March 14 2009

Battery.

The battery. Everyone talks about the battery. It is your biggest concern, biggest expense and whole EV is only as good as its battery. Typically it is also the biggest limitation in your EV, underscoring its utility. It is the only disposable item in the car you will replace sooner or later (for the same or different type) and keep replacing if you keep your EV long enough - sort of like an and oil change in conventional engine: no matter how gentle you drive, it wears out. There are so many choices, and at the same time - really so few worthy, at least at the time of this writing. After experimenting with Lead acid, NiMH, LiIon and now LiPo chemistries, it is quite clear that there are two winners here: NiMH and Lithium based battery chemistry, capable of delivering what a practical EV would demand. But this is from technical point of view. Reality is, there are political, economical and other reasons for advancing or suppressing what is best overall, so I'm not going to debate why you cannot have excellent NiMH battery in your EV. Ask GM. I'll focus on Lithium which is common type of battery made in millions for all kinds of gadgets and applications, and in recent years was scaled up enough to become the battery of choice for more-less advanced freeway capable EVs.

Brief run down on the particular chemistry flavors (mainly differentiated by the cathode material):

- Lithium Ion (LiIon)
- Lithium Manganese (LiMn)
- Lithium Cobalt (LiCo)
- Lithium Polymer (LiPO)
- Lithium Iron Phosphate LiFePO₄)
I'm sure other fancier materials will be applied in future as well as better manufacturing processes (such as nano-technologies, basically drastically increasing chemical reaction area)

LiIon is the oldest and most mature technology but falls out of fashion due to relatively unstable properties and not as good power and energy density as other options. There are plenty of info out there comparing between different types of batteries; sometimes objective, but most often biased (released by the battery manufacturer). I'm going to concentrate on LiPo choice for Audi - as you know from the simulation page this is the battery I'll use. As this page is not meant for academic education and I just describe what I did, I will only briefly compare LiPO type with the fashionable now LiFePo ₄ everyone seems to crave for. What is the advantage of it? It is considered safer battery in case you abuse it (in an accident or by inadvertent mistreatment by wrong charging or discharging). Manufacturers blatantly will tell you that their perfect battery will not catch fire as LiFePo ₄ is almost immune to it. Of course, it's a lie. In manufacturer's tests orchestrated to prove the point (read - boost sales) this might be the case that LiFePo ₄ and other type of battery are placed in the environmental chamber and conditions (overcharging or high temperature) gradually made harsher to see which type fails first. Say, LiFePo ₄ cell is compared to LiPo one. LiFePo ₄ cell resists abuse longer and once conditions are made so that LiPo fails or catch file, the test stops and LiFePo ₄ is declared a winner that don't catch fire. However, if conditions would get worse, LiFePo ₄ cell would burn just as cheerfully. Don't believe me? Check out for instance thislink accompanied with thesephotos - here was the vehicle equipped with LiFePo ₄ pack that "never burns". Whether the owner made a mistake or charger failed to stop, it's unclear, but this is not the point. The point is, as any battery salesmen, manufacturers obviously lie about LiFePo ₄ ability to burn, while this chemistry may well be relatively safer than LiPO or others. Bottom line - still burns very well. Welcome to copy these images, and next time a salesman will try to sell you safe LiFePo ₄ battery, show him this proof and ask if he will bet his life or at least his salary on the fact that LiFePo ₄ is impossible to ignite and watch his reaction. It's fun.

Another reason everyone runs after LiFePo₄ - most of the batteries of this type are made in China, so it can be bought cheap, for instance famous Thunder-Sky who use to sell crap (ask me how I know), but is more careful these days. Economics aside, let's compare the most important parameters relevant for your EV. Fundamentally, what do you want from your battery?

- Move your EV far enough (whatever YOUR definition of "far enough" is).
- Be powerful enough
- Able to get charged fast.
- Last long time ("long" means as close to the live of the vehicle itself as possible, ideally 10 years or so).
- Be safe
- Be physically light and small
- Be maintenance free (install and forget)
- Be a commodity (readily available from multiple sources)
- Provide sense of confidence in its performance, typically based on reputation which in turn is based on someone else's previous experiences,
- Be cheap
- Be cheap
- Be cheap

Last three points contradict (and always will contradict) with all the points above them.

So basically you want all the qualities above but don't want to pay a lot of money for them. OK, lets get to reality and as with everything in life, compromise. Don't expect me to define the "best" combination here, it's the same as trying to tell you what's the best vehicle out there in general. All depends on how much value you personally assign to each of these qualities and how you prioritize them. Only you know that.

In my case what I looked for is basically one main parameter - energy density. This is because I wanted highest capacity with minimal weight. The size of the battery then becomes a compromise between how far I can drive on one charge and how much battery I can afford. LiPo is the winner as far as gravimetric energy density (its Wh/kg ratio is higher than that of LiFePo₄). I want to raise voltage high enough and that dictates amount of cells. I'll target 192 cells battery. Why 192? Because I will have 6 blocks 32 cells each. 32 cells make up 118V block and that voltage has some advantages. Another choice is to have 48V blocks which depending on the BMS choice may be even more advantageous, but I'll have to start somewhere and will experiment later.

Here are my thoughts. Any battery (well, actually everything surrounding us in life) is affected by the temperature. There are 2 choices to deal with this:

a) adapt charging / discharging condition to actual temperature of the battery, compensating affected parameters per manufacturer's recommendations (if they have any...);
b) maintain about the same temperature of the battery regardless of the ambient temp. Then you don't have to deal with above, but need extra effort (and hardware) to implement it.

There may be a combination of both.

I chose method b) above. Basically, its cycle life will not be affected by the temperature, and thermal control is electrically trivial to do. The trick is to provide good thermal contact with individual cells. As everything in Audi, the battery will be liquid cooled (it's a customary expression, since sometimes battery will be warmed rather than cooled). Air cooling will require larger battery (gaps between cells for air circulation), ducts, noisy fans, etc. Pretty much the same arguments as water cooled vs. air cooled motor.

After debating how the BMS is going to fit and how do I take care of thermal issues, initial concept materialized and got captured in my work book. Basically, because the cells are sealed in poly pouch and have soft tab terminals, it presents challenge to hold them in place - there is no place you can bolt to, including terminals. So in each box the cells will be staked in a row and one large thick base PCB will cover all of them. The tab terminals will be inserted through the long cutouts in the base PCB, folded above and clamped between aluminum bars. The BMS boards will be positioned vertically and inserted into the connectors on the base PCB. This allows to replace the BMS boards with different type without touching the box. There will be heat sinking aluminum sheets between cells; they extend beyond sides and clamped between spacers equal to the cells thickness. These spacers will be bolted to the aluminum sides with machined water path. There is seal sheets between machined sides and the spacers, forming hermetic channel when squeezed together. To assure no leaks, there is a grove and o-ring around the bolts holding sides. That's the plan. It's easier to see than explain, so here below are photos of what I did.

Concept of the battery construction.
 Mean time cells were ordered. They arrived in a few large crates, straight from Kokam Korea.
The cells are about size of a notebook, just square. Another challenge upon closer inspection: they are not quite equal thickness across. Thicker toward terminal tabs, thinner toward the bottom.
Soft tabs are 0.2mm thick. One made of nickel and another - of nickel alloy, one is softer and more easily bent than another. They are flat so I will need a jig to cut to size and punch holes.
All the construction of course is done in CAD allowing to check fit. About 20 iterations and final drawings were done.
I made a simple jig to trim the tabs. Careful, it cannot be all metal, it can touch one terminal at the time.
Few days of work and all the tabs were cut and punched, and cells stacked to get ready for assembly into the pack.

There are quite a few parts had to be made to accomplish design goal. For the mass manufacturing this may not be the optimal design, but for few boxes and initial proof of concept it's worth the trouble. The trick is to design to use as close to stock material as possible - this will minimize machining time and expense. Here is what was made:

Top and bottom tab terminal clamps. Top one will also hold a cell's BMS PCB.
Spacers. There are 5 groups with staggered holes. This is done to spread the compression and distribute the heat as evenly as possible across entire sides of a box.
Front and rear "book ends" and left and right sides were machined for all the boxes. Beautiful work.
The end terminals connecting first and last cells to a box studs - these will be + and - of the box.
Stack of heat sink sheets with holes punched along edges.
The insulating plates holding bolts (studs) CNC machined from 12.7mm thick polycarbonate.
Bare base PCB overview.
Same PCB populated with connectors and bottom clamps bolted on.
The tab routed BMS PCBs. On this photo only SMT components are populated.
Precise calibration of the DC-DC converter of a node - this is fixed final voltage any cell will be charged up to.
Tedious and repetitive work, but one after another, all the nodes were SMT assembled, and then calibrated by hand. Here the trim pots being fixed with locktite.
Here I got help to pre-assemble the nodes. The top clamp was bolted to each node.
Finally electronics was tested and got ready to get installed! There are two layouts of he same circuit - right and left. This is due to the fact that even and odd cells have terminals facing different direction (say, even - positive to left, and odd- positive to right looking from one book end), but the PCBs have components on one side - toward one end (not alternating orientation). Again, it will be easier to see below than explain.

That's all regarding parts. Now, the fun part - assembly.

Initially I assembled the first box stacking up cells. This worked OK, but it was difficult to align them all and after assembly it turn out the row of cells is not quite parallel to the aluminum side, and so is the base PCB sitting right on top of each cell against their top seam:

Initial check if the tabs freely go through the cutouts, don't touch anything they're not suppose to and verifying clearances. So far so good.
Close up photo - the tab will be bent over the bottom clamp...
...and squeezed between it and the top clamp. Another tab from adjacent cell will be inserted between the clamps from the opposite side (not shown here). The bolt here is loose, just to see overall fit.

I started assembly on a plank of wood, laying down bottom bookend and stacking cells up. An insulating sheet was cut and placed on each cell covering its terminals right before aluminum sink sheet is placed onto the threaded rods against that cell. This is to prevent shorting cell's tabs by the sink.

One bookend is first and first cell is on top. Its tabs are inserted into the slots in the base PCB. Stainless steel rods will line up everything and tightening big sandwich at the end if assembly.
Next go side spacers.
Next - heat sink, another cell, heat sink, and so on. Granted, the cells are alternating face up and face down to get terminals interconnected in series. I decided to place spacers in after all the cells are stacked up.
Half way through. You can see the tabs sticking out of the slots in the base PCB. Careful - if they bend - they short. If they short - the tabs will act as fuses but spectacular arc destroying the tabs will accompany the show.
Same photo from the side.
After last cell is in place, the top book end goes last.

As I mentioned, after tightening the rods  I realized the t the stack is not exactly vertical. It is nearly impossible to press (and keep pressing) each cell against base PCB while assembling. It had to be another way. I decided to put them all on a jig, assembling the box up side down, so each cell will be laying on top of the base PCB and they all will line up under own weight as long as the base PCB is flat. Well, this is what the jig is for - it holds the edges of the base PCB along long sides allowing the cell tabs to get between these sides. I have to adjust the gap between jig's sides to be about 1mm wider than the distance of outside edges of the tabs. This proved to be just enough. Here:

The jig. Not very high tech, but very handy and does the job.
The aluminum sides and threaded rods are prepared.
The base PCB is on the jig face down. First book end is positioned and threaded rods are inserted.
This time the spacers will be placed together with each cell, not after assembly.
Preparing the first cell. The tabs are soft so I needed to watch out not to bend them. Not visible on the photo, but there is clear plastic between tabs and book end aluminum to prevent electric short.
The cell tabs are inserted half way into the base PCB.
Plastic insulator sheet is visible here.
Installing heat sink separator sheet.
It is pressed against first cell, and insulating sheet is placed right after this step, so next cells tabs, if touch separator before get into base PCB's slot, will not get shorted.
Next go side spacers.
Next sink separator sheet.
...and so on. Half way through.
Last cell is placed.

Because all the cells are up side down (terminals facing down), their weight automatically aligns them all on the base PCB. The PCB rests on the jig, so as long as the jig is straight, the assembly is as ideal as it can be; each cell protrudes through the PCB the same amount so the holes in tabs will line up with holes in the clamps after folding. This is important because I don't want to accommodate inaccuracy by oversizing punched holes thus reducing contact area of the tabs. The bolts clamping tabs are M5 and the holes are 6.35mm diameter - just enough slack but not too much missing material around holes.

The angle view from the bottom. You can see that the cells tabs are just fit between jig sides, no gaps.
The second book end is put onto the rods.
Side view. The whole stack can wobble side to side easily, which is good at this point - doing so self-alignes the cells settling everything in place.
This is critical - every so many cells I inserted thick foil shims to get total cell+shim thickness slightly more than the thickness of side spacers. Because only friction holds cells in place, I need to make sure when the stack is tightened, the cells are the ones which are going to be squeezed hard, not the spacers. Because of the cells thickness tolerances, choosing amount of shimming is trial and error process. Trying to compress the stack reveals whether the cells or spacers are going to get squeezed the hardest.
Another place for shim installation.
Initial compressing of the stack. The goal here is to line up staggered holes in spacers with holes in the seal sheets and machined sides.
A couple of adjustable C-clamps allow to squeeze right and left side independently.
Making sure the stack is true rectangular.
The side seal goes first.
Whoops, forgotten thread for the inlet fitting. Fixed.
Regular Teflon tape around NPT thread of the fitting.
The fitting being installed.
All 4 fittings are installed on the front book end. Later on excessive rods ends will be cut off.

To make sure no water leaks happen, the o-ring will be installed. First there is sealing compound applied into the groove for o-ring. The ring (2.3mm diameter) is thicker that the groove depth (1.6mm), so the ring is squeezed to guarantee good seal. Still to fill all the microscopic gaps between aluminum and rubber, compound is a good idea. It also seals the gab where the ends of the rubber cord meet butt to butt (it is not pre-made o-ring but a stock rubber cord on a spool).

Filling the grove with sealing compound.
Placing the cord at arbitrary starting point.
It follows serpentine grove.
Amount of compound is liberal enough to get squeezed out and cover the ring from top. Now it is ready for the side installation.
The side with o-ring lined up with the seal sheet.
Both are placed on side and bolted to the spacers. Of course the bolt pattern in the machine side match the pattern formed by stacked spacers. Same story on the other side.
Let's test for leakage. No problem after 1 hour.
Turning the box on one side. Now I can deal with the tabs.

Now time to do electric interconnects. Softer positive tab bent first as it is easier to get sharper radius. The negative tab of adjacent cell goes over. At this point I must be careful as all the cells become connected in series and there is 118V between end cells. And, it's about 4.7kWh worth of energy, so if this thing shorts, spectacular fireworks are guaranteed.

Bending the tabs over bottom clamps.
Installing the BMS PCBs. They are bolted to the top clamp.
Inserting the connector into the mating part on the base PCB. The top clamp held by 3 M3 bolts will prevent it from ever disengaging.
I made an error in the layout (which was not feasible to predict) - one of the bolts holding the clamp when unbolted is interfering with the transformer on the PCB. Assembled, there is no problem, but getting to this bolt is difficult. So I had to make this tool by grinding L-Allen key. This error is already fixed for future PCBs of this type, but for now I have to deal with all the boards already made this way.
Installing end terminal connecting the first tab to the stud. 
Another photo of this. Two bolts are enough to make reliable contact.
The terminal plate is ready.
Tightening the hut over the end terminal. The stud is M12 Stainless steel bolt with machined head. The head is recessed into the polycarbonate plate so doesn't tough the aluminum book end.
This box is the first (the most negative) in the chain, so the battery shunt gets installed here. Note, the stainless steel bolts everywhere only used for mechanical strength - they do not conduct electric current.
Another photo of the shunt installation.
Some nodes are equipped with thermistor touching the part of special separator plates that have narrow neck protruding the base PCB through the special cutout. Thermistor can touch this neck thus measuring temperature of the separator. However, this proved to be redundant as the water temperature is very good indicator of he pack temperature. Overheating of individual cells is undetectable by these thermistors and will have to be identified by the cell behavior under load. So most of the nodes were not equipped with thermistors although electronic circuitry just in case is there to do it. Will see how useful this circuitry is.
All the nodes are lined up nicely. Turn out to be very sturdy arrangement.

Finally, the box assembly completed. it is checked for water leaks and just has to be electrically tested. The stack was charged a bit and the BMS activated. The cells came from the factory amazingly uniform, the voltage difference between cells was no more than 0.02V. So there was no point to balance the pack, I just needed to make sure all the circuitry functions as designed. Yellow LED indicates balancing charging is taking place individually per cell. And because the pack is not full (each cell is less than 4.15V), every node should be active trying to bring its cell to that voltage level. So, this is easy to check - all yellow LEDs must be on as soon as balancing is enabled.

Test of assembled pack. It's hard to see the LEDs, but they are all ON.
Another view of the box and the BMS being tested.
This photo is prettier - taken without flash.
Well, this is about it. All I need now is to repeat all above described on this page 6 times and get my battery done. On this photo only 4 boxes are completed.
Another view of completed battery ready to get installed. They turn out to be quite OEM looking and worth my electric Audi A6.

Next I think will be the charging solution and I need to settle on the drive system. Now that EVISOL cheated me (and few others) and failed to supply working hardware, I may be forced to think about different drive system. There is quite a few things to do in the vehicle independent of the drive system, so this will allow me to wait before making final choice but not delay completion of the project. I must take break to prepare for EVS-24 in Norway which is about 3 weeks away, and will resume when I come back. So if you like what you see so far, stay tuned.

NEXT...