02Dec 2014

The National Transportation Safety Board (NTSB) released today its incident report on the battery fire that led to the grounding of the Boeing 787 fleet for many months. It is a thorough report, nearly 100 pages long, that examines in great detail all aspects of the lithium ion battery and its systems. The result is shared responsibilities for the battery fire, with blame targeting GS Yuasa, the Japanese manufacturer of the lithium ion individual battery cells, and Thales, the French system integrator responsible for the power conversion subsystem, as well as Boeing. Even the FAA shared some of the blame in how it managed the certification process.

The report reserved special blame on the poor design, test and manufacturing practices at GS Yuasa. In essence, the NTSB reports that defects in the assembly and manufacturing of the cells led to internal shorts in the battery. Inadequate testing during certification throughout the chain did not catch these problems. Specifically, these defects led to shorts and consequently thermal runaway in cells 5 and 6. The battery contained a stack of 8 cells connected in series for a nominal voltage of 29.6 Volts.  CT scans and disassembly of defective batteries showed the failure points — defects in the form of protrusions or wrinkling in the cathode-separator-anode winding that effectively created minute electric shorts between the cell’s terminals.

The first figure below shows a schematic from the NTSB report of the design of an individual cell. It consists of a sandwich of the lithium cobalt oxide (LCO) cathode layer, the separator layer, and the carbon anode layer. Aluminum and copper foils act as the electrical conducting planes for each of the cathode and anode, respectively. This material construction is nearly identical to the ones used in the consumer industry. These layers are then folded together to form the cell windings.

The next photograph shows how the cell windings are flattened before insertion into the metal packaging can. The NTSB pointed to this winding formation and flattening process as one of the culprits in the origination of the defects and shorts. This manufacturing process created wrinklings; this is a highly technical term referring to the buckling of the electrode foils, thus compromising their integrity right during the cell manufacturing steps. Additionally, the NTSB found that the electrolyte filling process was inconsistent with the practices used in the industry and possibly led to the incomplete formation of the SEI layer (it is a thin layer that forms between the anode and the electrolyte). In other words, the NTSB pointed a clear finger at the poor cell manufacturing at GS Yuasa.

So why weren’t these defects caught during the test and certification phases? Well, apparently the battery that made it on the actual B787 airplanes was different in design and properties from the batteries that underwent the tests. It seems that lots of heads at GS Yuasa and up the food chain to Boeing will be rolling — if they have not already. Shame on the engineers and the managers who supervised and led this process.

21Nov 2014

I described earlier the two pieces of the fast charging puzzle. There is first the power delivery from the wall socket to your mobile device, and then there is managing the charging power into the battery so it does not get damaged.

The first piece of the puzzle is all about power circuitry, something electrical engineers enjoy and revel in. But the second piece of this puzzle is about managing the chemistry in the battery. Here, I have some bad news: electrical engineers, even the most successful ones, are likely to have scored a C- or a D+ on their chemistry college classes; in other words, they are not comfortable with chemistry. Like all of us humans, we drift towards our comfort zones, so chemists drift to chemistry, and electrical engineers drift to building electronics. Getting the two disciplines to work together takes hard work and lots, really lots, of discipline. As a result, it is rare to find interdisciplinary development in the mobile world, especially between battery chemistry and electronics. We, at Qnovo, happen to be one of these rare exceptions.

Well, enough self-promotion; let’s get back to real life. What happens if overzealous electrical engineers decide to charge a battery as fast as they can. After all, they are electrical engineers and know how to design circuits that can pump a lot of energy in the battery — some even give it great names like Quick Charge. The result is simple: They will destroy the battery (if you use CCCV charging). The data in the next graph illustrates the extent of the damage.

The graph is a standard capacity fade curve for a battery. In this case, it is a 2,400 mAh lithium-ion polymer battery from one of the leading manufacturers. The vertical axis shows the remaining capacity of the battery (in Ah) as the battery undergoes cycling (shown on the horizontal axis).  As the battery is repeatedly charged and discharged, it loses its ability to hold charge, hence the degrading curve. This graph shows the capacity degradation for four distinct charge rates, varying from a slow 0.5C up to a superfast 1.5C, all using CCCV. At 0.5C, the battery charges in about 3 hours, and at 1.5C, it charges in a little over an hour.

So what is the graph telling these overzealous engineers? It says that at the slow charge rate, the user will comfortably obtain over 600 cycles of operation. That’s plenty to cover a year and more. But the superfast charge rate of 1.5C does so much damage to the battery that the battery will barely last a couple of months. In other words, fast charging with CCCV is bad! Isn’t time to use new methodologies that give us fast charging without damaging the battery.

The moral of this post: Don’t trust everything your electrical engineers tell you…well, that is unless they really scored A’s on their chemistry college courses!

14Nov 2014

You have heard it before from me: battery performance is getting seriously challenged. The battery vendors are having a difficult time, and I am being kind, in meeting the expectations of the mobile device manufacturers. Specifically, it is getting extremely difficult for the battery vendors to simultaneously deliver high capacity, fast charging, and long cycle life. But why is that? 

Let me digress briefly here and examine the average cruising speed of commercial airplanes over the past century. One will rapidly notice that airplanes have not gotten much faster in the last 50 years. They are more fuel efficient; they travel longer distances in greater comfort (well, if you pay for it); and they are far safer than they have ever been…but they are surely not faster. That’s because it gets very expensive to travel faster than the speed of sound (667 knots). The Concorde is a stark reminder of this economical limit. Chuck Yeager broke the physical limit of the sound barrier decades ago, but we can’t really change the economics around this limit.

Lithium-ion batteries are rapidly approaching a similar limit for energy density. Short of a major breakthrough in a new material system, we are staring at a difficult barrier somewhere between 600 and 700 Wh/l. With that, I mean achieving large-scale manufacturing with affordable economics that match the requirements of the mobile device industry. A key limiting factor is now the carbon anode material. It is possible that new carbon-silicon composite anodes can change this equation, but for the foreseeable future, these new composites will continue to suffer from poor cycle life and high manufacturing costs. Until then, the economics of rising energy densities will be severely disadvantaged.

Is there a scientific origin for this limit? There are plenty of reasons, but it is best to illustrate one: the impact of battery voltage on energy density. In a very simplistic description, higher energy density comes from packing more lithium ions inside the battery, as well as raising the voltage at the terminals (if you recall, energy is the product of charge, or ions, and voltage).

The voltage at the terminals is the difference between two voltages; that of the cathode voltage in reference to a fictitious lithium contact, minus the voltage of the anode, also in reference to that same fictitious lithium contact. This is illustrated in the graph below.

On the vertical axis, I show the voltages of both the cathode vs. lithium (top) and the anode vs. lithium (bottom). On the far left of the chart, i.e., when x is approaching zero, the graphite is void of lithium ions and cobalt-oxide is totally full of lithium ions. This is when the battery is “empty.” On the far right, when x is approaching one, the opposite is true; the battery is “full.” I specifically refrain from saying x = 0 or x = 1. At these extremes, bad things happen. When x = 0, the physical structure of the cobalt-oxide alloy is greatly damaged. This limits the low end of the battery voltage to about 3.0 V. In other words, never discharge your lithium-ion battery below 3.0 V; the risk of irreversible damage is great. Most smartphones actually cut off near 3.3 V (this is really when your phone reads zero percent).

At the opposite end when x = 1, lithium ions combine to deposit (or plate) as a metal.  The anode structure is also under immense mechanical stress. Additionally, when the cathode voltage rises past 4.2 V, the electrolyte begins to oxidize (and ultimately decompose). This effectively limits present-day lithium-ion batteries to a maximum voltage of 4.35 V with the understanding that the “bad stuff” begins to occur past 4.0 V, and becomes unsafe past 4.35 V.

To raise the energy density in the carbon/cobalt-oxide material system, one needs to raise the voltage and/or pack the electrodes as close as possible to each other. Well, raising the voltage past 4.35 V is getting very difficult. Finding electrolytes that can handle such voltages is no small feat. Additionally, the battery is now awfully close to the x = 1 point; in other words, the risk of lithium metal deposition  at the anode is dangerously large at high energy densities. Life is getting tough, and there is very little room left for the battery vendors to maneuver.

These are just a few physical insights behind the challenges that the battery designers and manufacturers are facing. Finding solutions to these challenges via brute-force material development is not the answer. If you find yourself stuck with these limits, talk to us at Qnovo!

13Nov 2014

If you own a mobile device and in need to charge it, the first thing you do is to find a (or your) AC adapter with a USB cable, then plug it into the appropriate USB charging port on your device…et voila! Come back some hours later and your battery should be full.

It is simple, as it should be. But have you ever wondered what happens behind the curtains? I will cover some of these details in this and additional future posts.

So first, before we delve into the electronic circuitry responsible for charging the battery, let us examine the electrical characteristics of the lithium-ion battery during charging. The battery is a complex chemical device, but electrically, it can be simplified into a two-terminal component; in other words, there are two electrical values of importance to us: i) the voltage across the battery terminals, and ii) the current flow, either into (i.e. during charge) or out (i.e. during discharge) of the battery.

The voltage across the terminals of the battery are directly correlated to the state-of-charge (SoC) of the battery — if you recall from this earlier post, it is the fraction of the battery charge relative to full.

During the charging phase, one would expect the voltage to rise across the terminals of the battery from the “empty” level (typically around 3.3 V) up to the “full” level (typically around 4.2 V or 4.35 V depending on the type of battery). This is precisely what the next chart illustrates for a lithium-ion battery with a nominal charge capacity of 720 mAh.

This chart of the battery’s charging characteristics looks rather busy but we can dissect it quite readily to glean some valuable information. Every lithium-ion battery, without exception, will have a similar chart, often included in its data sheet.

The right vertical axis shows the charge capacity (or SoC) as a function of charge time. It is shown in the long-dash curve. Zero is at zero, and 100% is reached after about 2.5 hours of charging.  The charging current itself is represented by the dotted line, and its values are on the far left vertical axis.

One can make a few key observations. First, the charging current has a steady value of approximately 720 mA, then begins to decay after less than one hour of charging. This first phase is called the constant-current (CC) phase; the second phase where the current is steadily decaying is called the constant-voltage (CV) phase. Some publications and blogs incorrectly label them as the “fast charging” phase and the “trickle charging” phase…this is absolute non-sense and illustrates total ignorance on the part of the writer. I will revisit this type of CCCV charging later — it is at the core of many ills that plague modern lithium-ion batteries.

The second observation we make is that the voltage of the battery indeed stays between 3.3 V and 4.2 V, but that somewhere around 50 minutes, the voltage is held steady at 4.2 V and remains there. This is precisely what the constant-voltage phase does; the internal charging circuitry will actually pin the charging voltage to a value of 4.2 V and keep it there until the charging is complete.

This maximum voltage value comes straight from the chemistry. At higher values, the electrolyte inside the battery begins to oxidize and decompose, thus posing a serious safety hazard. This is one of several reasons why an end user cannot, and should not, mix chargers (AC adapters) used for NiMH batteries and lithium-ion batteries. The voltages for each battery type are vastly different.

Finally, one wonders at what point is the charging process deemed complete? Naturally, you will say “100%”, but how is 100% defined? During the charging process, the convention is to halt charging when the decaying current reaches 1/20th of the capacity of the cell. In this particular battery, it corresponds to the current decaying to 720/20 = 36 mA. From the chart above, this is reached after 2.5 hours. But mobile device manufacturers are in a hurry and often fudge their numbers, so that’s why you will see the green light turn on much, much, earlier, shaving 30 or 45 minutes from the actual charging time.

11Nov 2014

A single lithium-ion 18650 cell is relatively small in size and in capacity. So how does Tesla pack 85,000 W.h in the battery pack of the Tesla Model S? The answer is very carefully.

The battery pack in a Tesla S is a very sophisticated assembly of several thousands of small individual 18650 cells connected electrically in a series* and parallel combination. A colleague alerted me recently to an outstanding teardown activity that an owner of a Model S is performing on his battery pack. This offers a great peek into how Tesla designed this battery pack.

First, it is very important to note that 85 kWh is a huge amount of energy. The voltages at the terminals of the pack are high, and the currents can also be dangerously high. In other words, safety is of paramount importance in the design, assembly, and equally the teardown of a large battery pack of this size.

The photographs that this owner has published are very telling and provide a great insight into the design of this pack. The first photograph shows the entire pack with its top cover removed.

Photograph of the battery pack for a Tesla Model S electric vehicle. Courtesy: Tesla Motors club user [wk057].

This photograph shows a total of 16 sections, or modules, electrically connected together. A closer inspection of one individual module shows that it contains a number of 18650 cells, all sitting next to each other in a vertical position. One can diligently count a total of 432 individual cells in one single module.

Therefore, the first observation we can make is that there are a total of 16 x 432 cells = 6,912 cells. The capacity of each cell is 85,000 / 6,912 = 12.29 W.h., or equivalently, 3.4 A.h. The individual cells are of the 18650 type, manufactured by Panasonic. They use an anode made of graphite, and the cathode is made of NCA (nickel-cobalt-aluminum alloy). The NCA-graphite architecture has a lower nominal voltage than the cobalt-oxide alloy commonly used in mobile devices. The nominal voltage of a NCA-based cell is 3.6V.  The owner measured the module voltage to be 19.63V when the battery was virtually dead. A dead (i.e., empty) cell has a voltage near 3.2V.

Photograph of one individual module from the pack.

Therefore, our second observation is that each module contains 19.63V/3.2V = 6 cells in series. Consequently, the module is configured as 72 parallel legs, each containing 6 cells in series (abbreviated as 6s x 72p).

The energy of one single module is 85,000 / 16 = 5,312 W.h. This is equivalent to the energy contained in about 100 (yes, one hundred) laptop PCs. A closer examination of the module assembly shows that each cell is wired to the main bus (the primary electrical path) through little fuses…this is an outstanding safety feature that will disconnect an individual cell that may have shorted with time.

Photograph showing the fuse wires connecting individual 18650 cells to the main bus.

Our third observation is that the entire pack consists of 16 modules connected in series, therefore the overall architecture is 96s x 72p. The stack voltage is nominally 96 x 3.6 = 345 V, but would be as low as 310V when the pack is nearly empty, and 403 V if the battery is at 100% full (but Tesla does not recommend that you charge the battery to 100%).

Our fourth observation is about weight. Panasonic specifies a weight of 46 g for each 18650. The weight of all 6,912 cells comes out to be 318 kg or about 700 lbs. The weight of the entire battery pack is estimated by various sources to be 1,323 lb. So the 18650 account for approximately 53% of the weight of the pack — the rest is due to electronics, cooling systems, wiring and safety.

Judging from my earlier post on cost trends, the estimated cost is about $1.50 for each 18650 cell. I am assuming that the Tesla sourcing team is very influential in demanding attractive pricing from the cell manufacturer, Panasonic. This equates to a cost of approximately $10,000 for the cells used in the pack. Given a delivery rate of about 35,000 cars for 2014, that equates to nearly $350 million that Panasonic will collect this year from selling cells to Tesla…and the number may grow to $1B in 2015.

Adding another estimated $5,000 for the cost of the electronic battery management systems, and one has a preliminary material (BOM) cost of $15,000 for a pack used in the Tesla Model S. That equates to less than $200 for each kWh of stored energy. It also works out to about $50 of battery cost for each mile of driving range. It’s amazing what one can derive from a handful of photographs!

* A series configuration means the positive terminal of the first cell is connected to the negative terminal of the second cell. The voltage at the free terminals is now the sum of the voltages at each cell. The series combination allows raising the voltage of the battery pack to much higher voltages.
† A parallel configuration means electrically connecting the positive terminal of the first cell to the positive terminal of the other, and the negative terminal of the first gets wired to the negative terminal of the other cell. The voltage at the terminal of one cell is identical to the voltage at the terminal of its sister cell. A parallel configuration allows the addition of capacity without raising the voltage.