diy solar

diy solar

LiFePO4 Capacity - Voltage chart when under load/charge/nil

nate_syd

New Member
Joined
Apr 14, 2021
Messages
161
I understand the downside of estimating capacity by voltage etc...

There's a few voltage charts around however you get different results depending if the battery is being charged/discharged.
If the battery is resting at 50% & you apply a load - the voltage instantly drops, however the capacity hasnt really changed.
If the battery is resting at 50% & you start to charge - the voltage instantly rises, however the capacity hasnt really changed.

So I wanted to create a chat that can be referenced in each condition & get a "more accurate" picture given the varying conditions.

Using Andy's work here: https://off-grid-garage.com/battery-data/
I've compiled some voltage-capacity charts under charge & discharge!
 

Attachments

  • Voltage-Capacity.PNG
    Voltage-Capacity.PNG
    20.7 KB · Views: 210
  • Capacity v Voltage.zip
    513.2 KB · Views: 82
The thing to be aware of is that this chart will be accurate only for those specific cells and setup. As cell IR varies with different cells .... or even the age of those cells .... or someone else has different resistance thru their bus bars, wires, and connections ... the voltage profile will be different.
 
This is about the 5th time I've seen someone try this with lifepo4 on this forum and they're all almost worthless for anyone who doesn't have your specific battery pack and your specific meter with whatever voltage inaccuracies it has.

The issue is many meters are so inaccurate at the mv scale you might be off by 20%.

Not to mention the variation in cell IR having an effect on the voltage you'd get at any given soc.

I'll pull up my data and make my own real fast and you can see the variation. It'll be for a 4s pack but I'll just divide by 4 and call it close enough to illustrate the point.
 
Turns out real fast isn't so fast. My computer is currently trying to start on fire with a hundred thousand data points.
 
Here you can see its somewhat close to yours, but you'll notice my 40% is your 50%. Ish.

And meter error will drive this variation up a lot, so it might be plus or minus 10%

10% error in my pack gives me a range of 48ah, which is a full 2 days of roof fan usage lol

241 AH Table.png
 
This is about the 5th time I've seen someone try this with lifepo4 on this forum and they're all almost worthless for anyone who doesn't have your specific battery pack and your specific meter with whatever voltage inaccuracies it has.

The issue is many meters are so inaccurate at the mv scale you might be off by 20%.

Not to mention the variation in cell IR having an effect on the voltage you'd get at any given soc.

I'll pull up my data and make my own real fast and you can see the variation. It'll be for a 4s pack but I'll just divide by 4 and call it close enough to illustrate the point.
In general, I agree with you, especially when it comes to meter accuracy (many of which, like BMS, are +/-5mV, meaning as much as ~4% SOC variation between measurements in the ‘flats’ (down to ~0.3% variation below ~14.5% SOC and ~0.2% variation above ~99% SOC).

On the other hand, variation due to IR and even cabling and contact resistance can be eliminated if waiting for cells to settle and reading at equilibrium.

It would be enormously useful if someone with suitably-sensitive equipment could monitor battery / cell SOC, and for each readpoint, allow the battery to reach equilibrium and then measure individual cell voltages using the same high-precision multimeter.

This would provide a LiFePO4 cell voltage versus SOC table that would at least represent a meaningful baseline that members could attempt to compare against.

Any tables provided from measurements while any currents were flowing is essentially meaningless (for the reasons you mention).
 
The thing to be aware of is that this chart will be accurate only for those specific cells and setup. As cell IR varies with different cells .... or even the age of those cells .... or someone else has different resistance thru their bus bars, wires, and connections ... the voltage profile will be different.
yeah 100%.

This chart has been floating about & is quite popular - https://diysolarforum.com/threads/lifepo4-24v-voltage-chart.22388/
I wanted to get something that was closer to realistic conditions. Its not an exact answer - but an answer based on a lot more data & sound research work.
 
yeah 100%.

This chart has been floating about & is quite popular - https://diysolarforum.com/threads/lifepo4-24v-voltage-chart.22388/
I wanted to get something that was closer to realistic conditions. Its not an exact answer - but an answer based on a lot more data & sound research work.
Found this: https://mdpi-res.com/d_attachment/wevj/wevj-06-00752/article_deploy/wevj-06-00752.pdf

It’s unfortunately not in table form, but figure 2: ‘OCV vs SOC curve’ is exactly what we have been discussing.

I’ve bottom-balanced my battery so that after an overnight rest, all of my cells are within 2-5mV of 3.040V, meaning that according to table 2, I’ve discharged my battery down to about ~4% SOC (compared to the tables in this thread that indicate closer to ~11%).
 
In general, I agree with you, especially when it comes to meter accuracy (many of which, like BMS, are +/-5mV, meaning as much as ~4% SOC variation between measurements in the ‘flats’ (down to ~0.3% variation below ~14.5% SOC and ~0.2% variation above ~99% SOC).

On the other hand, variation due to IR and even cabling and contact resistance can be eliminated if waiting for cells to settle and reading at equilibrium.

It would be enormously useful if someone with suitably-sensitive equipment could monitor battery / cell SOC, and for each readpoint, allow the battery to reach equilibrium and then measure individual cell voltages using the same high-precision multimeter.

This would provide a LiFePO4 cell voltage versus SOC table that would at least represent a meaningful baseline that members could attempt to compare against.

Any tables provided from measurements while any currents were flowing is essentially meaningless (for the reasons you mention).
Definitely not enormously useful and anyone with a high precision meter would just use a shunt.
 
Working on the data & making some progress... Getting some good results & predictability within the flat part of the curve.
This is really just saving me from buying a wifi shunt & having to pull data out of it.

So far the logic is...
Use static values start/end curve:
3.45v = 95%
3.4v = 90%
3.1v = 10%
3v = 5%

Then adjust cell voltage depending on charge/discharge current:
cell voltage = cell voltage + (battery power / 40000)

capacity = ((Cell_voltage - 2)^11)*3)
which is: cell voltage - 2, then x itself 11 times, then x3
That will get us 20-80% pretty smoothly

Logging lots of data now, so will continue to refine
 
And are you doing this with your cells only or do you have a significant number of different cells?
 
And are you doing this with your cells only or do you have a significant number of different cells?
With my battery pack...
16S - 8xCATL 310's & 8x280 EVE's

Take the full voltage & /16 to get the cell voltages that I use for the calculations.
 
Oof. Yeah that's really not going to be useful information.

The fact that you're mixing cells like that makes it even worse. You're just averaging them out. When I did it, it was just to illustrate the point that the results wind up different from pack to pack but your different cell types will make the discrepancy much larger.

Your numbers are even drifting farther from mine as well, further showing the point that voltage and SOC are at best vaguely correlated and that this method shouldn't be used.

It'll work for your pack only for some time but eventually even that will be off with age.
 
Back
Top