A123 capacity vs charge voltage (and other effects)

texaspyro

100 kW
Joined
May 12, 2010
Messages
1,408
I recently added a battery analyzer mode to my capacitive discharge FET based tab welder (http://endless-sphere.com/forums/viewtopic.php?f=2&t=2633&start=690#p300329).

I used it to analyze the capacity of a 2300 mAh A123 cell that had been charged to various voltages. The cell was charged with an HP E3614A cc/cv lab supply set to a 6A current limit. The cell was charged until the current fell to 50 mA. It was then removed from the charger and discharged (at around a 1C rate) to 2.5V. Extracted power was:

3.65V -> 2293 mAh
3.55V -> 2282 mAh
3.45V -> 2268 mAh
3.40V -> 2274 mAh
3.35V -> 2157 mAh
3.325V -> 1526 mAh
3.30V -> 953 mAh
3.25V -> 611 mAh

There appears to be no significant difference in capacity when charged above 3.40V. Somewhere between 3.30V and 3.35V cell capacity drops off a cliff. A123 recommends a float charge voltage of 3.45V

Charging to a voltage lower than the standard recommended 3.65V may improve battery life, particularly if you don't trust the accuracy of the cutoff voltage of your charger. A value of 3.50V looks like a good place to set an iffy charger... midway between the recommended charge voltage and where the capacity begins to fall off. Charging at lower voltages takes longer than charging at higher voltages. Those runs at 3.30V and below took forever for the charge current to taper off.

---------------------------

Here is some capacity data on the effects of different discharge rates:
Testing on 2 different new A123 cells hot off the charger (and charged outside where it was 100F) Cell 2 was a bit hotter off the charger. It was brought in from outside and tested without any chance to cool.


Basically A123 cells have no capacity loss from 1C to 5C discharge rate. There may even be a very slight boost at the higher rates due to increased cell temperature.

Cell 1:
@1C: 2248 mAh, 6862 mWh
@5C: 2237 mAh, 6848 mWh
@7C: 2242 mAH, 6704 mWh

Cell 2:
@1C: 2277 mAh, 7243 mWh
@5C: 2337 mAH, 7249 mWh

------------------

Cell temperature rise data:

At 1C discharge, +1 degree F in open air, +3.5F insulated.
At 5C discharge, +23 degrees F in open air, +33F insulated
At 7C discharge, +40 degrees in open air

-------------------

I did some testing to determine the A123 cell voltage that indicates 90% discharge. Unfortunately it seems to depend upon the cell discharge rate (or perhaps temperature).

At 1C a voltage of 3.075V was a good indicator there is 10% capacity left.
At 5C the voltage was 2.850V
At 7C the voltage was 2.822V

-------------------

Next I popped a cell in a 5F freezer for a couple of hours, removed it, and then immediately discharged it 5C (12 amps) in an insulated, room temperature box. First plot is a normal (non-chilled) cell. The next two are the voltage and wattage curves of the chilled cell... can you spot the cold? The total cell discharge capacity was not affected by the cold, but its ability to hold up the voltage and power output was drastically affected.

5C Volts.jpg

5F Voltage.jpg

5F Watts.jpg


Here is a plot of A123 cell voltage recovery after being fully discharged at around 1C. After discharge at 1C to 2.500V there was a 4 milliamp load on the cell.

cell_recovery.jpg


-----------------------------

I charged some A123 cells with a CC/CV charger (constant current/constant voltage). The CC phase was set at 6 amps. With the 2300 mAh cells, the cells would ideally fully charge in 23 minutes, but the chemistry does not allow that. At 18 minutes the charger started the CV phase where the pack voltage remains constant and the charge current drops. It took another 18 minutes to reach the recommended cutoff current of 50 mA.

So just how much extra capacity does one gain in the final CV charge phase. A fully CV/CC charged cell had 2305 mAh capacity. The same cell that was charged for only the 18 minute CC phase had 2093 mAh. You gain around 9% capacity from the final CV phase (that takes as long as the constant current phase).

-----------------------------

How are A123 cells for self-discharge? I bought some A123 cell developer kits from a guy. They came with the original sales invoices. He bought them in June of 2006 and never used them! They had never been charged in over 4.5 years! The cells gotta be toast, right? Wrong! The 12 cells measured from 3.189 to 3.306 volts. Most were above 3.3V

The factory ships them at 50% charged so they should have left the factory with around 1150 mAh in them. I put them on my analyzer and they still had from 209 to 989 mAh in them. 7 of the 12 cells had over 900 mAh left, the others 209, 517, 683, 789, 810 mAh. It looks like the self discharge rate of these virgin cells (stored at room temp) was around 5 percent PER YEAR!

Can you expect these kinds of self discharge in the real world? Probably not, but you should be able to get close to them. These were virgin cells. A good way to increase a cell's self discharge rate is to abuse them. Also, the self discharge rate of a paralleled string would be governed by the weakest cell in that string. My weakest cell had a 18%/year rate (note: rates calculated using simple math... i.e. capacity drop/years, not proper exponential decays)
 
charging roughly half depleted prismatic 15Ah A123 cells with a 150 watt PS, I monitored voltage at cell level .. with my charger set at 3.65v
the cell sprints up to 3.45V and just holds steady for a period of time.. then rapidly climbs to 3.65v..
also set charger to 4.05V in an attempt extend beyond the 3.45V barrier voltage.. didnt work.. voltage still holds steady 3.45V then will rapidly escalate upwards to the charger set voltage...
 
The 3.45V area is where your power supply is operating in constant current mode. It is supplying all the current that it can (or that you have allowed it to) output and has reduced its output voltage to maintain that current.

When the cell has absorbed all the juice it can at that voltage, the current flowing into it begins to fall and the power supply starts raising its output voltage to keep the current flowing at the maximum allowed rate. Once the power supply voltage reaches the max voltage you have it set to, the current into the cell begins to fall as it fills to its maximum capacity at that voltage.

You should stop charging A123 cells when the current drops below 50 mA. Or, you can "float" them indefinitely at 3.45V
 
texaspyro said:
Charging to a voltage lower than the standard recommended 3.65V ma improve battery life, particularly if you don't trust the accuracy of the cutoff voltage of your charger. A value of 3.50V looks like a good place to set an iffy charger... midway between the recommended charge voltage and where the capacity begins to fall off.


Yeah I also found there's little capacity when charging my LiFePO4 battery over about ~3.42V/cell. A lower charge voltage would work fine if you are charging single cells or a pack with perfectly balanced cells however when bulk charging a string of series cells a higher per cell voltage is used to try to make sure all of the cells get fully charged. Also when using a BMS you do want to use a charging voltage somewhere above the level which will activate the shunts, generally 3.6-3.7V/cell, again to top balance the pack.

-R
 
well thanks texaspyro ..

that power extracted table is of great information..

more precision is still needed ..
the values between the 3.30v - 3.35v
for the 2300mAh cells the capacity is only approx. 1A difference
however for larger capacity cells that could be alot larger..

Do you think there might be a formula or would
0.01v equate to 200mAh ?

3.35V -> 2157 mAh
3.30V -> 953 mAh
 
Thanks for sharing TexasPyro. I take this to mean I can just bulk charge to a 3.55v/cell average and only worry about balancing if cells are out greater than +/- .1V . Your data seems to clearly define the reason that the one guy (I forget his username) may have been correct about the whole balance changing thing, though he didn't have the data as to why. For me this seems pretty clear cut as to how to manage my A123's that I've had shelved for a year and just this week started my plan to put them in service. Your post is perfectly timed, and great news from my POV.

Is there any way I can beg you into testing what happens at the bottom too. Is there a similar clearly defined line of no capacity beyond? I'd be most interested in a 10% or more remaining capacity point if one exists that is fairly clearly defined?

John
 
John in CR said:
Is there any way I can beg you into testing what happens at the bottom too. Is there a similar clearly defined line of no capacity beyond? I'd be most interested in a 10% or more remaining capacity point if one exists that is fairly clearly defined?

Take a look at the graphs in the photos at the link in my first post. Once the cell gets below around 2.9V it is pretty much all over. The graph drops like a rock. There is VERY little time from 2.9V to 2V at a 4C discharge. Maybe a couple of percent of the cell capacity.

Using 2.9V as a LVC may be (just) a little extreme. You could hit that on typical current peaks, but you won't be doing it for long. At 4C it took 800 seconds for the current into a 0.25 ohm load to drop from 9.3A to 8.3A and only 20 more seconds to drop from 8.3A to 3.7A (where it hit the 2.5V cutoff that I had set)

I'll see if I can plot the cumulated mAh and figure out a reasonable 10% remaining capacity voltage. It looks like something around 2.95V will be it.

One issue when I was charging the cells at lower voltages (below 3.35V) where the cell had little total capacity, it took forever for the current to reach it's charge cutoff level.

And those capacity values in the first post were at around 1C average discharge current. At 4C I tend to see around a 10% reduction in capacity. A lot of that could be going out in I*I*R losses in my crappy 0.080 ohm clip lead connections to the cell.
 
Texas Pyro... the irony in the timing of your post just kills me :roll: . I was just about to post some findings I recently made along the same lines, but not quite so overarching as your own. 8)

I have a 20ah 8s2p Headway pack and a Chargery DB8 stand-alone balancer I have experimenting with. It is also worth noting that one of the cell pairs in this pack straggles some due to some damage incurred a while back (long story), so this pack is ideal for evaluating the effectiveness of various balancing techniques.

I have been trying to incorporate a 2.5 amp SLA charger that I can run in two different CC/CV modes (27.6v and 30.2v). Using the 30.2v setting and the DB8 the cell voltages settle at 3.76v per cell, but the DB8 with 350mv balancing shunt current can get quite warm while it is working hard to get the job done. Using the 27.6v mode, the balancing process (much to my surprise) was actually faster and resulted in 3.46v across the board with the DB8 emitting substantially less heat.... and no single cell exceeded 3.75v thoughout the charge process. At that point I switched the charger to 30.2v mode. The cells quickly shot up to 3.76v in unison and absorbed a mere 0.05ah in doing so. That amounts to only 0.25% additional capacity... downright negligible. I was in disbelief after that first round, so I repeated it four more times with identical results :mrgreen:

This proved to me that balancing them at 3.46 volts was not only possible, but actually preferable in my case because it made the DB8 balancer work much less hard. So using the 3.45v per cell (27.6v bulk charge) setting has proven to be the sweat spot for this particular combination of charger and balancer. 8)
 
John in CR said:
Is there any way I can beg you into testing what happens at the bottom too. Is there a similar clearly defined line of no capacity beyond? I'd be most interested in a 10% or more remaining capacity point if one exists that is fairly clearly defined?

Ok, I modified my code to be able to plot/dump accumulated mAh of discharge. At a 4C discharge rate the cell reached 90% of its capacity at a voltage of 2.85V (a little lower than what I expected). The cell had been charged to around 90% of full capacity when the test started (OK, I burned off 10% of a full charge tweaking the mods to my code). I'll try some tests with a full charge and at 1C.

What was rather unexpected was the shape of the mAh curve. A rather straight line. I was expecting some curveatude at the ends. The falloff may have been masked since it is so quick. I don't have a zoom for my LCD graphs (but the raw data can be dumped).
 
I did a couple more tests with fully charged cells. At 1C discharge, 90% capacity was reached at 3.070V. At 5C it was 3.068V. End of discharge was set at 2.500V

It looks like capacity does not change a whole bunch from 1C to 5C discharge.
 
texaspyro said:
I did a couple more tests with fully charged cells. At 1C discharge, 90% capacity was reached at 3.070V. At 4C it was 2.909V. End of discharge was set at 2.500V

It looks like at 4C you lose 10% or so capacity over what you could get at 1C. Mostly due to I*I*R losses in the cell internal resistance and wires.

When you say capacity are you referring to Whrs or Ahs? Also is that to 2.5V LVC for both 1C and 4C? I wonder how much of that is due to the bigger load dragging it down at the end of the discharge when the IR is at it's highest levels. I wonder what would happen if the LVC was reduced to 2.0V and/or the cell discharged with a 4C load was continued on a 1C after it had hit LVC on 4C. Does your test unit use a 4 wire system, if not that could make a significant difference?

I wonder if the reason for recommending a higher charge voltage is more down to charging time? It would be interesting to see the charge plots for different charge voltages CC/CV and see how they compare.

Good info, thanks for sharing.
Paul
 
cell_man said:
When you say capacity are you referring to Whrs or Ahs? Also is that to 2.5V LVC for both 1C and 4C? I wonder how much of that is due to the bigger load dragging it down at the end of the discharge when the IR is at it's highest levels. I wonder what would happen if the LVC was reduced to 2.0V and/or the cell discharged with a 4C load was continued on a 1C after it had hit LVC on 4C. Does your test unit use a 4 wire system, if not that could make a significant difference?

For capacity I was referring to mAh.

Discharge ended at 2.5V. The dropoff is so steep at that point that there is essentially no difference going to 2.0V. Take a look at the current plot in my first post. It is dropping vertically. It only takes a few seconds to drop from 2.5V to 2.0V. The 1C discharge curves are pretty much the same shape.

My battery test system (which is just my capacitive discharge FET based welder) does not use a physical 4-wire connection, it implements a virtual 4-wire connection in software. What you do is enable the load and measure the voltage at the cell with an external meter. You then adjust a "wire resistance" parameter until the unit's voltage reading matches the value at the cell. Once it knows what the connection resistance is, it can calculate the voltage drop in the wires due to the current flow.

For the connection that I was using that is 0.080 ohms. I have checked the calculated cell voltage many times at 1C and 4C and it is pretty stable and accurate. I have since made some much better clips with short, heavy gauge wire and will connect them directly to the unit with ring terminals. They have a resistance under 0.010 ohms (BTW, those spring loaded black plastic battery holders are TERRIBLE. They have over 0.1 ohms per cell).
 
For capacity I was referring to mAh.

I haven't used Ah as a measure in years. Wouldn't wh give you a more accurate measure of capacity, since mah does not take into consideration decreasing voltage during the discharge cycle? IIRC docs headway discharge rate tests showed a discrepancy in Ah that was evened out in WH.

John I wonder if you were thinking of me I've taken heat for single-cell charging to 3.60 for a while, guess I can drop to 3.55 now.

-JD
 
Hi Texaspyro,

I'm sorry I think there might be something else going on that is effecting your results. Please don't get pissed at me :) You shouldn't be getting that much loss in capacity at 4C discharge. I've seen independent tests, A123 test data and my own testing and it all suggests there is no marked reduction in capacity until the discharge rate gets to much higher levels. Attached is a document that includes some independent tests. I'll try running a cell at various discharge rates and get some figures, but I can't promise.

View attachment A123 cell info.pdf
 
oatnet said:
I haven't used Ah as a measure in years. Wouldn't wh give you a more accurate measure of capacity, since mah does not take into consideration decreasing voltage during the discharge cycle? IIRC docs headway discharge rate tests showed a discrepancy in Ah that was evened out in WH.

Pretty much all battery cells are spec'd in terms of mAh capacity, so that's watt I use. :roll:

And mAh does take into account the decreasing voltage. I am discharging with a fixed resistor. As the cell voltage drops, so does the current through the resistor.

To calculate the cell capacity I sum the milliamps of current drawn over each sample interval (5 seconds in my case):
amps = cell_voltage / load_resistance; // load resistance includes wire and cell connection
mah += (amps * 1000.0F) * (sample_interval / 3600.0F)
 
cell_man said:
I'm sorry I think there might be something else going on that is effecting your results.

There could be... perhaps my load resistor is changing as it heats up. It is a 0.250 ohm, 100 watt resistor made of 4 paralleled 25 watt sand resistors. It is only burning 20 watts, so I don't think it is changing that much. Also the cell that I have been testing is from a DeWalt pack that had a flakey BMS. It seems to be good. I chose it since it gave very repeatable (+/- 10 mAh) capacity values at 1C.

I'll do some more high current tests with the good battery clips and a new cell. Will also measure the resistor when it is hot.
 
texaspyro said:
Pretty much all battery cells are spec'd in terms of mAh capacity, so that's watt I use. :roll:

Pretty much everyone here uses wh, not AH (and definitely not mah when you are talking EV-sized discharges), so. :roll: :roll: back atcha!

But if you want to work with manufacturer numbers - cells are also spec'd in terms of Nominal Voltage - do you also use nominal voltage to calculate the number of cells you can serial without popping your controller's caps? Of course not, it is a useless spec. Also, until very recently, the spec'd AH were usually generated in fantasy land, so we stopped using it.

We aren't discharging through resistors in controlled conditions, so ah is pretty much useless for us, especially when comparing different chemistries. Seems more accurate and much much easier to get wh from a CBA-II or CycleAnalyst than work through your formulas, but thats me. Nothing wrong with you doing your own thing, if you don't mind the extra barrier to effective communication.

As cell-man points out, your numbers don't look right. Like I said, a few years back we all thought that higher c meant less power from the cell, but found that the higher-c discharge ran longer at lower volts and ended up generating about the same watts. I think the HeadWay 38120s actually gave a little more wh at high C, I think we theorized cell heating contributed to higher yields.

-JD
 
oatnet said:
Pretty much everyone here uses wh, not AH (and definitely not mah when you are talking EV-sized discharges), so.

And EVERY battery maker specs their batteries in mAh or Ah, so that is the measurement that I am making. It is rather trivial to calculate Wh instead of mAh. W=V*V/R, A=V/R... just an extra multiply. In fact there is a POWER plot in the menus. There was also an ENERGY plot (Joules a.k.a WATT-SECONDS) but I changed that to mAh.

Since I am currently measuring single 2300 mAh A123 cells, I report results in mAh. If I was testing bigger cells, I would use Ah. My little LCD screen has limited space when labeling plots. By using mAh (without a decimal point), I get 5 significant digits. If I used Ah with a decimal, I'd get 4 digits. (And the hardware is capable of 5 digit accuracy on its measurements) I can display up to 99999 mAH... call it 100 AH. Beyond that it scales back to Ah.

My hardware is really no different than a CBA-II, except I use a fixed resistor as a load (and I can spot weld/cut metal/anodize and plate metal/supply regulated power). They use some FETS to simulate a resistor. The basic operation is the same. Apply a load to the battery and measure the cell voltage. Calculate the current (or measure it across a shunt resistor). Then calculate the power or current integrated over the discharge time.
 
cell_man said:
I'm sorry I think there might be something else going on that is effecting your results. Please don't get pissed at me :)

I have dispatched a load of Ninja assassins to your location... :twisted:

And there does seem to be something going on with the higher current discharge... It looks like my external load resistor is shifting value by quite a bit as it heats up. By the end of the cell discharge it was 0.200 ohms... a 20% change! I need to heat sink it/stick a fan on it/get a better type of resistor. The internal load resistor for 1C type measurements has a fan.

Also, I put on the good clips/leads. They have a 0.005 ohm resistance vs 0.080 ohms. I am now getting a 12A/36W/5C discharge
 
I have fixed my high power load resistor. I separated the 4 individual resistors and soldered them together at the corners of 2.5x2.5 inch 10 mil copper end plates. I have a fan blowing through the array. Along with the much better clip leads, I get over a 5C discharge rate (12 amps). It now looks like that there is no significant difference in capacity between 1C and 5C discharge rate.

The cells do get pleasantly warm (around 25F temp rise) discharging at 5C. At 1C there is no detectable heat rise. I'll dig out the Whal thermocouple calibrator/thermometer and get some better numbers.

I found some 0.7 ohm, 50 watt, 3%, "gold" aluminum cased Dale resistors. I am going to build an even better load resistor with 3 of them...

Testing on 2 different new A123 cells hot off the charger (and charged outside where it was 100F) Cell 2 was a bit hotter off the charger. It was brought in from outside and tested without any chance to cool.

Cell 1:
@1C: 2248 mAh, 6862 mWh
@5C: 2237 mAh, 6848 mWh

Cell 2:
@1C: 2277 mAh, 7243 mWh
@5C: 2337 mAH, 7249 mWh
 
Damn, just when I was expecting a parcel.... otherwise I would have left town :)

I know how difficult it is to iron out the bugs when building something from scratch, I'm sure you'll get it sorted. I thought a man of your skills would have knocked up a constant current load, although I personally feel a resistive load isn't so bad. It's just a bit difficult to rely on an assumed resistance value, especially when the voltages get low and the current gets high, contact resistance can also be rather variable IME. I made a resistive load setup for checking 8 series pouch cells using a cell log to monitor and disconnect the load on LVC. It's reasonably consistent, but you need to keep the copper bus bars I use for linking the tabs and the tabs themselves nice and clean or it can effect the reading significantly. Ideally you really need to measure the current with a shunt, that way any additional contact resistance shouldn't effect the capacity values as much.

Anyways, good work, thanks for sharing.
Paul

Edit: didn't see the above before posting. Those figures are more like it :)
 
cell_man said:
I thought a man of your skills would have knocked up a constant current load, although I personally feel a resistive load isn't so bad. It's just a bit difficult to rely on an assumed resistance value, especially when the voltages get low and the current gets high, contact resistance can also be rather variable IME

Yes, I could easily build a REALLY GOOD CBA device. This one is a freebie extension to my FET based capacitive discharge welder. It is just firmware glommed onto the welder... at least for the 1C discharge. Larger discharge rates need that external $2 resistor. The welder has the ADC and internal 1 ohm switched load and the ability to switch several hundred amps continuous externally without breaking a sweat. I can calibrate the 10 bit ADC in the micro to produce sub-millivolt absolute accuracy with 5 second gate times. Basically everything to do battery analysis on the cheap.

With a little cutting and jumpering, I could modify the welder PCB to do an adjustable constant current load. It has 18 IRFP2907 FETs that can handle 200 amps continuous each (but I would need to add heatsinking). It has a quad op amp that could be a shunt amp/gate drive. Once the board was hacked up, it could not be used as a welder, but would make a killer CBA.

I think that if I was going to build a real battery analyzer, I would take one of my HP3457A 7.5 digit mulitmeters (they can be had for a few hundred dollars), some relays and load resistors (even lamps would do), and a little firmware on a MegaDonkey. I have really good GPIB firmware for talking to the meters. They can do 4-wire Kelvin reading, amps, microvolt res voltage readings, etc. A multi-cell analyzer that rivals any commercial product would be doable in short order...
 
I measured the cell temperature rise. At a discharge rate of 5.4C (avg 12.449 amps) the cell temp went from 78F to 98F over 635 seconds of discharge. Room temp was 74F. It had been charging outside at 82F.

I also tested a cell at 1C. I had to insulate it in a dewer to measure its 3.5F temp rise. Without the insulation and sitting on a table, the rise was under 1F.

I have a cell chillin' in the freezer at 5F... we'll find out how they behave in Eskimo land...
 
Thanks to Texaspyro and others for their constant stream of practical useful information that can be understood by all!

Is it fair to say that smaller battery packs, say 24v 10Ah, with therefore fewer cells, are easier to keep balanced, and therefore harder to pull out of balance, than larger packs using more individual cells?

Thanks.
 
Back
Top