Matryoshka and EV Batteries

I promised to keep my latest article relatively short and sweet, as the last piece on the SAE’s Autonomous Driving levels turned into a bit of a tome. So, 10 minutes' reading maximum here – let’s get into it!

One of the hottest topics in the EV battery field is solid-state batteries: what are they? When are they coming? What will they cost? But breathe easy – I am not going to go into that, albeit fascinating, topic today. Discussions about batteries tend to focus almost exclusively on breakthroughs or potential breakthroughs in cell electro-chemistry, like the abovementioned solid-state technology. Understandable, but far from the whole story. Maybe 25-35% of the cost of the complete battery pack, and 15-25% of its mass, is comprised of other systems – the mechanical housing or ‘casing’ itself, thermal systems, BMS (Battery Management System), BDUs (Battery Disconnect Units), high-voltage busbars and the mind-boggling rat’s-nest of low-voltage wiring that plagues most batteries. So this article will concentrate on the mechanical design of these batteries, and most particularly on the technology known as ‘Cell to Pack’ – possibly the second-hottest topic in batteries these days.

‘Cell to Pack’ is a very simple and generally very logical idea. Most EV batteries today are based on a ‘boxes within boxes’ design principle. Lithium-Ion cells – be they cylindrical ones, like those used by Tesla and Rivian, or the squarer, flatter prismatic or pouch cells used by most other car-makers, are packed into ‘modules’. A module is basically a self-contained mini battery. Groups of cells are packed mechanically together into a (typically rectangular) module housing. The cells are interconnected – wired together - typically with soldered or welded connections, in the appropriate parallel/series architecture to achieve the design terminal voltage for the module. The module also contains the required sensors to monitor cell voltage, current and (crucial for safety) temperature. The module tends not to have any ‘intelligence’: they don’t have a module ’brain’ – the brain tends to be the centralized battery BMS, that monitors all of the modules of the battery pack centrally…

?Fig 1 - an LG Chem 'pouch' element. Image credit EV Revs

With me so far? It might be clearer if we have a closer look at a very typical module – that used by GM’s Bolt. OK, it’s not the newest of EV designs, but is in fact very similar to the fresh-out-of the-box Ford Mustang Mach-E design. The Bolt uses pouch-type cells supplied by South Korea’s LG Chem Energy Solutions. Each pouch, in the form of a long, slightly flexible aluminium foil envelope with connection tabs at each end, has a nominal terminal voltage of 3.65V and an energy capacity of roughly 0.60 kWh. In the Bolt and in the Mach-E, these are configured together in a 3P or ‘3 in parallel’ layout to give a 3-pouch cell unit, still with a terminal voltage of 3.65 V, but now storing up to 1.80 kWh of energy. LG Chem then pack these units into 8 modules of 10 cells, and 2 slightly smaller modules of 8 cells, giving a total of 96[1] cells in series, with a 350 V nominal pack terminal voltage, and a theoretical 57.4 kWh pack energy or State of Charge (SOC) – which GM sometimes cheekily round up to 60 kWh…

Fig 2. A pretty typical EV battery - that of the GM Bolt, clearly showing the module 'boxes'. Image credit GM.

Looking at the pictures of the pack, you can clearly see that this gives us a ‘box within a box within a box’ type structure. The 3-pouch cells are housed in plastic-and-aluminium racks, which are then housed together in further boxes that look a little like large truck batteries, which are in turn loaded into the very robust aluminium-and-steel battery housing itself. This whole assembly – the battery pack – is then bolted firmly into the underbody of the Bolt. Look at most BEVs out there today – from VW’s IDx series to the Nissan Leaf, Renault ZOE or most of the Chinese OEMs and you’ll see a basically similar approach.?

So – a modern EV battery really is a series of boxes within boxes, or as some folks have observed, a bit like the iconic Russian Matryoshka doll.

I guess some of you might be saying “hang on, those OEM dinosaurs know nothing about designing EVs”. Surely Tesla have a better approach?

Fig 3. Even Teslas use boxes within boxes...in this case with modules built up from cylindrical cells. Image credit 057 Tech.

Not really. Teslas also house their – typically cylindrical – cells in ‘boxes’ or modules, then pack those boxes within the bigger box of the battery housing slung under the floor. And Tesla’s LFP prismatic-cell batteries, used in some Model 3s, look pretty similar to the architecture described above – and hence to 90% of other BEVs on the market today.

So, OK, boxes within boxes, Russian dolls, we get it. So what?

Well, this is where Cell to Pack or CtP ‘technology’ comes in. It’s not really a technology to be honest, it’s more a packaging principle. The idea is to get rid of all these intermediate boxes, and pack the cells directly into the last, biggest box, the battery housing itself. So, no more modules, folks: we populate the cells directly into the battery housing tray itself. Interconnect them (‘wire them up’) in whatever parallels-series configuration that we need to get the desired terminal voltage, and we’re off to the races.

Now this seems a no-brainer. We can get rid of redundant mechanical design, simplify and reduce mass. Simpler systems - all good? And if we look at some of the claims – from CATL, the largest battery supplier in the world and a Cell to Pack pioneer, as well as various independent studies, we can estimate very significant pack-level energy density gains (in kWh/kg) - 10 to 20%, depending on whose numbers you believe.

?It’s tempting to go further. Why do we need a battery housing at all? Can we not get rid of this ultimate, biggest Russian doll of all, and simply use the vehicle’s own structure? After all, the vehicle itself IS, generally, a big metal box? Surely we can pack cells directly into the floor-pan, pillar structures, hell, why not even pack them into the ‘dead’ spaces like inside chassis members, empty extrusions etc? (obviously having designed the structure appropriately to protect the cells in crash…this is important, kids). This approach is sometimes called ‘Cell to Chassis’ or ‘Cell to Structure’ and is a logical and elegant extension of the Cell to Pack idea…and Tesla are pushing in this direction, having pushed the idea during the September 2020 Tesla Battery Day.

?In general, I love this way of thinking. I am all for eliminating redundancy, using one element to perform two or more tasks, and hence reducing mass. Colin Chapman would have loved Cell to Pack, and probably would have pushed hard for Cell to Structure designs. So I should be the world’s biggest fan of CtP or CtS architectures. And in general, I am, and I predict that these will indeed become the dominant approach to designing EV batteries - at least for passenger cars - in the next few years.

?But as always there is no such thing as a free lunch, and those rushing headlong towards Cell to Pack or Cell to Structure layouts should maybe pause to consider two potential trip hazards:

1.??????Modules…are modular.

Intermediate modules lead to some mechanical design inefficiency, it’s true. But they also make it very easy for car-makers to upgrade batteries when cell technology improves, with relatively little re-engineering of the battery pack, vehicle or production process. To illustrate this point, consider the case of the little Renault ZOE, still the 2nd-best selling BEV in Europe[2] after all these years, and many attempts to push it off the podium.

When the car was launched in late 2012, I was ridiculously proud that we had managed to cram a whole 22 kWh of useable energy into the 290 kg battery pack, developed with LG, using an early-generation version of their pouch cells that GM, Ford and others later adopted. This gave the car a then class-leading range of 130km[3] (130 miles).

Picture my embarrassment when, in 2016, my colleagues who had taken over the project when I moved onto other things, managed to cram 41 kWh (again, useable energy) into the very same battery package, for a mass increase of only 15 kg, and – crucially – without having to heavily design the actual vehicle floor-pan, or the battery or vehicle production processes. The range of the vehicle virtually doubled to 400 km (250 miles), making it truly useable for most small family car use-cases.

I was blown away by this remarkable achievement, in just 4 years. And what made it possible was the modular design of the battery. In these few short years, the brilliant LG electro-chemists had cooked up a new cell chemistry, but one that could be housed in the same mechanical module footprint as before – allowing Renault to update the battery chemistry with relatively few changes to the vehicle, system or production process design. An (almost) plug-and-play design change.

Fig 4. The modular ZOE battery design that helped Renault upgrade the vehicle's battery technology without a full design tear-up. Image credit Renault SA.

Now, imagine that we are in a more deeply integrated Cell to Pack or (even trickier) Cell to Structure architecture. ANY changes to cell physical footprint, terminal design, interconnection method, cell thermal behaviour etc etc. is far more likely to require the carmaker to make significant battery pack, or even vehicle layout changes. Simply updating cell chemistry within an existing module becomes a thing of the past.

This may become a significant issue as there are signs that vehicle (and certainly EV platform) production life-times are actually increasing, but improvements to Li-Ion cell chemistry are accelerating. A basic EV platform architecture that lasts 10+ years (as both the Tesla Model S and Renault ZOE already have, albeit with significant improvements to both within that decade) may see three, four or even five generations of Li-Ion chemistry…and modular designs may make it easier to upgrade those platforms.

2.??????A million miles may not be enough…

The second trip hazard may be the growth in shared mobility. It seems inevitable that the privately-owned ‘retail’ car model is set to decline. The ride-hailing and ride-sharing models that are already used by millions or Uber, Lyft or Didi customers every day demonstrate this. The rise of the AD L4 ‘Machines’ in the second half of this decade will only accelerate this tendency. So we will see a new generation of vehicles being designed not to be sold or leased to private individuals, but to be sold (or more probably, leased) to the operators of massive human-driven or machine-driven fleets. Examples are already coming off the drawing boards and into production reality – Rivian’s ‘Amazon’ van, GM/Honda/Cruise’s Origin, ZOOX’s whatever-it’s-called, and more recently Waymo’s new Zeekr vehicle, designed in collaboration with Geely.

Now, the economics of these markets mean that these vehicles – like heavy trucks, or commercial fleet vehicles used by delivery and postal services, will need to last a very long time. Much longer than the typical 160,000 km or 100,000 miles (or 7 to 9 years) that most private passenger cars have to last. Think maybe factor 5x or 10x…

Fig 5. Shared mobility vehicles like ZOOX's - whether they are human-driven or autonomous - will have to last much longer than a typical privately-owned car. Image credit Elektrek.

So what’s that got to do with modular vs. Cell to Pack or Cell to Structure battery designs? Well, we’ve seen various companies announce the long-awaited ‘million mile battery’ – CATL, Toyota and Tesla have all made public predictions that they have chemistries in the pipe-line that will allow batteries to survive many, many years or even decades in service, even in intensive fleet-type use. This is great. BUT batteries do not always ‘die’ simply through homogeneous electro-chemical cell fatigue through repeated charge-discharge cycles. Many failures are more trivial and less predictable, caused by accumulated damage through heat, vibration or corrosion – a failed welded/soldered/crimped electrical connection, a broken cell sensor, a loose screwed fastener to an individual module. In these ‘point’ failures, today’s modular ‘box within a box’ designs have a huge practical advantage. It is relatively easy to ‘drop’ the battery, open the top cover, diagnose the defective module, and replace that module – and that module only. Then reverse the process and put the vehicle back on the road. Let’s be clear – its not that easy a process – we are dealing with high-tension systems here after all, so we need trained personnel and adequate facilities to do this safely, but it’s not that hard, and not that expensive.

Now imagine once again, a more deeply integrated’ CtP or CtS design. It’s gong to be much, much harder to ‘get at’ a defective cell, and even if we can physically access it somehow, the connection is unlikely to be a simple screwed fastener. Replacing individual cells will probably require relatively advanced soldering or welding operations – probably not something we will be able to do safely in even a relatively sophisticated service/aftermarket facility.

In conclusion, these two hurdles do not present strong enough arguments against Cell to Pack or Cell to Structure designs – and as I've suggested above, they are likely to dominate the passenger EV market over the next few years. But I would just sound a little note of caution for the designers of commercial vehicles, or custom-built fleet-mobility vehicles – for those specific applications, it might be worth carefully considering whether the gains of CtP/CtS approaches outweigh the costs in module serviceability. Maybe those Russian dolls have a future, after all…

[1.??????The GM Bolt battery is hence a so-called 3P96S (3 cells in parallel, 96 in series) battery, in the jargon used by battery engineers.

?

2.??????It finally got shoved off the top step by the Tesla 3 last year…after many years up there

Measured at?M