After seeing some questionable performance from our new LiFePO4 battery array at Rally Creek we decided we need to know more about how our system is working. We quickly uncovered an undersized charge controller for our 800W panels being capped at 400W by our charge controller, but even with that limitation our runtime was much shorter than expected. The readings on our batteries were telling us we should have plenty of power for our minisplit from our 10 kWh LiFePO4 battery array to run for 6-hours straight. Instead we say maybe 2 hours maximum runtime.
After these results we ran various on-site tests with each battery in isolation. Still not seeing consistent results, we decided to do some tests in a more controlled environment. We’ve temporarily put the lead-based AGM batteries back online at Rally Creek and took the LiFePO4 battery array back to our home base in Charleston for more tests in a controlled environment.
We are testing in something closer to a real-world scenario but are using an AC-based charge controller (HTRC P20) at 20 amps to have a steady charge to fill the batteries. We then use a Ryboi 1,000W modified sine wave inverter connected to a variable-setting digital space heater to provide a near constant load for discharge testing. We added a Renogy 500A Battery Monitor with Shunt so we could have more precise battery load readings. We do NOT charge and discharge at the same time and ensure any equipment related to either process is disconnected when doing our tests.
E-Lektech 12V 300Ah LiFePO4 Takeaways
- The on-battery BMS percentage state of charge (SOC %) readings are wildly inaccurate.
When the battery has been at rest for a few hours it is usually almost accurate, but even that proves to questionable. - The inverter (or any other load) will shut down BEFORE the battery reaches 0%
While under load the BMS as well as the shunt, which means any intelligent inverter or charge controller as well, will get a LOWER VOLTAGE READING than when all the equipment shuts down and the battery comes back to a rest state. - It takes a LOT MORE than 300Ah of power to fill up a 300Ah battery.
Filling the battery from 10% to 70% goes fairly quickly, you push 20 amps into the battery for an hour, you get almost exactly 20Ah of usable power.
However at 88% charge, the increase in power needed to add just 1 more amp-hour to the battery is notable, taking TWICE as much power into the battery to get that 1 extra amp house. - You will NOT get to use the full 100% of the battery in an efficient manner.
Under load you cannot access that last 5% of the battery, probably good as you really want to stay above 10% and preferably 20% SoC for long term battery health.
When the battery is over 85% you need a LOT MORE power pushed to the battery to get usable power back out.
E-Lektech 12V 300Ah LiFePO4 Inaccurate BMS
The battery management system (BMS) , shunts, and from what I can tell most devices that give a state of charge (SoC) value for a battery base the estimate off of the voltage reading on the battery. Take a battery with no load, put a multimeter on it and the voltage returned should give a good estimate of the state of charge — how full the battery is.
Turns out the built-in BMS has what we assume is a low-end version of this and appears to be wildly inaccurate. Our Renogy Battery Shunt seems to be far more accurate and gave pretty much identical voltage readings as our multimeter readings. In addition, the shunt’s corresponding state of charge reading was very close to the SoC/Voltage charge provided to us by E-Lektech.
Here is the E-Lektech SoC/Voltage chart:
| State of Charge (SoC) | Voltage (V) |
|---|---|
| 0% | 11.60 |
| 5% | 11.80 |
| 10% | 12.00 |
| 15% | 12.20 |
| 20% | 12.40 |
| 25% | 12.60 |
| 30% | 12.80 |
| 35% | 12.83 |
| 40% | 12.86 |
| 45% | 12.89 |
| 50% | 12.92 |
| 55% | 12.98 |
| 60% | 13.04 |
| 65% | 13.07 |
| 70% | 13.10 |
| 75% | 13.13 |
| 80% | 13.16 |
| 85% | 13.24 |
| 90% | 13.32 |
| 95% | 13.38 |
| 100% | 13.44 |
Here is how the BMS compared to the Shunt (and multimeter, which showed identical readings but with 2-points of precision) during various no-load charge/discharge readings.
| Load | BMS | Shunt | Variance |
|---|---|---|---|
| Low | 100% | 100% | 0% |
| High | 100% | 100% | 0% |
| Low | 99% | 99% | 0% |
| – | 90% | 99% | -9% |
| – | 93% | 99% | -6% |
| High | 93% | 99% | -6% |
| High | 53% | 95% | -42% |
| High | 53% | 83% | -30% |
| Low | 77% | 83% | -6% |
| Low | 82% | 83% | -1% |
| Low | 93% | 83% | 10% |
| High | 85% | 83% | 2% |
| High | 48% | 60% | -12% |
| High | 38% | 45% | -7% |
| High | 29% | 35% | -6% |
| High | 23% | 20% | 3% |
| High | 20% | 13% | 7% |
| High | 18% | 12% | 6% |
| Low | 16% | 9% | 7% |
| – | 25% | 9% | 16% |
| High | 24% | 9% | 15% |
| High | 24% | 8% | 16% |
| – | 12% | 8% | 4% |
| Low | 20% | 8% | 12% |
As you can see in the chart above, the Battery Shunt (blue columns) showed a predictable and expected decline in state of charge as our load discharged the battery over time. It is predictable and what we would expect including the plateaus where we rest the system to allow the voltage to stabilize and get a “clean” reading of battery voltage and charge.
In comparison, the BMS reporting is all over the place. At one point it reads a precipitous drop from 93% to nearly 53% almost immediately. Then it stays there for over an hour while drawing over 800W from a 3600W battery. The BMS then corrects itself somewhat during rest, but then way overcorrects showing 93% SoC when in reality we should be closer to 80%. Even after our final discharge the BMS starts out reading 24% when the shunt and multimeter under load show 11.9 volts (around 8%). It correct down to 12% when the system shuts down and there is no load, but then after some rest and with the shunt activated to take a reading corrects up to 20%.
This low end is where things get interesting. We talk about that next.
E-Lektech 12V 300Ah LiFePO4 Early Shut Down
We can see above that the state of charge readings on the BMS are fairly useless. We are not going to look at that any longer as our charging tests showed similar results. The external charger, the shunt, the multimeter all give far better real-world insight into what is going on inside the world of electrons inside the battery. However, this leads to another interesting situation — low end performance of the battery.
When under load, especially the high 800W+ draw while running our tests, the voltage readings are FAR DIFFERENT than when the battery has no load and is at rest. Turns out the BMS and/or the shunt will shut down the circuit when the voltage drops too low. The inverters, both our Ryobi test version as well as the Rally Creek installed Renogy 2000W Pure Sine Wave Inverter, also seem to shut down as soon as they reach a low voltage threshold. No matter which device mashes the “kill switch” on the system, one thing is clear — under load you will not get the battery down to less than 12.0 volts using any type of semi-intelligent equipment (shunts, inverters, and such). Maybe, if we directly connect a 12 volt fan or LED light strip the BMS will let that trickle the system down to 11.6 volts, but we are talking real-world usage and for us that means using an inverter (and a shunt from now on) which will stop us from killing the battery by truly draining it to zero.
That means in the real world, under load, the system (inverter or shunt) will get a reading of < 11.8 volts from the battery. It shuts everything down. That’s a good thing.
Immediately after the shut down, the BMS, the shunt, and the multimeter will show a higher voltage. I like to think of it as if it were and old-school gas gauge on the 1980s car I drove for a while. If you stopped to fast with a low tank of gas, the gauge would read empty then slosh around a bit and eventually settle showing you have a bit more than empty in the tank… just enough to make it to the station.
In repeated tests, our system would shut down, inverter staying off then after a minute the BMS, Shunt, and our multimeter would show there was 12.3 volts in the battery. That is around 18% SoC according to our chart. We would turn the system on, voltage would immediately drop… 11.9 volts, 11.8 volts… done. System shuts down a minute later. The voltage would quickly recover to 12.3 volts.
That means we can really only draw down the battery to around 18% SoC. We could tweak our shunt down to 11.6 volt shut down or lower and risk straining the battery (the BMS is supposed to prevent that) but we know for certain the Renogy 2000W inverter will shut down at around 11.8 volts anyway and our Ryobi stand in will likely do the same. Even if the BMS would allow us to draw the battery down to 11.6 volts (0%) in reality, once the load shuts off the BMS will likely read at least 12% not 0% as the voltage will almost certainly creep back up to 12+ volts after a brief rest.
Bottom line – in the real world our floor is around 12.3 volts or 18% state of charge as the maximum discharge level.
Better for battery health anyway, but definitely need to adjust our formula that indicates you can use 90% of the battery with LiFePO4.
So our usable battery capacity is 100% – 18% or 82% of the rated capacity. Which brings us to another issue, reaching 100% state of charge.
E-Lektech 12V 300Ah LiFePO4 Charging Rates
Turns out the charging of a LiFePO4 battery is not a linear process where you put in 20 amps in an hour and you later have 20 amp-hours to take out. Not even close. It turns out that as these batteries charge it takes a LOT MORE ENERGY to find places to put those electrons. In our testing we found that with this particular battery (and we are guessing most/all LiFePO4 batteries) do not charge at the same rate over time.
What we found in our charging tests is that this particular battery charges fairly evenly when the battery is at the lower charge state (40% or so in our real world test) up to about 78% state of charge**. At the lower state of charge , under 78% or so, we find we have an near 1:1 ratio of power in to power stored. At around 78% this has a notable uptick where we need to put in more than TWICE the power to get the same amount out. For example, put 40Ah in you get 20Ah stored for later use. When you hit around 88% this then increases exponentially and you have to put in FIVE TIMES the power, 100Ah in you get 20Ah to use later. As the battery gets closer to 100% charge you are putting in more than 10x the power you will get out. Thankfully this only last for a short time.
However – in the world of solar-driven off-grid power that brings up a very real issue. If your solar panels need to fill up a 12V 300Ah battery they are going to need to generate a LOT MORE than 3600 watts of power to do so. We’ve not finished the math yet and we are about to run some “depleted to full charge” tests to find out, but our estimate is looking like you’d need to push in closer to 600Ah of 12 volt power (14.4 volts if I recall properly to be technically correct), or around 7200 watts.
That means you are better off sizing your battery array so that you stay in the sweet spot of < 88% charge to make that solar work a lot less to push the electrons into the battery.
Bottom line : our top-end charge is really closer to 88% SoC, not 100%. That drops our “ceiling” to 88% and gives us a lot less power to play with.
** We used the BMS reading for much of our recording until we learned later how poor the BMS is at showing actual charge state — we will have more accurate reports in the future.
E-Lektech 12V 300Ah LiFePO4 Usable Energy
We are still testing and crunching numbers, but out there in real-world off-grid usage we are learning that our LiFePO4 batteries, at least this first one we’ve tested in detail, are able to serve up around 70% of their rated capacity (88% max charge from solar – 18% max discharge before shutdown). That means our E-Lektech 12V 300Ah LiFePO4 battery is expected to provide 210Ah of usable daily-use power when charging with our solar array. This is mostly due to the huge increase in power needed to get the battery to 100% charge.
Sure, there are days where we will have full sun and now use for the extra power so we might as well push more power into the battery to get it to 100%. Yes, it will generate waste heat, but in these cases we will have 82% of the rated capacity (246Ah versus 210Ah). However, if there is a cloud day or two there is going to be a whole lot less extra power to push that battery from 80% to 100%.
In face we can see many days where that power increase around the 78% SoC mark where it takes twice the power to charge the battery will be the ceiling. On these days we will only have 60% of the capacity available, around 180Ah. This is marginally better (by 10%) than our old-school and much cheaper AGM batteries.
In summary, it looks like the real-world usable power of LiFePO4 batteries like this E-Lektech battery is going to be between 60% on the low end and 70% on a typical day. On the best days, which will be rare, you might 82% capacity but I wouldn’t count on it.
Time to re-do some math on what it takes to run our cabin on solar + LiFePO4 batteries.
E-Lektech 12V 300Ah LiFePO4 Summary
In our real-world testing using an AC-charged battery thanks to our 12 volt HTRC charger, we were able to start from a 100% charge. After forcing the system to restart after automatic shut-down several times, we were able to draw an average load of 856 watts for just over 4 hours. Turns out that was damn near the full 3600 watt-hours of power the battery is rated for (12v @ 300Ah). If we didn’t force the restarts we would have been just under at closer to 3200 watts, and have enough battery capacity to extend it’s life.
That’s a respectable result and we need to update our comments and stance on the usable capacity of this battery. This battery can provide damn near the 3600 watt-hours they promise.
In everyday use we’d limit it to a 15% depth of discharge to improve longevity, but our top-end charge limit is going to remain a concern. As such we are going to perform our storage needs based on a 70% typical usage scenario = 210Ah at 12v or about 2.5 kWh.
That aside, E-Lektech needs to invest in a better BMS that reports more accurate state of charge as it led to our original misunderstanding of available power (the BMS might have said 40% SoC left when in reality it was closer to 20% — and thus why we were shocked our minisplit drawing 1kw shut down so quickly).
Discover more from Rally Creek USA
Subscribe to get the latest posts to your email.