LiFePo4 Power -- Controlling Alternator and Inverter/Charger

[Delfin]

Originally Posted by Dockhead

If what he says is true, then we should be able to measure SOC accurately using voltage, no?


And if it's true, then indeed I guess you could pick any desired SOC, set the "float" voltage to the corresponding voltage, and the battery would drift down towards that SOC and then be held there.

Not so much. Mine will read 26.2 volts from about 80% SoC to around 30% under modest load. There is no way for me to know what the SoC is from voltage, except when full or below 30%.

[Inkog]

This is a chart from Battleborn for their 100Ah LiFePo4 Drop-ins showing SOC as voltage while charging and discharging at a specific amperage. I've found it highly accurate as compared to my Victron 712 battery monitor.

I don't have any experience running a 24v LFP bank, but commonsense tells me this could be plotted for any nominal voltage.

You would divide the discharge/charge amperage by your bank size. I would think calculating for a 24V system would be even more accurate at the resolution of a tenth volt.

[Delfin]

Originally Posted by Inkog

This is a chart from Battleborn for their 100Ah LiFePo4 Drop-ins showing SOC as voltage while charging and discharging at a specific amperage. I've found it highly accurate as compared to my Victron 712 battery monitor.

I don't have any experience running a 24v LFP bank, but commonsense tells me this could be plotted for any nominal voltage.

You would divide the discharge/charge amperage by your bank size. I would think calculating for a 24V system would be even more accurate at the resolution of a tenth volt.

The discharge figures make some kind of sense. Battleborn can't differentiate the difference between 80% and 40% SoC based on voltage, while Lithionics says between 80% and 30%.

Given the variability of discharge loads in the real world, I suspect the "Terra incognito" SoC based on voltage is a bit wider than this. But perhaps sitting at the dock the loading is stable and you could set float to maintain a SoC give or take 40% capacity, more or less.

[Inkog]

Originally Posted by Delfin

The discharge figures make some kind of sense. Battleborn can't differentiate the difference between 80% and 40% SoC based on voltage, while Lithionics says between 80% and 30%.

Given the variability of discharge loads in the real world, I suspect the "Terra incognito" SoC based on voltage is a bit wider than this. But perhaps sitting at the dock the loading is stable and you could set float to maintain a SoC give or take 40% capacity, more or less.

If you're just trying to use a charger to supply loads while not over overcharging, you'd simply set the float voltage to the battery resting voltage, or 13.4v (example 12v LiFePo4 battery above) Long term - you may want to drop that down to 90% or 13.3v in this case for longevity.

For example, I'd set my solar chargers to float at 13.4v, as the bank discharges every night, and my shore power charger to ~13.3v where it may be maintained there for long periods.

[nebster]

Originally Posted by Dockhead

If what he says is true, then we should be able to measure SOC accurately using voltage, no?

Indeed: you can estimate SOC surprisingly accurately during the flat part of the curve, especially if you have a higher voltage pack. The slope is something like 1.7mV per %SOC, per cell, between 20 and 80% SOC.
*
So for a 4s battery you should see roughly 7mV per %SOC. For my 16s battery it's around 27mV for each 1% SOC.

And if it's true, then indeed I guess you could pick any desired SOC, set the "float" voltage to the corresponding voltage, and the battery would drift down towards that SOC and then be held there. That sounds very good, actually. Is that what you do, Tanglewood?

Definitely, you can set a float that will achieve a given drift-down SOC, approximately. (A smarter float value would be current-compensated, but a fixed voltage is pretty good with relatively stable loads that are a small fraction the pack's capacity.)

Getting the exact resting SOC you want is very sensitive to how your charger measures the voltage (and from where). It's easier to dial in a float voltage over time with a coulomb counter to help than to try to get it precise a priori, because real batteries have hardware in the way that alters the measured voltages slightly.

[nebster]

Originally Posted by john61ct

At higher rates, the curve gets steeper, as the difference between "loaded" voltage and at-rest OCV is much greater.

The curve, which is approximately linear between 20% and 80% SOC, does not change slope between 0.01C and 0.25C. I don't have data for larger loads, but those are probably not relevant for most house batteries being discussed here.

Furthermore, to be clear, the voltage drop/rise away from the OCV is irrelevant in the linear regime. (It does increase at higher rates, but it does so in a proportional way that does not alter the slope of the curve.)

With a given bank at a particular SoH and very precise **and** accurate voltage measurements, it is possible to correlate V and SoC at various dis/charge rates, held steady.

That task is much easier if the volts at each SoC increment are measured OCV isolated and at rest, but the correspondence table will still change with different brands and models, and even with one bank as it ages.

The second sentence, bolded, does not match my experience.

In theory OCV is the gold standard for looking up SOC versus V. In real life, the "surface charge" on LFP sticks around so long that it biases the measured voltage toward either the charge or discharge curve for so long that it is impractical to get a reading.

Put another way, my battery will move from a neutral OCV to within a tiny fraction of the 0.05C load voltage, at the same SOC, in under 10 seconds after the load is applied. But it takes at least 30 minutes, and probably much longer but I got tired of waiting, for the battery to return to close to OCV.

The second clause of your sentence does not match my experience, either. I agree that the curve should change and be slightly different over time and from bank to bank. In practice, though, my results, after adjusting for measurement offsets and interconnect losses, almost precisely follow curves published by other owners who have prismatic LFP batteries and who have plotted voltage versus SOC at various rates.

In the linear regime between 20 and 80 SOC, then, I suspect most well-maintained cells are much stabler over time and manufacturer than you suggest. If you have direct evidence to contradict my personal measurements and observations, please do share.

[nebster]

Originally Posted by tanglewood

Got my LFP bank with all the same loads and charging, full is 54.4v, and empty gen start is at 49V. That’s a spread of 5.4v, or nearly double the LA spread. And the voltage doesn’t sag nearly as much as LA with bigger loads, and recovers right away. So I would argue that voltage is a good indicator of LFP SOC, and a better indicator than for LA, both of which run counter to popular lore.

Yep, these mirror my experience, except I use a little bit narrower (more conservative) voltage range on typical days.

[nebster]

Originally Posted by Delfin

Not so much. Mine will read 26.2 volts from about 80% SoC to around 30% under modest load. There is no way for me to know what the SoC is from voltage, except when full or below 30%.

That's surprising to me. I would expect you'd have four digits of precision available somehow, but even with just 0.1V precision I would expect to see:

80%: 26.4V at the terminals at 0.05C
70%: 26.3V at the terminals at 0.05C
.
.
.
30%: 25.8V at the terminals at 0.05C

I mean, it might not be exactly that, and there could be some offset for fuses or other stuff in the circuit, but I would still certainly expect that you should see a measurable difference between 80 and 30.

[nebster]

Originally Posted by Delfin

The discharge figures make some kind of sense. Battleborn can't differentiate the difference between 80% and 40% SoC based on voltage, while Lithionics says between 80% and 30%.

I'm not sure if the Battleborn cells internally are actually larger and being sandbagged for safety/longevity, but the chart above definitely suggests a straightforward voltage difference between 80 and 40*SOC at a fixed load.

Given the variability of discharge loads in the real world, I suspect the "Terra incognito" SoC based on voltage is a bit wider than this. But perhaps sitting at the dock the loading is stable and you could set float to maintain a SoC give or take 40% capacity, more or less.

That's true*for some values: with a lower-precision setting/meter it would be very hard to hold a low SOC if the rate is going to vary a lot (either because you have huge intermittent loads/chargers or because you have in proportion a smaller battery). It gets much easier to use float to hold an SOC in the 80+% range OR if you have a high ratio of bank capacity to load size.

[CatNewBee]

Originally Posted by Dockhead

Fantastic information!


I have always known that voltage in a lead bank is an excellent measure of SOC. I tested this extensively to convince myself to throw away my amp-counting battery meters, some years ago.


This is valid even when reasonably small loads are attached, but not indeed when there is any charging source present. That's why Smart Gauges work brilliantly when you're using battery power, but don't tell you anything useful when you have any kind of charging source connected.


But how about during charging? I think that's what we care about here, isn't it?


But we could set a combination of voltage and charge current for switching to float, couldn't we? Wouldn't that be the most accurate way?

As with LiFeYPO4.

I use solar to charge my house bank, cut off at 14.2V, what translates to 89..90% SOC, my Victron BMV always goes then to 100% and the Amp counter differs by +/- 15Ah at a 1000Ah nominal bank, with 1200Ah between empty 2.8V and full at 3.65V the BMS amp counter is around -130Ah when the charge stopps and goes to float, no additional amps and no overcharging whatsoever.

This is from facts, not copy and paste fiction.
Charge current between 100A and 5A depending on radiation and loads usage. (0.1C charge, discharge up to 0.5C or 500A)

[Delfin]

Originally Posted by nebster

That's surprising to me. I would expect you'd have four digits of precision available somehow, but even with just 0.1V precision I would expect to see:

80%: 26.4V at the terminals at 0.05C
70%: 26.3V at the terminals at 0.05C
.
.
.
30%: 25.8V at the terminals at 0.05C

I mean, it might not be exactly that, and there could be some offset for fuses or other stuff in the circuit, but I would still certainly expect that you should see a measurable difference between 80 and 30.

My voltage curve does seem flatter than that. I have done a half arsed capacity test, drawing down 580 Ah of a nominal 600 and barely broke 25 volts. After 450, I'll still see 26.0 volts. I have a Blue Sea voltmeter with .xx resolution installed on the LFP bank, but it seems to bounce around enough in the 1/100 range that I didn't see how voltage could be easily used to determine SoC as compared to coulomb counting, but you clearly have more accurate tools than I do.

I don't need to float, since I can divert charge sources to the LA starter bank easily, but it looks like if I did want to float, a voltage of 26.1 might be feasible. That said, I guess I am still not convinced that holding the pack at some nominal voltage with constant small loads won't have undesirable consequences.

[CatNewBee]

Originally Posted by Delfin

My voltage curve does seem flatter than that. I have done a half arsed capacity test, drawing down 580 Ah of a nominal 600 and barely broke 25 volts. After 450, I'll still see 26.0 volts. I have a Blue Sea voltmeter with .xx resolution installed on the LFP bank, but it seems to bounce around enough in the 1/100 range that I didn't see how voltage could be easily used to determine SoC as compared to coulomb counting, but you clearly have more accurate tools than I do.

I don't need to float, since I can divert charge sources to the LA starter bank easily, but it looks like if I did want to float, a voltage of 26.1 might be feasible. That said, I guess I am still not convinced that holding the pack at some nominal voltage with constant small loads won't have undesirable consequences.

Try it for a couple of cycles with having a coulomb counter connected. You can see then how much you discharge and charge. As I said, I use 14.2 V as setpoint with 5...10min absorption (minimal setting), it would translate to 28.4V. Battery is full to specks at 3.65V per cell, 14.2V translates to 3.55V. This is far enough from the shoulders, but already after the low linear part, so quite accurately measurable and a good set point for the Victron MPPT. I have repeatable results in charging within a range of 1... 2% variation.

If you have a different chemistry you may find a slightly different setting more convenient. It is try and error until you find your spot that let's you sleep better.

[CatNewBee]

Here the settings, in the victron mppt they are 14.3V and 13.5V, the BMS and my voltmeter agree on 14.2V at the cell poles at this setting.

[john61ct]

The idea of "surface charge" afaic is only at full SoC, above 3.37V after resting isolated 24 hours

At the definitions of Full I use for normal-cycling, just drawing .005% of capacity brings OCV down to 3.35 or so, depending on the bank.

That would be my 100% starting point for a "correspondence table" between V and SoC

But these finer points of behaviour do vary greatly from one maker's batteries to another.

Even by size and form factor within one brand.

The cylindrical A123 compared to their pouch cells for example, although presumably the same purity of chemical if manufactured around the same time period.

This approach does need to be calibrated by the owner for **that** specific bank, and requires very accurate gear and a fair bit of effort.

Personally, as long as the Full point and too-low danger zones are well defined, a decent SoC meter will be a more reasonable universal recommendation for all but professionals and the OCD hobbyists among us (not meant to be a negative 8-)

[john61ct]

By aging I do not mean in calendar terms but functionally, by performance and capacity dropping.

With quality LFP cells well-coddled that does take many years to even start.

It stands to reason, when the 100% SoC definition has shifted down by 10% or more, precise measurements will reveal slightly different chemistry behaviour, including the V vs SoC curve shifting.