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

[goboatingnow]

Originally Posted by roetter

No constants in my code, but a few defines. A bit rusty on the C++ stuff. Have a class for Battery and for Alarm so far. The code is running well and less than two hundred lines including comments and empties.

#define HIGHALARMPORT 3
#define LOWALARMPORT 4
#define HIGHALARMON HIGH
#define HIGHALARMOFF LOW
#define LOWALARMON LOW
#define LOWALARMOFF HIGH
#define HVC 3.6
#define LVC 2.8
#define HIGHHYST 0.2
#define LOWHYST 0.2
#define TRUE 1
#define FALSE 0

Purely for educational purpose I am playing with some other means of reading the cell voltages.
I am playing with differential op-amps using dangerous Internet half-knowledge. I use an LM348 as differential amplifier so see if I get more precise results in creating voltages to measure, to compare to just the resistor bridges. However, the first cell reads off. So instead of getting 3.22 for each cell, I get 3.72 for the first cell. All other cells create 3.22 as a difference amplification value.

Strangely though, reading from ground to the second cell gives 6.44 which is correct. Maybe this is because this is not a rail-to-rail op-amp?

Anybody?

I am waiting on some fancier items that I saw today in a circuit on Ruggedcicuits.com. They have a circuit to create individual cell voltage to ground using op-amps and reference voltages. Circuit #5 I ordered the parts to have some more fun.

Yup that's the circuit , see it sets the dynamic range so your only reading from 2.9v. Just note the supply voltage warning , you'll need a greater common mode range for more cells

Yiu could use Lm732x series or OpA series , both have 32v common mode , rail to rail , the LM is a bit overkill with its 100ma drive capability !!! , but it makes it tough.

Dave

[goboatingnow]

Originally Posted by roetter

No constants in my code, but a few defines. A bit rusty on the C++ stuff. Have a class for Battery and for Alarm so far. The code is running well and less than two hundred lines including comments and empties.

#define HIGHALARMPORT 3
#define LOWALARMPORT 4
#define HIGHALARMON HIGH
#define HIGHALARMOFF LOW
#define LOWALARMON LOW
#define LOWALARMOFF HIGH
#define HVC 3.6
#define LVC 2.8
#define HIGHHYST 0.2
#define LOWHYST 0.2
#define TRUE 1
#define FALSE 0

Purely for educational purpose I am playing with some other means of reading the cell voltages.
I am playing with differential op-amps using dangerous Internet half-knowledge. I use an LM348 as differential amplifier so see if I get more precise results in creating voltages to measure, to compare to just the resistor bridges. However, the first cell reads off. So instead of getting 3.22 for each cell, I get 3.72 for the first cell. All other cells create 3.22 as a difference amplification value.

Strangely though, reading from ground to the second cell gives 6.44 which is correct. Maybe this is because this is not a rail-to-rail op-amp?

Anybody?

I am waiting on some fancier items that I saw today in a circuit on Ruggedcicuits.com. They have a circuit to create individual cell voltage to ground using op-amps and reference voltages. Circuit #5 I ordered the parts to have some more fun.

I don't think the LM 348 is a rail to rail amp , so the top cell maybe exceeding the OP amps common mode range , not easy to tell if there no circuit.

Dave

[roetter]

Here is a link to a video of camper guy who uses an Arduino to monitor all kinds of things in his camper. Makes a nice integrated data center.

Camper control panel using Arduino - YouTube

[roetter]

LiFePo 150Ah - used from Balqon 1st test report:

I had discharged the batteries quite a bit during all my testing of the Arduino equipment over the last 10 days or so.

Today discharged until LVC. Actually, several LVC points as can be seen from the confusing graph. I was using a 1500W inverter with a fan heater set to stage 2 which draws about 50A. The "up-bounces" are from running into LVC set points either on the Arduino or the JLD5740. I then got more daring and reprogrammed the LVC's even lower.
Attachment 63264

Then I did the first charge using a 50A bench power supply set to constant current of 50A. I promptly messed up starting the Junsi and did not record the charge cycle. The charge cycle lasted 3.66 hours with 50.1 A set at the power supply and 48.9 reading at the Victron BMV. As per my calculation I put in 183Ah (3.66h*50A) or 179Ah. A little low for 250Ah Winston cells. I am missing 70 Ah.

I then did a 10 min discharge at 50A, a five minute resting period, followed by a ~10min charge at 50A until HVC. I modified a regular extension cord by putting a relay on the 0-phase and connecting it to my HVC relays. Works great. Shuts down the bench power supply beautifully.
Attachment 63265

Now, maybe you LiFePo experienced guys can give an analysis of the various cells.

- The red cell appears to be lagging the other cells, and should be balanced upwards.
- The yellow cell appears to have more internal resistance than the other 3 cells.
Anything else?

I am wondering?
To be able to use full capacity under high loads and charge amps, should the cutoffs be modified by the amperage going in or out.

For example low cut off 2.8V - C*Amps.
or high cut off 3.6V + C*Amps.

C could be 0.02V/A, so 100A would change the cutoff by 0.2V.

[ebaugh]

Originally Posted by roetter

LiFePo 150Ah - used from Balqon 1st test report:

I had discharged the batteries quite a bit during all my testing of the Arduino equipment over the last 10 days or so.

Today discharged until LVC. Actually, several LVC points as can be seen from the confusing graph. I was using a 1500W inverter with a fan heater set to stage 2 which draws about 50A. The "up-bounces" are from running into LVC set points either on the Arduino or the JLD5740. I then got more daring and reprogrammed the LVC's even lower.

Then I did the first charge using a 50A bench power supply set to constant current of 50A. I promptly messed up starting the Junsi and did not record the charge cycle. The charge cycle lasted 3.66 hours with 50.1 A set at the power supply and 48.9 reading at the Victron BMV. As per my calculation I put in 183Ah (3.66h*50A) or 179Ah. A little low for 250Ah Winston cells. I am missing 70 Ah.

I then did a 10 min discharge at 50A, a five minute resting period, followed by a ~10min charge at 50A until HVC. I modified a regular extension cord by putting a relay on the 0-phase and connecting it to my HVC relays. Works great. Shuts down the bench power supply beautifully.

Now, maybe you LiFePo experienced guys can give an analysis of the various cells.

- The red cell appears to be lagging the other cells, and should be balanced upwards.
- The yellow cell appears to have more internal resistance than the other 3 cells.
Anything else?

I am wondering?
To be able to use full capacity under high loads and charge amps, should the cutoffs be modified by the amperage going in or out.

For example low cut off 2.8V - C*Amps.
or high cut off 3.6V + C*Amps.

C could be 0.02V/A, so 100A would change the cutoff by 0.2V.

The bank is just a little out of balance. The red cell is low and the yellow cell high. For observed testing, I do occasionally charge to 3.6V max, mainly to balance. But I would cutoff normal charge at 3.45V, maybe 3.5V based on the highest cell.

I don't understand the difference between observed capacity and rated capacity. If the cells are new, you should see slightly better than the cells rating. I would try to get to the bottom of this.

Generally speaking for house bank apps, there would be no reason to modify LVC based on current. Perhaps if your peak demand was greater than .5C, say 150A for the inverter on a 200Ah bank. But then I would suggest a larger house bank.

HVC is different, especially if you have solar or wind powered trickle charge sources AND want to charge to 100%. Not so much for high rates, but very low rates from say Solar or Wind sources. At say a 10Ah charge rate you would want to terminate at a much lower voltage than if charging at 100A to reach 100% capacity. This is why generally I suggest 3.45V to terminate charge, it will mostly keep you out of trouble. To make it an exact science, it is a function of both bank capacity and charge rate. The problem is I don't think anyone has the "formula" to apply to determine the values.

[SVI]

This one is nice: A quick demo of the Cell Log 8 Breakout Module « Electric Porsche 944

[T1 Terry]

Originally Posted by SVI

This one is nice: A quick demo of the Cell Log 8 Breakout Module « Electric Porsche 944

For a 98v battery, not a lot of use for house batteries though. The builder is RWAudio from the DIY EV forum, I asked him a while back if he was interested in making single logger boards but he said it wasn't worth it to him as his interest was in electric vehicle battery monitoring.

T1 Terry

[SVI]

Connect the 98v battery in different way - and you will have the same 12.8v house battery. The board can switch off individual (or group of cells) based on individual low/high voltage... in charge or discharge.

[T1 Terry]

Originally Posted by SVI

Connect the 98v battery in different way - and you will have the same 12.8v house battery. The board can switch off individual (or group of cells) based on individual low/high voltage... in charge or discharge.

i think you would be just making an extremely complex system out of what could be a very simple system to monitor, to me it looks to be a move in the wrong direction. The idea is to make the house battery monitoring as simple as possible, by breaking a large capacity 4 in series battery into multiple small capacity 4 in series batteries connected in parallel would require monitoring and controlling the voltage levels in each small 12v battery, why would you do that when you can just monitor 4 groups of parallel cells in series?

T1 Terry

[SVI]

Originally Posted by roetter

Interesting idea, but:

I am not sure how you would like to see that as a PC app? Maybe you can explain.

Some thoughts from my side:
- A PC has much higher power demands than the Arduino. The board uses about 25mA at 5V. This processor is used in all kinds of low power devices, including the Junsi CellLog. So you would not like to have the PC running all the time.
- I am planning to put a SD card on the board to record the voltages with time stamps. The card can be read by any PC and analyzed by Excel.

- The settings for low and high voltages, etc. could be set using a simple interface on the LCD display.

What are you planning to do on the PC?

Actually, I like your approach, it looks like Arduino is very applicable for this job.

Couple of quesions:
- how many analog inputs latest Arduino can handle? How many outputs?
- you can add the temperature sensors (they are cheap) to every cell - usefull to find problems when charging.

I plan for 480ah 12.8v battery setup - 32 x 60ah cells in 4S12P configuration.
Can I monitor the voltage/temperature of every of the 32ah cell?
I want to put 12 relays to turn off any 4S string if any of the cell in the string is below LVC or above HVC.

I am on the right track?

PS I may skip the Arduino and go with 4 Cellogs and 4 4S2P batteries... But it looks like the Arduino way is more fun

[SVI]

Originally Posted by SVI

Actually, I like your approach, it looks like Arduino is very applicable for this job.

Couple of quesions:
- how many analog inputs latest Arduino can handle? How many outputs?
- you can add the temperature sensors (they are cheap) to every cell - usefull to find problems when charging.

I plan for 480ah 12.8v battery setup - 32 x 60ah cells in 4S12P configuration.
Can I monitor the voltage/temperature of every of the 32ah cell?
I want to put 12 relays to turn off any 4S string if any of the cell in the string is below LVC or above HVC.

I am on the right track?

PS I may skip the Arduino and go with 4 Cellogs and 4 4S2P batteries... But it looks like the Arduino way is more fun

PPS it looks like I need several Analog-Digital Converters to connect all analog voltage sensors to Arduino (max is 16 inputs). That's a "veeery dark forest" for me...

[ebaugh]

Originally Posted by SVI

Actually, I like your approach, it looks like Arduino is very applicable for this job.

Couple of quesions:
- how many analog inputs latest Arduino can handle? How many outputs?
- you can add the temperature sensors (they are cheap) to every cell - usefull to find problems when charging.

I plan for 480ah 12.8v battery setup - 32 x 60ah cells in 4S12P configuration.
Can I monitor the voltage/temperature of every of the 32ah cell?
I want to put 12 relays to turn off any 4S string if any of the cell in the string is below LVC or above HVC.

I am on the right track?

PS I may skip the Arduino and go with 4 Cellogs and 4 4S2P batteries... But it looks like the Arduino way is more fun

The conventional wisdom is to connect cells first in parallel then in series. Then only each paralleled cell needs to be monitored for voltage. You can spread a few temp sensors around, but you probably don't need one for every individual cell. This lowers the number of analog inputs to 4 for voltage monitoring, regardless of the cell count on a single 12V system. Then 1 more for every temperature sensor.

There are no advantages to wiring the other way for a single battery bank serving the same loads. You might make a case for 2 banks, but the reasons to do that or not are the same for LFP as FLA. Again, I think the conventional wisdom is against splitting house banks into 2 units.

[SVI]

Originally Posted by ebaugh

The conventional wisdom is to connect cells first in parallel then in series.

I do not know what is "a conventional wisdom"

My logic is - 4S then 8P - if I monitor every cell, I will be able to recognize and replace any faulty cell before it ruins the complete string.

If I make 8P and then 4s - I will not be able to monitor every cell (as 8P will equalize to the same voltage), only every string. Any one faulty cell will be drawing amps from other health 7 in the same string, and ruin them over time.

It is possible to extend the number of analog inputs on Arduino by additing 2 16ch external ADC to 2 digital Arduino inputs, or by multiplexing the analog inputs.

Several temperature sensors is more for the piece of mind, some sort of "fire alarm".
Also, it is possible to install 30A current sensor and add Amp/h counting function.

[ebaugh]

Originally Posted by SVI

I do not know what is "a conventional wisdom"

My logic is - 4S then 8P - if I monitor every cell, I will be able to recognize and replace any faulty cell before it ruins the complete string.

If I make 8P and then 4s - I will not be able to monitor every cell (as 8P will equalize to the same voltage), only every string. Any one faulty cell will be drawing amps from other health 7 in the same string, and ruin them over time.

It is possible to extend the number of analog inputs on Arduino by additing 2 16ch external ADC to 2 digital Arduino inputs, or by multiplexing the analog inputs.

You can detect a faulty cell in the arrangement I suggested, by noting one of the large cells getting significantly out of balance with their sister cells. Programmatically monitor the max difference, which is the difference between the high and low cell. Except at top and bottom of charge, that will be within .02V. Anything else on a well balanced pack indicates a problem.

If you want to, an Arduino can be extended to have an almost unlimited number of analog ports via the serial port. Probably the cheapest way is more Arduino's networked together. Multiplexing the ports is probably possible, but outside my experience. Lots of parts and bread boarding.

[T1 Terry]

It looks like you are trying to develop a problem so you can build a solution :lol: If you put 8 x 4 cells in series into a parallel connection involving load and charging at a common point you will build a problem that will keep you so busy fixing you won't have time to be sailing, I thought the basic plan was for the system to work the other way around, the battery to look after itself so more time could be spent sailing.

T1 Terry