing  x1 goes back into the CC/CV charger.

Meanwhile, after testing x2 will be charged to 3.6 volts before removal from the charger and stored. Every 6 months it will be then fully charged and tested before re-charging to 3.6 volts and storage.

Capacity testing will be by discharge to 3.0 volts. 

 

Initial Test Results after Two Months

Having established a base line for the cell capacity when fully charged I have now floated one cell at 3.9V for two months with 4 discharge tests during that period. After the third test where I had minor contact resistance issues I soldered a connector onto the cell for discharge testing for the fourth and onward tests.

I am yet to see any meaningful degradation in capacity when floating at 3.9V. I will do a full charge/discharge test for both the floated cell and the stored reference cell next month.

Inductor Sizing for Boost Convertor

When I reviewed the datasheet for the MT9284 boost converter I noted the suggested inductor size of 4.7uH to 22uH. The datasheet contemplates led currents of 20mA. I'm pushing the chip almost 10 times past the 20mA. 

Which begs the question what inductance should I be using? I did a series of measurements and after a lot of thought concluded the right amount of inductance is around 8.2-10uH. The following chart shows measured efficiency, with outliers removed, and two trend lines for 10uH and under, and 10uH and over. By inspection these trend lines intersect around 9uH. If I drop the 10uH observations the intersection of the trend lines shifts towards 8.2uH.

I've ordered some more inductors to confirm this is the sweet spot. 

This will be a good starting point for the lower loss boost converter I will be receiving in the same parts shipment.

The Rat Flasher - Efficiency Measurements for Boost Convertor

It all seemed so straight forward. Move from two cob leds in series to 4 cob leds orientated every 90 degrees around a case, each with their own boost convertor. How hard could that be?

Trivial, until you try to get every last drop out of the boost convertor. The boost converter data sheet never really envisaged hitting leds so hard for such a short pulse. So when I measured efficiency I found the absolute best I could get was 72% at which point I was hitting the cob led with 250mA with a power dissipation of 2.1W. 

Four of those flashing together was going to be really bright. However, that meant a peak supply current of around 4 x 1 Amp, or 4 Amps.  And a 4 Amp surge meant I had to add larger decoupling capacitors and consider the micro browning out from the voltage dip.

I found the efficiency was dependent on the battery voltage. A drop of 0.5V meant efficiency fell from 72% to 64%. Which means I will have to revisit the PCB at some point to use some bigger traces and eliminate the headers used to connect the battery.  A voltage drop of 0.5V when drawing 4A is just 125mΩ!

A slight tweak in the code means it is now easier to lengthen the time between sets of pulses. The rat flasher had not been in the ceiling for 3 nights, and no vermin had returned, so I applied a cautious tweak. With the adjusted timing I expect I can get 3 nights of solid flashing before the LiFePO4 battery goes flat. Over 24 hours the average current drawn should be around 40mA. I have a solar cell ready to go on the roof which should re-charge the battery during sunny days. With the arrival of spring there will be plenty of sunshine.

Longer term I need to move to a better boost converter. The boost converter this project evolved from uses a NMOS switch with a Rds of 0.5Ω. I expect a step change in efficiency by moving to a boost convertor with on resistance of 80mΩ. And a revised circuit that essentially charges a large capacitor and dumps that through the leds at much higher current, and hence brightness. 

There's no point in showing you a circuit of how not to do this. Another iteration is warranted so please be patient. If your desperate to drive those critters out of your roof and are happy with a less than optimal solution then by all means contact me.