Solar Battery Blues

The IoT Solar Battery Voltmeter project resulted in a dead battery. It now does not accept a charge, and its voltage dropped below 12V very quickly when discharging, indeed when it was not being charged. There is nothing else to do but to replace it.

20 minutes after sundown the battery voltage went from 14V to 10.9V

Now before I get too misleading, it is not a good idea to use a car battery in place of a deep-cycle battery. I started experimenting doing just that around 12 years ago mainly because they cost about half of the deep cycle ones.

3KW sine-wave UPS APC Matrix 3000 hacked to accept 4 36Ah Yuasa 50B24R

The idea was to over-rate the car batteries and hopefully they will then have a worthwhile lifetime and cost. The reason for this is all cars have a residual load on the battery when not in use. This comes from the digital clock, stereo, alarm and the various car engine electronic modules. It is thought the load here should be 85mA for a new model and 50mA for an old model.

So in general you cannot replace a deep-cycle battery with a car battery, unless the load is less than 100mA.

After all these years progress have overtaken the lead acid car battery. I recently bought a 20Ah (at 3.7V) lithium-ion power bank for RM87. This is somewhat comparable to my 36Ah (but at a stonking 12V)  Yuasa at RM195. Now lithium-ion batteries are deep-cycle so they are definitely worth investigating.

This lithium-ion power bank is rated 20Ah25V and costs RM87

After all this is why I built the IoT Solar Battery monitor, to investigate batteries and their chargers.

So, apart from obvious misuse, exactly what went wrong? The load was only 200mA, and Yuasa 50B24R ECO-R GS car battery, it should be more than enough. 50B24R is nominally 36 Ampere-hours. If it completely discharged (again, not a good thing) it could output 3A over the 12 hours of a tropical night.

The actual current draw was 0.2A at 5V. This should translate to 0.083A at 12V. Assuming the 12V to 5V conversion efficiency of 75%, we can downrate it to 0.111A at 12V. At 111mA the battery should last 324 hours or 13 days. So the battery is way over-sized and yet lasted only nine months much less than the expected 3 years lifetime I get from the same battery in my car.

The problem could also be in previous tests on the solar battery I had completely discharged it 3 or 4 times. Now car batteries do not like to be over-discharged and they will fail prematurely. So, the 36Ah Yuasa 50B24R was right there at the edge, and the previous deep discharges may have tipped it over.

A simpler cause can be that the solar panel's charging rate could not keep up with the discharge rate at night, and over a few days the battery progressively got weaker and eventually completely discharged. The solar panel will put out a variable charge depending on the weather and we have had unseasonably rainy weather recently. Indeed the reason to build the IoT Solar Battery Voltmeter was to investigate this.

It could also be that my Gamma solar charge controller was optimized for deep-cycle batteries and somehow it damaged the car battery by overcharging. One clue is that the battery charging voltage is often 14 to 14.8V, clearly over the 13.8V trickle charge limit for a fully-charged battery.

In any case, the IoT Voltmeter consumed too much of the battery charge it was monitoring (this left only 30mA for the load!), and we could do with a leaner voltmeter.

Of course we could power the voltmeter separately using another battery, or even from mains power. Mains power would have been ideal, except we live on top of a little hill that seems to get struck by lightning a lot, and to get the weeks-long data we need, an isolated system like a battery-powered voltmeter seems like a good idea.

Alternatively we can try using inductively-coupled wireless charging like the Qi, often used to charge smartphones.

The Qi wireless charger & receiver combination can deliver some 500mA at 5V, enough to power the Raspberry Pi-based IoT Voltmeter

Indeed we will try both ideas, but it seems wise to try to reduce the power consumption of the Voltmeter. This brings us to the PIC18F14K50-based voltmeter minus the Raspberry Pi. The PIC18F14K50 USB RS485 board I used consumed only 10mA at 5V. This should give me a maximum load of 90mA for a N60 car battery.

The PIC18F14K50 converted to transmit its data via RS-485 takes only 10mA..

Now I can get it to transmit data back using its RS-485 port, indeed I have already done so, but RS-485 involved copper wiring back to my webserver and this broke the mains isolation.

To maintain isolation I need a wireless data transmission method, and it seems like a good idea to try Bluetooth. There is an existing product for sale that will do exactly that, but the price is a little high:

The CTEX CTX bluetooth battery monitor costs RM305

The Arduino HC-06-S bluetooth slave is only RM15

 Besides, rolling your own is not only cheap, but might be fun. Happy trails.