Before he was CEO of Lucid Motors, Peter Rawlinson was the Chief Engineer of arguably the most important electric car ever: The Tesla Model S. While at the Goodwood Festival of Speed this past summer, my colleague Matt and I had a chance to chat with Rawlinson, and this story he told us about how a seatbelt packaging problem led to the Tesla Model S’s battery pack voltage is absolute gold.
I’ve said it before: If I were allowed to have a favorite car company, right now it would be Lucid. As an engineer obsessed with optimizing (I could have just said “as an engineer”), I find Lucid’s mission of building vehicles that maximize miles per kilowatt hour (i.e. that can travel farther on a smaller, lighter, cheaper battery) to fit in perfectly in with everything I believe in. Lucid is taking the hard route. The company is not relying heavily on styling or horsepower or marketing — it’s putting in the work necessary to downsize drive units, consolidate multiple components into single assemblies, dial in aerodynamics, and ultimately squeeze as many miles as possible out of every last cell in the battery pack.
Rawlinson talked all about his company’s philosophy in our interview, but this article isn’t about that: This article is about the engineering legend’s previous job at Tesla, and details how a packaging challenge related to a seatbelt pull-test drove the design of the original Tesla Model S’s battery pack. It’s fascinating stuff.
I start this bit of the interview by asking Rawlinson to point out some examples of designs he and his team had to create from scratch. It’s been over a dozen years since the debut of the Tesla Model S, so we’ve all become used to a number of EV attributes: The battery packs are flat, structural parts of the vehicle body; the drive modules contain a motor, inverter, and ~10:1 gear reduction; a heat pump offers improved efficiency in the winter, and on and on. These are engineering attributes of EVs that we’re now well acquainted with today, but back when the Tesla Model S was under development 15+ years ago, some of this wasn’t obvious. So I was curious to hear some engineering firsts that Tesla had to just… figure out.
Rawlinson discussed the drive module design (with integrated inverter) and a few other innovations, but the story about the seatbelt and how it drove the architecture (and thus voltage!) of the original Tesla Model S’s battery pack.
“I can give you a little example,” Rawlinson replied to my question about engineer-firsts. “We wanted to have a flat floor in Model S…but in order to make the seat belt pull test work — you know, you can have a pull of an elephant… through the seat belt mounts — if you put that into a flat diaphragm, it’s not going to be strong enough.”
“So we wanted it perfectly flat without a cross member, so the only place we could put a cross member was through the battery pack and actually use the cross member of the battery pack, which is underneath the floor and through-bolt it and take the seat belt loads through this.”
Rawlinson went on to describe how the passenger’s H-point (the location of the hip of the seated driver) was determined based on styling and safety considerations.
“We had a predetermined style and actually have quite a long hood and a very swept-back windshield on Model S, and that drove the H-point of the driver relatively far back because it was a very traditionally style vehicle,” he said. “And that actually meant that we had to move the B puller back further than we’d like. And it made for a very small door aperture and a very compromised rear seating position because it had been styled around the traditional long hood. We also couldn’t put the driver too far forward because in an unbuckled crash situation, you have a neck breaker because the head would rise and hit the windshield and it was a snapped neck.”
“So and then there was an ingress-egress issue for the A pillar — so all these things considered, we did trajectory modeling on the computer for the mannequin in crash and we looked that with ingress-egress ergonomic modeling and determined the H-point. And from that we could determine where the anchor points for the seatbelt would be. And that would have to go right through the battery pack.”
“I remember laying this out and actually I got a printout and I was doing it [by hand] and there was this Damascus moment when I realized, actually, if we divided the rocker length into seven bays, I could get a crossmember bang where I needed the seat belt. And it was one-seventh. And I thought that’s it, we’ve got to have seven battery modules in the rocker length to make this seat belt pull test work.”
The crossmember pointed out above was where the seatbelt would be mounted (see seatbelt below), but because it was so flat, it needed to be stiffened; this would be done by bolting that crossmember to a stiff crossmember in the battery pack below.
“We could architect the battery around those seven bays. So then you look at the hexagon fill and we use 18650 cells. And I’m looking at thermal runaway and how much spacing we can have with it between the cells, to the 10th of a millimeter optimizing it. And then I’m looking at the practicality of how we’re going to do the ultrasonic wire bond of those.”
You can see the seven bays of the battery pack below. Notice the two thicker crossmembers that line up with the two crossmembers in the body (shown above). Those crossmembers are what the seat (and thus seatbelt) fasten to.
“So I start laying out cell blocks within the seven bays. And we find that we can get 900 cells in each of those bays. But we need to split it into two because of the way the architecture was configured. So that meant the module was down to 450 cells.”
“And we’re able to do that, split it 75 in parallel, six in series, and that way then by adding two extra modules in the front, which was just enough space to allow for wheel drive in the future, we had seven modules plus seven, that’s 14, two in the front that’s 16, each with a series count of six, and that led to 96 series count and that determined the voltage that Model S ran at.”
Below you can see the modules that fit into the seven bays above, including the ultrasonic wire bonding connecting them all together:
“The voltage,” Rawlinson concluded, “emanated from the seat belt port anchor structure requirement.”
Let’s Look At The Patent Drawings
To summarize: Because of that seatbelt pull test, Tesla had to put its seat belt crossmember in a certain location, and in order to divide the battery pack (which had a certain length) into roughly equal sections, the battery pack had to be split into seven “bays.” Divide those up the middle and add the two up front just aft of the steering gear (between the front two wheels) and you end up with 16 modules.
With each module having six groups (of parallel cells) in series, you’ve got a total of 96 groups in series. Since groups of parallel cells all have the same voltage as a single cell (roughly 3.7 volts), the battery pack has 96 groups*3.7 volts per group = about 360 volts.
That’s how the Tesla Model S’s voltage was determined. By a seatbelt. Amazing stuff!
Let’s have a look at the patent. Here’s the abstract:
A vehicle seat mounting assembly is provided in which the seat mounts are attached to body cross-members that are, in turn, mechanically coupled to battery pack cross-members contained within a battery pack enclosure mounted under the vehicle.
The seat (and thus seatbelt) is mounted to the body, which the battery pack (which itself contains crossmembers) mounts to via some bolts. You can see the bolt holes in the pack in figure 1. below.
Number 503 in the image above points out all the holes in the battery pack that allow a bolt to go through the pack and into the body crossmembers on which the seats (and thus seatbelts) mount:
Here’s a close look at the crossmembers in question; you can see that the battery pack crossmembers 601D and 601E line up with body crossmembers on which the seat (and thus belt) mount. 
You can see the seat rails, 1105, in the image below, sitting on the body crossmembers (1101 and 1102), which sit just above the battery crossmembers 601D and 601E shown above:
(Note: the intro part of this article mentioned just a single crossmember (the rear one) presumably because the one that took the seatbelt loads had to be set in a certain location, and the second crossmember (the front one) could be adjusted to go anywhere by just changing the seat rail length).
Here’s a nice cross-sectional view showing a bolt (1209) coming up from the bottom, through the battery pack, and threading into a weld-nut in a crossmember. 1213 is what the seat rail mounts to from above:
Here’s another view, but this time showing the seat rail (1105):
Here you can actually see the crossmember inside the battery pack; you’ll notice there’s an upper member (1201) and a lower member (1203):
Why the two members? Well, per the patent:
The use of upper and lower members for each of the battery pack cross-members 601A-601H provides a convenient means for holding the battery modules in place, specifically by capturing the module mounting flanges 801 within region 1407 during battery pack assembly.
It’s just fascinating, behind-the-scenes stuff from one of the world’s most talented engineers: The Tesla Model S’s battery design was hugely determined by a seat belt requirement. Incredible.
My opinion of Tesla has sunk so low I forgot they did and do employ talented engineers in a few select cases. I figured this was going to be something more along the lines of some ridiculous seatbelt motor retractor reporting active time as a proxy for the voltage in the 12V.
Model s was build it with as little comprising as possible. The current one I guess is like that in some aspects but definitely has that closed ecosystem we know what’s best feeling going on now. It is interesting how they have gone back and forth using the battery as a structural member vs easy swap out. The Chinese have really got battery swaps down. It’s too bad they didn’t explore that more then the single one stage thing. I guess the cost was probably too much too justify. But the same can be said about a charger network and I would bet if you assessed Tesla the charger Network is with the most.
About 12 years ago, I bought some of the early LED A19 light bulbs with he help of a government rebate program for about $4 per bulb. The bulbs had a nice aluminum heat sink in the base, and are still faithfully lighting parts of my home today.
Now, after 12 years of Engineering Optimization, LED bulbs cost $5 a piece, and fail consistently within about 2 years thanks to their “optimized” plastic heat sink.
Sometime within that same period, engineers were able to optimize the automatic transmission, presumably by sifting through scrapyards for failed and obsolete old tech, then thinking “hey, that’ll work fine.” This ingenious process of endless optimization brought us the legendary Jatco CVT.
I think the best products, from a consumer standpoint, are the ones they come out with in the first few years of development. At that point, they don’t yet know the failure points, so they can’t design the product to reliably last just past the warranty period.
Conflating optimization with planned obsolescence is just wrong..
“Planned Obsolescence”
That’s actually a much more accurate term, and yes, it should not be misrepresented as “optimization.”
Neat.
Worthy of a patent? Not in my opinion. This is basic problem solving (for the mechanical aspect).
I’d be rich if I had a patent for every time I had to find room for a fastener in my career.
I think this article really illustrates why engineering a single platform that can accept multiple drivetrain options (e.g. ICE, hybrid, EV) is likely never going to result in a profitable product for an automaker. In order to have a product that is relatively easy to manufacture, uses the minimum amount of materials, and is appealing to the customer, a ton of engineering optimization has to take place. In other words, while we might want to have the ability to buy the same car with whatever drivetrain we desire, no automaker is going to be able to give us that and actually remain a viable business.
I mean, lots of compliance EVs (ahem, Stellantis’ Junior, Avenger, e2008, etc.) already exist which simply shove an EV motor under the hood and call it a day, with the batteries more or less where the transmission and fuel tank would be. I bet they’re plenty profitable, but definitely not space or efficiency optimized as a pure EV would be.
Correct the STLA Large, Medium, and Small platforms all are designed to accept ICE EV or HEV powertrains. As you said are they the most optimized? no, but the cost savings of developing new products on 4 (I left out STLA Frame) platforms is significant opposed to starting from scratch.
Even with the new Cherokee debuting in late 25 (will likely be 26) they made the last minute decision to switch to an HEV powertrain over EV as they initially planned and the fallout from that change in decision is minimized so much thanks to the shared platforms. STLA also actively weighing the costs to switch the next Compass rollout from EV – as planned – to HEV, and its a realistic option thanks to these platforms.
If Stellantis can pull this off and actually make an EV that people will want to buy and that they can profitably sell for a reasonable price, I would be really surprised. I hope they do.
It is my understanding that the only automaker to actually make a profit on EVs is Tesla. Possibly also BYD, though likely a very slim profit. Every other manufacturer loses money on them, typically a great deal of money. However, if you have a credible source that says otherwise, please share it.
That’s not the case anymore. Stellantis in Europe is profitable. Hyundai/Kia has been profitable on their EVs for a couple years now. GM is now profitable as of last quarter on theirs, and even Rivian has a positive gross margin on theirs now. It usually takes a generation to learn where to optimize, but Tesla isn’t special when it comes to profitability
Have any of these manufacturers managed to earn enough on these EVs to cover their development costs? That’s the true determinant of long term viability.
That depends on how you define it. Every automaker is still building out the factories so they just haven’t had enough time to pay those things down yet. Hyundai/Kia are net profitable though so for them I’m guessing the answer is yes. GM will need another year or so of sales to cover the costs of the build out.
Factories are expensive but you only have to build them once
BYD and several other Chinese companies make EVs at a profit, especially on exports. Even if they still receive subsidies at this point, BYD’s Chinese sales numbers are far too high to survive at a loss. Additionally, there’s several really small automakers in China making mini EVs (not LSEVs) effectively as a commodity, presumably at a profit. Automakers outside of China struggle to profit on EVs mostly from overcoming one-time development costs combined with low sales volume, and in the US, high range requirements necessitating large batteries using the expensive NMC chemistry.
For example, Ford’s Mach E has only sold about 250k units worldwide in 5 years, while making significant powertrain iterations (new battery packs, new motors) that drive up project cost even further. The F-150 is even worse at 75k units over 3 years on a significantly modified platform (the normal F-150 likely receives only relatively minor platform updates gen to gen). Both aren’t able to command high prices either.
Let’s start with the most ridiculous part of this design discussion:
“Floor must be flat”
As with most decisions, this one is irrational and emotional, leading to designs like this that did not need to happen if one were to remove this stupid requirement.
“Truck must be shaped like a wedge”
“Steering wheel must be shaped like a yoke”
“All functions must be on a touch screen”
etc….
“Innovation at gunpoint” is not genius
And yet, the original Model S is the definition of an engineering masterpiece.
Tis. I still love the looks of the car, and think it’s the best Tesla design. Actually, it’s the only one I genuinely like. The 3, X, and Y are okay, but the S is quite pretty.
Agreed. Saw one the other day and its a very sharp design; everything since has not come close.
Is it though? I would call it rather compromised and certainly rather poorly built with PLENTY of stupid baked into it. Those idiotic door handles alone…
Why is that a ridiculous requirement? It improves aero and doesn’t impede passenger space.
Flat bottom surface of the car, yes, good requirement.
Flat bottom surface of the interior, pointless. It does not matter if there is a little bump on the floor under the seats. Hence my suggestion that it is a dumb requirement.
I believe Porsche actually allows for a few indents into the battery back to increase footroom.
Lucid does as well.
If you are shooting for a brand attribute to differentiate your product in the marketplace, it can be important.
It has been my experience that people notice the flat floor of EVs.
My Corvair has a flat floor and it is lovely. It annoys me when people create huge consoles to take up that roomy footspace.
I agree, this was motivated by the “interior floor must be flat”. Talk about cascading consequences. But its interesting in that it seemingly points out that they did not have the powertrain specs decided or set, but were already allowing the design of the car to dictate those specs back to them. Very interesting.
However, genius at gunpoint is still genius.
genius at gunpoint is genius
innovation at gunpoint is not
Dude has skills, but the dude has better luck. You can’t succeed without both. Sure, he is probably a genius, but I gander that his luck far outweighs his genius.
But, once you are in the position, all of that disappears, as you are now in control, and it does not matter how much of a genius or lucky you are. He is past the genius and luck, is my point, and just utilizing his capital better and maintaining momentum.
So many conflicting feelings. What a sensitive young man you must be.
Tesla isn’t alone is choosing to design around a seemingly arbitrary constraint. The first two that come to mind are how Porsche puts the 911 engine in the exact opposite location of the Corvette…
A 911 is the very definition of working very, very, very hard to continue making a bad decision work. German bloodymindedness at its finest.
How about the Beetle and its variants? Same engine layout, not a whole lot of changes throughout its 65 year run.
The Beetle was fine – they were never intended to be fast. The layout made a LOT of sense for a cheap, slow car that could haul a family around. Not THAT unusual a layout back in the day. But as a basis for a very high performance sportscar? Porsche should have said to Hell with hanging the engine out behind the rear wheels many decades ago. The only reason the Cayman didn’r run circles around the 911 was Porsche intentionally sandbagging it. Or the 944 for that matter. Imagine how much faster a 944 with the 928’s aluminum V8 but without the road-hugging weight of the lard-assed 928 would have been?
“Imagine how much faster a 944 with the 928’s aluminum V8 but without the road-hugging weight of the lard-assed 928 would have been?”
I can do you one better:
https://m.youtube.com/watch?v=9aNXb9uMXA4
Boring. Do better.
An LS in a Porsche IS more interesting. The 944 engine was essentially just a halved version of the 928 engine making this swap a yawn.
The LS is better on every metric over the Porsche engine. It’s cheaper, simpler, more compact, more powerful, more fuel efficient, more reliable and makes for a better balanced car.
But if its more “interesting” you’re looking for I’m sure someone somewhere has shoved a TDI into a 944.
It’s boring. Everyone sticks LSs into everything because they are *cheap* (in all senses of the word) and ubiquitous. And at the end of the day imost of them are boring pushrod lumps that really don’t make THAT much power stock considering it’s ridiculously large displacement. And they sound terrible, as all ordinary American V8s do.
Porsche’s V8 even today makes rather more power per liter, and certainly rather more power period than any LS you are getting for cheap out of a junkyard. Put one of those in a 944 and I might be vaguely interested. And it won’t sound like a Camaro. If I wanted a Camaro, I would just buy a Camaro.
Well you do you. To me an engine swap is about making the car better, not worse.
“And it won’t sound like a Corvette. If I wanted a Corvette, I would just buy a Corvette”
FIFY. An LS powered 944 is a lot closer to a C4/5 Corvette than any Camero.
I’m not into Frankenstein engine swaps to start with. If the factory didn’t do it, you aren’t going to do it particularly well either, 99.9% of the time. Just buy a car with the right engine in it to start with.
Well you do you.
TONS of changes
Nothing like the changes to the Corolla or Civic though.
Civic and Corolla are just nameplates that have been slapped on wildly different cars for decades. Largely meaningless at this point. the Beetle (Really Volkswagen Type 1) was a direct linear evolution from beginning to end of the same car.
Which is exactly the point.
That said the “Beetle” did change into a Golf
The only Beetle was Golf-based. Since the original was never officially named “Beetle”. 🙂
-Captain Pedantic
Sorry Captain, VW says otherwise:
Beetle
Not long after it first appeared on German roads, the Volkswagen Type 1 was dubbed the Käfer – or Beetle. That name survived the translation into English, and dozens of other languages, and became the model’s official moniker by the late 1940s. (Other nicknames: “the bubble” in Denmark, “coccinelle,” or ladybug, in France, and “turtle car” in Thailand)
https://www.vw.com/en/newsroom/lifestyle-and-heritage/vw-names.html
Revisionist history. In period, VW never, ever officially called the Type I any thing other than the Type I, despite it’s nicknames.
I guess all the original ads from the ’70s calling it a “Beetle” are just revisionist too:
https://www.atticpaper.com/prodimages/life1972/vw_72price.jpg
https://i0.wp.com/www.curbsideclassic.com/wp-content/uploads/2012/03/1974-fuel-injection.jpg
Funny those ads never once mention Type I.
It’s *advertising*. The official title of the “Beetle” was the Volkswagen Type I. Full stop, do not pass Go, do not collect $200. Just as the official title of the car that nominally replaced it was the Volkswagen Golf/Rabbit.
No shit. ALL product names are marketing.
Most cars don’t have unofficial nicknames. But the VW Type I does – it’s “Beetle”.
Another example – everyone calls that sexy front-engined Ferrari Couple from the late ’60s early ’70s the Daytona – but that was never that car’s official designation. Just a widely used nickname. It’s a Ferrari 365GTB, even if Ferrari themselves occasionally call them Daytonas today.
So here’s the difference: Ferrari themselves recognize the name as an unofficial name:
“Ferrari 365 GTB4
…
Many fans know it by its unofficial name of “Daytona”.”
“https://www.ferrari.com/en-EN/auto/365-gtb4
Whereas VW recognises the name “Beetle” as an official moniker as per VWs own website and used “Beetle” as the name it was sold as during its production.
So yes, the Type I was indeed officially named “Beetle” by VW. Whether that name was officially granted posthumously or not it was granted.
– Major Pedantic
And you are simply wrong. VW calling it the Beetle is revisionist history, it never was while it was in production.
-President Pedantic
Even revisionist nullifies “never”
– Emperor Pedantic
You will never win this, so you might as well stop playing.
Intergalactic Supreme Being God-King Pedantic
I won a long time ago. You just weren’t paying attention.
– me
You’ve been losing for years.
Nice to have a fan.
I do luv ya buddy! You complete me. I wish I knew how to quit you. <hugs>
Every automotive project I’ve worked on has arbitrary constraints.
Often contradictory.
My favourite one was an engine that had to have best in class power output, lowest in class parts cost, and fit down the exact same block machining line with no tooling changes. You present the “best” compromise at the approval gateway and find out what the client really wants.
In the case of the Tesla flat floor/styling/seat belt load issue if the resultant battery voltage had been unacceptable they would have changed the floor. Or the styling (but they never change the styling).
Do automakers, who design vehicles outside of America, make the same kind of compromises for safety of individuals who don’t wear a seatbelt?
There’s a lot of parallels to the Darwinian arguments I’ve heard against wearing a seatbelt.
Not to nearly the same extent, as a general rule. For example, in the US airbags are designed to restrain an unbelted adult, while in the rest of the world they are generally much less powerful and only intended to “supplement” a seat belt. Though with today’s multi-stage airbags it’s sort of moot. If you have your seatbelt on, you get less bang in the face than if you don’t.
Personally, I think they should be disabled if you aren’t buckled up, and a spike should deploy from the wheel and dash. The gene pool could really use some chlorine these days.
The art of engineering is coming up with the optimal set of design compromises.
As someone who’s on the nth complete redesign of a product over the past few years, 100% agreeing.
Don’t fret. One day you’ll get to the oth.
I laughed way too much at this one lol!
So true. I remember being taught this point when I was in engineering school. All engineering is based on limitations. If it where not the floor it would have been another hard point of the design. I am sure the flat floor design had become a hard point for engineering before the belt issue arose.
I know there’s always packaging requirements that limit things, but as a 20 year veteran of the diesel world, 17 of which I pulled wrenches, we used to always joke how “They have this huge truck, and the engineers jammed everything in this one corner.”
One of my mentors worked for Audi, and had to replace the fuel tank in an R8. He said “I’m convinced they just put the tank on a stand and built the car around it.”
So kudos to the behind-the-scenes look at what engineers have to deal with, but I don’t think that’ll make techs hate you any less when they can’t reach that bolt or have to disassemble everything to access one part.
Ha, one wonders what the R8’s heater core would be like to replace: https://www.theautopian.com/why-swapping-a-heater-core-is-one-of-the-most-miserable-car-repairs/
That same joke is often applied to changing the air filter on a Goldwing. “Honda started with the airbox, and built the rest of the bike around it”. If you look up pictures of “goldwing air filter change” the bike looks like it was hit by a missile when sufficiently disassembled to access the filter. If you’ve done it a few times and have the procedure down, you can do it in maybe 2 hours, but it’s probably double that for a first-timer, and having detailed instructions or a video is pretty much mandatory. There’s loads of details about order things have to come off to avoid damage, hidden fasteners, ect.
The trade-off is since it’s a boxer engine, the valve covers are just….there….so valve checks are a breeze (where on a lot of bikes, accessing the valves is a colossal task and often requires partially dropping the engine).
The fun of designed for ease of manufacturing first and foremost, with about zero thought to fixing the thing later.