LiFePO4 Batteries: Discussion Thread for Those Using Them as House Banks

[diugo]

Originally Posted by T1 Terry

What brand and how did you determine they were at 20%SOC? I would be worried if I received a cell below 3.2v still fresh in the box.

They are TS-LFP260AHAs, initial voltage was 3.292±3mV. I discharged them using a 5-amp load for about ten hours, down to around 3.08V, which was as low as I wanted to go before paralleling and bottom balancing them and giving them their first charge.

[T1 Terry]

Originally Posted by diugo

They are TS-LFP260AHAs, initial voltage was 3.292±3mV. I discharged them using a 5-amp load for about ten hours, down to around 3.08V, which was as low as I wanted to go before paralleling and bottom balancing them and giving them their first charge.

did they give you manufacture certificate with the cells? It will have the cell part number, manufacture date and voltage it left the factory along with the QC persons name. 3.292v from an unused cell is what you should see, did the cells take their full capacity charge after you dropped them down to 3.08v? You do realise that 3.08v is not fully discharged, 2.8v is fully discharged, you still had a lot of capacity to go.

T1 Terry

[Lagoon4us]

OK to use a cell logger to signal say a 200amp breaker to shut the discharge off in the event of 'over or under' voltage on one or more cells, what actual components do you need in a circuit to activate a heavy breaker like that from a small signal?

Diagram?

Cheers

[diugo]

Originally Posted by T1 Terry

You do realise that 3.08v is not fully discharged, 2.8v is fully discharged, you still had a lot of capacity to go.

You're wrong about that. Even at a mere 5A (0.02C), voltage was dropping fast enough that I felt it unnecessary to take the cells any lower before their first charge. If you want to quibble and say I left ~2% still on the table, I'll make that concession.

There was no paperwork whatsoever supplied by Balqon. The cell cases, of course, bear a date code and serial number.

[hellosailor]

Just curious about how folks are measuring voltages, since I'm seeing "specs" calling for 3.45v or 3.95v or similar values and diugo is measuring voltages to the thousandth of a volt??

Let's say you are aiming for 3.50 volts, and you are using a meter that is accurate to 1% on DC volts. It could indicate 3.535, which would show up as 03.56 on a conventional 20VDC scale. Or 3.465, which might show as 03.47. And if the meter had a float of 2-3 digits on the rightmost digit, as most do, that means it could indicate anywhere from 3.56 to 3.44 when the real voltage was 3.50.

Of course if the meter was accurate to 1/2% like a good Fluke...that could still be off by 2/100's of a volt (3.48-3.52 when you're looking for 3.50) and STILL subject to float on that rightmost digit.

So I'm just curious, with all this talk of measuring hundredths of a volt, how is that being measured? And what is the rated precision and accuracy of the fancy monitor/control systems that are displaying voltage, too?

Maine-
You need to spend more time in the supermarket. Really, haven't you ever bought a half-empty box or bag of something, and been told "May have settled in shipping" ? That's obviously a Lay's battery, the plates settled just like the chips do.

[diugo]

Originally Posted by Lagoon4us

OK to use a cell logger to signal say a 200amp breaker to shut the discharge off in the event of 'over or under' voltage on one or more cells, what actual components do you need in a circuit to activate a heavy breaker like that from a small signal? Diagram?

I'm looking into that exact question myself, as we speak. My shunt trip appears to draw about 0.6A around 13V. I am in the process of a full discharge, so I'll test it again at 12V. I'm hoping to use a 6A, 60-millohm high-side switch as the intermediary. If it works I'll share the details and a diagram.

[diugo]

Originally Posted by hellosailor

Just curious about how folks are measuring voltages, since I'm seeing "specs" calling for 3.45v or 3.95v or similar values and diugo is measuring voltages to the thousandth of a volt??

Great question! My Fluke 87 displays millivolts in its 4VDC range. Accuracy is spec'd at ±0.1% ±1 count, so a displayed 3.088V is actually 3.088V ± 4mV.

For actual live monitoring, I am using an Arduino-like microcontroller with a 12-bit ADC, scaling up to a max of 17 volts. 12 bits is 2^12 or 4096, so it is also accurate to ± 17/4096 = 4mV. Thus I only display cell voltages to the hundredth.

[T1 Terry]

If you are trying to determine an exact voltage to compare it to a state of charge, the thousands of a volt matter, but the conditions to replicate the procedure are probably even more likely impossible to do. These cells aren't that fragile, plus or minus 5/100th of a volt at cell level is about as precise as you need to go and that's only for determining fully charged at rest or fully drained under a 0.5C load, 1/10th of a volt at the very top and bottom happens quickly, but in the middle it means very little. I wouldn't be attempting to do the bottom balancing personally, that's for those guys using them in EV's they need to know they will all go flat at the same time if they plan to use every little bit out of the battery. Far better off top balancing and draining the 20% SOC, you will never hit the bottom then and your cells will last a long time. 3.6v is the about as far as you want to charge the cells, the time between 3.6v and 3.65v at higher charge rates is nearly the blink of an eye stuff, a charge controller with a 0.5v hysteresis is exact enough, just allow for it in the max charging voltage, 0.5v is only 0.125v per cell across a balance pack, a 13.8v max that floats between 13.85v and 13.75v isn’t going to be an issue. I even use slow switching with my plasmatronics controller with 0.2v hysteresis just to give the solid state relays a break, they get a bit warm rapid switching 100 plus amps :lol:

If you need a low voltage cell cut to happen automatically you must be planning to leave the batteries unattended under load for quite some time. The question is, what happens to what ever the load was when the power is turned off? If it was the fridges, you are looking at a lot of fish burly and a fridge restock, what else would be important enough to leave running for a long period while you weren't there? You know how much the fridge uses per 24hrs, just make sure that’s left in reserve before you go walk-about if you don’t have solar charging.

If you want to cut the load i the case of a deeply discharged cell, the main relay would need to be wired so it ratcheted and drive it with a small solid state relay or transistor circuit from the cell logger alarm terminals. If you have the equivalent of a Jay Car store you can buy one of these kits, KC5434, it's an opto coupler driving a transistor driving a 30 amp mechanical relay, but any solid state relay can be driven directly by the alarm port.

T1 Terry

[daddle]

Originally Posted by diugo

Accuracy is spec'd at ±0.1% ±1 count, so a displayed 3.088V is actually 3.088V ± 4mV.

Plus the one count, 0.001, gives ±5mV, no? Usually the percentage is expressed as a percent "of full scale". Ya, picky.

You need good consistent technique to measure at these accuracies in electronics. And proper honest datasheets for it to make any sense. And product which within specification. Note the voltage your fingers apply to the probes...

I would be happy with one part in a hundred (1%, 40mV) measuring on a boat. I would accordingly conservatively base my limits with some room for error. Do the battery datasheets a give temperature compensation figures? Can the internal temperature of a cell be estimated?

The cell monitoring circuits will need to be designed and maintained carefully to reliably sense millivolt levels.

[Lagoon4us]

Originally Posted by diugo

I'm looking into that exact question myself, as we speak. My shunt trip appears to draw about 0.6A around 13V. I am in the process of a full discharge, so I'll test it again at 12V. I'm hoping to use a 6A, 60-millohm high-side switch as the intermediary. If it works I'll share the details and a diagram.

That's the go just make sure you put it in dillspeak please, i'm happy to minimise a BMS down to disconnect function controlled only by cell loggers and a timer.

We are keeping the existing LA circuit separate but powering anchor winch, thruster and engine start. Lithium for house and sailing only.

Makes sense to us till confidence builds...

[Lagoon4us]

Just need this translated into dillspeak, sorry Terry



If you want to cut the load i the case of a deeply discharged cell, the main relay would need to be wired so it ratcheted and drive it with a small solid state relay or transistor circuit from the cell logger alarm terminals. If you have the equivalent of a Jay Car store you can buy one of these kits, KC5434, it's an opto coupler driving a transistor driving a 30 amp mechanical relay, but any solid state relay can be driven directly by the alarm port.

[Lagoon4us]

If you need a low voltage cell cut to happen automatically you must be planning to leave the batteries unattended under load for quite some time. The question is, what happens to what ever the load was when the power is turned off? If it was the fridges, you are looking at a lot of fish burly and a fridge restock, what else would be important enough to leave running for a long period while you weren't there? You know how much the fridge uses per 24hrs, just make sure that’s left in reserve before you go walk-about if you don’t have solar charging.

I guess my thoughts are to protect the batteries from not being monitored correctly OR a failure in the solar charge whilst away, appreciate the load comment in that there's ample amps there in reserve but if doing hi voltage why not do low.

Cheers

[diugo]

Originally Posted by daddle

Plus the one count, 0.001, gives ±5mV, no? Usually the percentage is expressed as a percent "of full scale". Ya, picky.

I don't see "full scale" in the specs. Maybe it's a carryover from when voltmeters actually had needles?

I already took the 1 count into account when I posted the 4mV accuracy.

[deckofficer]

Heck Doug, I remember rectifier tubes before diodes. Needed that DC B+ voltage for our triode tube before we had transistors. Quite a transition back then, going from voltage gain devices to current gain devices.

[daddle]

Originally Posted by diugo

I don't see "full scale" in the specs. Maybe it's a carryover from when voltmeters actually had needles?

I already took the 1 count into account when I posted the 4mV accuracy.

It must be relative to full scale. What else could it be? The numerical accuracy can hardly go up as the voltage reading drops. A reasonable way to proceed would be to measure voltage differences, but the cells need to be separated for that.

Just a warning that it is easy to delude yourself that your measurements are accurate when measuring millivolts in 4 volts. The meter's digits look so confident.

The LFP cell characteristics see similar to those of semiconductors where exceeding a maximum rating leads to immediate damage to the part. Good designers stay well away from those values, often sacrificing considerable performance. Same goes for noise and other inaccuracies, they build up remarkably fast.