Playing with A123 20Ah lithium cells

[Daox]

Originally Posted by bwilson4web

I'm starting to look for A123 charging protocols. I didn't find a clean description at their web sites. So these are my initial thoughts.

I'm looking at two, battery management chips:

• LTC6803 (Linear) - includes integrated MOSFETs for discharge balancing. However, discharge balancing limit appears to be 80 ma for a single cell, 0.25W. With up to 12 cells, two are needed.
• BQ76PL536 (TI) - external balance MOSFETs, current is design selectable. With a limit of 3-6 cells, three are needed configured 5 cells each.

Near as I can tell, these are series, voltage measurement chips with modest temperature monitoring. The LTC6803 has a modest cell balancing, up to 80 ma, one cell at a time, 0.25W capability. The BQ76PL536 drives an external MOSFET which increases the discharge current that can be used to discharge a cell.

I've been thinking about cell balancing and two approaches come to mind:

1. Discharge cells to a lowest dV uniform state and then when all are at the same low, state, let charging continue to when the first cell achieves a limiting, dV. To achieve reasonable times, the discharge needs to handle a hefty current, 1-5A. Once initially balanced, the discharge draw-down would be fairly short so 1A may be feasible.
2. Top limit, uniform using a shunt regulator. This regulator can quickly become pretty complex if one tries to keep it efficient. A simple, series Zener stack would work but the total charge load would remain constant until the last cell reaches the Zener limit. Again, as the cells become balanced, the time during which individual cells complete their charge becomes smaller but at the end, the Zeners are handling the total heat load. Tapering the charge current at the cell voltages increase could mitigate the Zener load at the end.

You've had a lot more experience with balancing LiON cells. Any insights to share?

Also, looking at the solder tabs, have you considered spot welding? It would help if we had test articles but spot welding two or three 'dots' per tab would both block air and ensure metal-to-metal bonding. It would be mechanically simpler and only intra-tab insulation would be needed.

Thanks,
Bob Wilson

If you read the first post in my BMS thread you can see how I am doing things with my setup. I haven't really looked into the BMS ICs yet. I'm trying to get my current version working before moving on to using those.

As for balancing, I don't believe active balancing is really necessary. When initially assembling a battery pack you should choose how you are going to balance. You either balance at the top (fully charge each cell), or balance at the bottom (discharge each cell to a set voltage). There are advantages and disadvantages of both techniques. In general, bottom balancing seems to have greater advantages IMO.

The guys who have been running lithium cells for a while now have been monitoring their packs and have noticed very very little unbalancing over time. Personally with my PHEV kit, I have top balanced my pack mainly because it is easier. In about a year I plan to hook up my charger to each cell and see how much energy I'll have to put into each cell to top them all off. I am guessing it will be very uniform.

For charge protection things are pretty simple. The celllogs monitor all the individual cell voltages and when the first cell hits 3.65V the BMS shuts the charger off. This completely eliminates the CV portion of the charge. This charges my 39Ah mottcell cells (the cells in my PHEV kit) to roughly 95%. So, there is a small capacity penalty, but that also leads to longer cell life. This eliminates the need for any shunting and is thus more efficient too. It also elminates the need for a smart charger since the brains have been moved to the BMS.

[Daox]

Originally Posted by bwilson4web

Well done! Love your charts which I can't find from the vendor.

Just a couple of questions. I've re-ordered them to look at charge and discharge:


Looking at these excellent discharge curves, I get the impression regardless of the starting voltage, the bulk of the energy starts at ~3.278-3.302V. Charging over this voltage does not appear to give any corresponding increase in total energy. So I'm wondering what is the smallest peak charge voltage that would give a starting discharge of 3.278-3.302V? Would a peak charge of 3.5V still put the cells in a state that they still start their discharge at 3.278-3.302V?

I'm quite happy to help. The biggest hurdle for me in using lithium was understanding the differences between it and lead. I'm also finding it interesting to compare my findings with the A123 cells vs the older and prysmatic type mottcell cells.

From the charts, it would seem like 3.5V would give you a fair amount of capacity. I will test it and post the results. If you are really interested in the raw data I can post a link to the software (which is free) I am using for my charger and upload my datalogs.

At the end, it looks like 3.017-3.055V is also where the available energy pretty well falls off. You went ahead and stopped it at 2.813V but I'm thinking of cell powered, battery management. I might stop it at 3.0V only to make sure the BMS logic continues to run without external power ... for a long time. My rough estimate from your charts is about 2Ahr, 10% of the rated capacity so I might go a little lower depending upon BMS load.

That might be doable IF you have a constant and low load. These tests only show up to a 2C dis/charge rate. They are designed to take 10X that! The voltage sag obviously increases as you draw more current. Unfortunately, 40A is the limit of my charger. The other big factor is temperature. I've noticed with the mottcell batteries in my PHEV kit, that in winter the voltages sag quite a bit more. This is more than enough to totally screw up that low voltage setpoint. So, as the temperature changes you'd have to account for that.

What first attracted me was the initial 3.431V on both charge cycles. Given the discharge went to 2.813V, we're looking at ~0.62V 'recovery' after the discharge ended. After the cells are discharged to 2.813V, have you measured the unloaded, recovery voltage, what they float to once the discharge ends?

The reason I ask is the open circuit voltage after discharge, what ever it float up to AFTER discharge to 2.813V, helps us understand the initial V*I, power needed to start the charge cycle. It also gives an idea of the dV needed to force charge into the cell.

In a similar fashion, what happens to the cell open voltage once the charge completes? Does it droop?

I haven't specifcally noted their voltage recovery after discharge, but it would be similar to what you see as the voltage at the beginning of the charge cycle. Looking at the charts it seems to be around 3.24V. The reverse is true for the voltage drop after charging. It looks to be around 3.43. When I do my the above testing, I will confirm these numbers.

Thanks for the great questions Bob!

[bwilson4web]

You're going after a 12V replacement and I'm going after a 48V eBike replacement for a lead-acid battery. In my case, this is my first thoughts:

Risky areas to resolve:

• series shunt regulator - I've specified a 3.6V Zener but there is also a 3.3V that might be a better part. Bench testing will help resolve it. Also, I need to SPICE the Zener and current limiting resistor with both parts. From DOAX's charts, it looks like the 3.3V might be a better part because the current limiting resistor gives a slight, over-voltage to achieve peak charge.
• MOSFET + Schottky relay - I may have the wrong symbol for the MOSFET and it isn't clear if I'll need a reverse biased Schottky to pass regen current into the string. I've also shown a directly driven gate and this might be a case where reverse bias of the MOSFET puts the gate and attached microcontroller at risk. My fall-back is a relay but "UGH!"
• 30A vs 50A - the eBike has a 0.5 hp motor which should be ~15A sustained but I haven't measured the peak current draw.

I haven't seen a series, shunt regulator like this before but it is an obvious solution. Using ordinary Zeners means I need to model the thermal effects. In theory, as the part warms up, the Zener threshold voltage should go down which increases the draw from the peak cells towards a normalized voltage for all cells. In other words, helping to level-charge the peak cells. The key will be a collective heat-sink for the 15 Zeners.

I have not show thermistors for cell-strings. If I can have one for each of the LC6803s, it should be fairly straight forward. I did go with Linear instead of the TI part because I only need two instead of three chips and it has a 1/4W, built-in, per-cell, MOSFET. After all, what is 'time' to a computer and it may be this low, rate might be enough for cell balancing and not require the series shunt.

Bob Wilson

[bwilson4web]

Originally Posted by redpoint5

. . . Would these cells be ideal for an alternator kill switch setup?

If I assume a 5% increase in FE using an alternator kill switch, . . .

Daox pretty well nailed it. In my case, I only have a very heavy, limited range, lead-acid battery for the eBike which has an effective radius of ~3-4 miles. I'm expecting the effective radius to increase to 15-18 miles and my work is 10 miles away. My dual-stack design means no 'range anxiety' and the modest charge rate, ~250-300W load, means any 110VAC outlet will work.

It is my expectation that the $800 of LiON cells and $200 BMS will give me a fair-weather, electric bicycle. It will also give light-duty, cardio-time during the commute (it still has pedals.) Yet I won't add much more time to the morning commute once I have the route worked out.

Right now, I'm buying ~10 gal every three weeks, call it $40/3 ~= $14/week commuting cost. Due to weather, I'm not expecting to go 'gas free' as much as cut that down saving about $4-5/week. So we're looking at $1000/5 ~= 200 weeks or nearly four years to achieve 'pay back.' But if I pull off another 30-40 lbs, it will be much cheaper than some exercise club bill. <grins>

Plus ... I get to sneak-up on pedestrians. <GRINS>

Actually the real advantage is getting some 'hands on' time with LiON batteries and a battery management system. The experience gained is the 'tuition in the school of hard knocks.' It will satisfy my curiosity.

Bob Wilson

[Daox]

Originally Posted by bwilson4web

You're going after a 12V replacement and I'm going after a 48V eBike replacement for a lead-acid battery. In my case, this is my first thoughts:

Risky areas to resolve:

• series shunt regulator - I've specified a 3.6V Zener but there is also a 3.3V that might be a better part. Bench testing will help resolve it. Also, I need to SPICE the Zener and current limiting resistor with both parts. From DOAX's charts, it looks like the 3.3V might be a better part because the current limiting resistor gives a slight, over-voltage to achieve peak charge.
• MOSFET + Schottky relay - I may have the wrong symbol for the MOSFET and it isn't clear if I'll need a reverse biased Schottky to pass regen current into the string. I've also shown a directly driven gate and this might be a case where reverse bias of the MOSFET puts the gate and attached microcontroller at risk. My fall-back is a relay but "UGH!"
• 30A vs 50A - the eBike has a 0.5 hp motor which should be ~15A sustained but I haven't measured the peak current draw.

I haven't seen a series, shunt regulator like this before but it is an obvious solution. Using ordinary Zeners means I need to model the thermal effects. In theory, as the part warms up, the Zener threshold voltage should go down which increases the draw from the peak cells towards a normalized voltage for all cells. In other words, helping to level-charge the peak cells. The key will be a collective heat-sink for the 15 Zeners.

I have not show thermistors for cell-strings. If I can have one for each of the LC6803s, it should be fairly straight forward. I did go with Linear instead of the TI part because I only need two instead of three chips and it has a 1/4W, built-in, per-cell, MOSFET. After all, what is 'time' to a computer and it may be this low, rate might be enough for cell balancing and not require the series shunt.

Bob Wilson

I'm no electrical engineer (far from it in fact), but I'm a bit confused why you are using the zener diodes? The LTC6803 has mosfets inside it to activate shunting to each cell at a much more precise voltage than what those zeners can do.

edit: I read the datasheet a bit more (hadn't read much at all) and see what you are doing.

[bwilson4web]

Originally Posted by Daox

I'm no electrical engineer (far from it in fact), . . .

There are electrical engineers at work with less understanding. It is as if upon getting a diploma they decided they had all knowledge and lost any curiosity . . . if they ever had any. <sigh>

Originally Posted by Daox

. . . edit: I read the datasheet a bit more (hadn't read much at all) and see what you are doing.

There are precise, shunt regulator parts but they only sink enough current to operate a power transistor gate and require two bias resistors to set the value. So to make a functional equivalent to a precise Zener, it would take a shunt regulator, two bias resistors, and a third to bias a power transistor.

Although not as precise as I would like, a Zener might be 'close enough' since its thermal profile appears to be 'positive' in a good way for my purpose (aka., thermal runaway in my favor is no vice.) Now you can see why the voltage over 3.3V is so important. The current limiting resistor also raises the effective Zener threshold and hopefully enough to provide the additional voltage needed to achieve peak charge. Designed to handle 1A, it should be precise enough if I tie all of the Zeners to a common, non-shorting, heat sink.

One thing I've learned with the MSP430 is an ADC does not perform a precise, exact metric. Rather, it needs a series of measurements and averaging or weighted-averaging to get a stable value. I'm expecting the LC6803 will have similar characteristics . . . a poorly documented 'feature' of ADCs.

You mentioned the Prius cells are cold-weather, temperature sensitive. I don't see any technical data at A123 that addresses these characteristics. But on Monday I'll send them a letter and using your charts, ask for similar technical data. . . . I'll declare your testing to be at 75F/24C and ask for similar data as a function of temperature. In my case, between 50F/10C and 104F/40C. But I would accept your charts as gospel (assuming your wife doesn't catch the wires running in and out the fridge. <grins>)

One piece of good news, I only need 15A:

I might add some pad and go up to 20A.

Bob Wilson

[bwilson4web]

Hi,

Google found this source for technical details about the A123 battery:
http://revolectrix.com/support_docs/item_1229.pdf

The biggest finds are:

• 140F - dramatic loss of service life
• 170F - cell failure

What I don't have are the cold temperature effects.

BTW, they make series, A123 compatible, chargers that handle up to 10 cells. I've sent them a note asking if there is a combination that can support the 15 cells I will need.

Bob Wilson

[Daox]

Very interesting link Bob. Thanks for adding that. It would be interesting to see the effects of an updated study since that was from 2007 and is not 5 years old. I wonder how much progress has been made in that time period.

[bwilson4web]

Hi,

Originally Posted by Daox

Very interesting link Bob. Thanks for adding that. It would be interesting to see the effects of an updated study since that was from 2007 and is not 5 years old. I wonder how much progress has been made in that time period.

I sent a note and will follow-up with a phone call later today (first get rid of the Monday morning alligators!) But one thing bothered me this weekend and this morning until I realized my question.

The tabs are aluminum? Aluminum is one of the most reactive metals around and it is exposed to whatever the electrolyte is in the LiON cell. I used to fill balloons with hydrogen by using the lye, coke bottle, and aluminum foil 'rod' trick. I would have thought nickel would be a better tab material.

It has no magnetic characteristics?

Not to worry, once I get mine I'll do some heavy duty testing. <grins>

Bob Wilson

[Daox]

Originally Posted by bwilson4web

The tabs are aluminum? Aluminum is one of the most reactive metals around and it is exposed to whatever the electrolyte is in the LiON cell.

The aluminum and aluminized copper tabs are simply extensions of the anode and cathode. You might find this video interesting...

http://ecomodder.com/forum/showthrea...deo-15502.html