Yesterday, Tesla stock closed at a price per share of $188.18, and while that’s down from the company’s 52-week high of $299.19, it still makes Tesla more valuable than the next four most valuable car companies combined – Toyota, Porsche, Mercedes-Benz and BYD. But it wasn’t always that way. When I started at Tesla as employee number 350 in November, 2009, the company was seriously struggling financially. We were living hand to mouth, which is not unexpected from a small start-up. Still, despite our size, we were ambitious. We had just started the design phase of the Model S and we had high hopes for the car, but a low budget.
It may sound surprising today, but Tesla didn’t have billions in the bank back then, and developing a new car is a very expensive proposition. It takes hundreds of millions of dollars to design, develop, and tool up a new vehicle. Just to give perspective, when Ford developed the new Taurus in 1986, it spent $2 Billion. 25 years later, the costs had only gone up from there.
So at Tesla, we had to get creative to save money wherever we could. You can see the effects of this in the chassis, especially in the steering system and rear suspension.
Here’s a video where I explain all of this:
Steering Racks Are Absurdly Expensive
One of the most expensive parts of a car’s suspension system is the steering gear. Designing and tooling a new steering gear can cost well over a $1,000,000. The housing is a complex casting, and the rack bar (an example of which is shown below) is a very complicated forging requiring some very fancy tooling to create the teeth and the helical groove for the electric power assist system.
Even in the old days of hydraulic steering systems, steering gears were expensive to make. There was no way a small start-up company like Tesla would be able to afford such a large bill. On top of that, Tesla was an unproven company. We did not have a long history of profitability and success yet. For many traditional suppliers, agreeing to design and develop something as complex and expensive as a steering gear would have been a major risk they were unwilling to take.
The way around that is to find a steering gear that already exists which the supplier and the current owner will agree to let you use. If you’re lucky, you could find a steering gear that fits perfectly, and where the supplier has enough additional production capacity to support your needs. Production capacity was unlikely to be a problem for most suppliers since, at the time, we were only expecting to sell about 20,000 Model S’s per year. But finding just the right gear was going to be a difficult task.
You can’t just go up to another car company and ask them for the details on all their steering gears to see which one might work for you. No one’s going to even answer the phone. You can leave a message but don’t expect to hear back from them.
The answer was to work through the suppliers. Fortunately, ZF-Lenksysteme, a well-known German steering supplier, agreed to work with us as long as we were able to find something in their current production catalog that would suit us.
The first step in this process was to determine what we actually needed. We did this by designing a front suspension while pretending that we could get whatever we wanted. We created a front suspension design we liked, and which would fit in our car. Out of that process came two very important numbers – the length of the steering rack, and something called the C-factor.
What We Were Looking For: Rack Length
View this post on Instagram
The length of the rack is the distance between the left and right inner tie rod joints. This dimension is fixed by the length of the rack bar and by the design of the steering gear housing and would require a complete redesign of the steering gear in order to change.
The reason this dimension is so important is that the position of the inner tie rod joint relative to the upper and lower control arm inner bushings establishes how the wheel will steer as the suspension moves up and down (the inner tie rod swings up and down, and so does the control arm; the difference in the arcs at the end of those “swings” has huge handling implications; this phenomenon is called “roll steer.”). If the rack length is too long, i.e., the position of the inner tie rod ends is too far outboard, then the length of the tie rods will be too short relative to the control arms and in the suspension shown above, the wheels will steer inward (toe in) as they move up or down. The opposite will happen if the rack is too short and the inner tie rod ends are too far inboard.
Of course, the vertical and fore/aft position of the inner tie rod are also very important to the proper functioning of the suspension, but this can be easily changed by moving the entire steering gear up/down or fore/aft. There is no way to modify the inboard/outboard position of the inner tie rod joint once the length of the steering gear has been established.
C-Factor
The other dimension that is critical to the function of the steering rack is called the C-factor, and it refers to the distance the rack travels during one full revolution of the steering wheel. This number is usually in the 45-65 mm/rev range, and it determines how fast the steering feels. A higher number means the rack moves more for each turn of the wheel and so the steering will feel quicker than if the C-factor is a lower number.
The C-factor needed for a specific steering gear is determined by an analysis of the overall steering ratio, which is the ratio between the degrees of steering wheel rotation compared to the degrees of steering at the front wheels. For most passenger cars, this number is on the order of 14-17:1, meaning for every 14-17 degrees the steering wheel turns, the front wheels will steer 1 degree.
We Found A Land Rover Rack That Would Work, But It Was Right-Hand Drive. So We Flipped It
Once we had established what steering gear we wanted, it was then a matter of seeing what was in production that might be close. We provided our requirements to ZF-Lenksysteme, and they quickly determined that a rack they were soon to put into production for the Land Rover Evoque was very close. Here’s the Evoque right-hand-drive rack:
And here is a rack from a Model S:
You can see how they are identical except for the ends of the tie rods (these are fairly cheap to tool), which had to be different to connect with our knuckle.
We would have to do some minor modifications to our suspension design to make it fit, but these would be easy to do and wouldn’t compromise the functionality of our design. The biggest problem was that the Land Rover gear was designed to be placed behind the centerline of the axle (called “rear steer,” not to be confused with rear axle steer) while we wanted to place ours ahead of the axle centerline (called “front steer”). The problem is that gears designed to be placed in front or behind the axle move in opposite directions. In a rack designed for front steer, the rack will move to the right when you turn the wheel to the right while in a rack designed for rear steer, the rack will move to the left when you turn the wheel to the right.
Clearly, we couldn’t simply take the Land Rover rack and place it into our suspension design. The car would steer to the left when you turned the wheel to the right!
Our solution was to use the right-hand drive rack from the Land Rover, flip it upside down (so the right side was now on the left and vice versa, and the bottom mounts that normally attach to the subframe are now pointing up), and place it in front of our axle for our left-hand drive application. Flipping the rack had the effect of reversing the direction the rack would travel, meaning it would now move in the correct direction when you turned the wheel and it also had the effect of moving the pinion, where the steering wheel and column attach, from the right side to the left side of the vehicle. This turned the Land Rover right-hand drive rear steer rack into a left-hand drive front steer rack that we could use.
That left only two things for us to do, or so we thought.
The first was to figure out how to attach the rack to our vehicle structure. Flipping the rack over meant that the bolts that mounted the rack would now end up on the top side of the housing and far away from our structure. The solution was to add two towers to the front subframe that reached up high enough to get to the bolts.
The second thing left to do was negotiate a commercial arrangement with Land Rover that would allow them to agree to let us use the steering gear they paid for, made with the tooling they purchased. Fortunately, we had some excellent negotiators on our team, and we came to a mutually beneficial agreement.
Another Problem: The Land Rover’s Steering Rack Was Too ‘Quick’
While the length of the Evoque rack was close enough that we could use it, the C-factor was quite a bit higher than we wanted. We knew we might have an issue, but we thought we could probably live with it. When we built our first prototypes however, we quickly discovered that the steering was much too quick.
The car was darty and difficult to drive on the highway. It wasn’t relaxing at all and would be uncomfortable for our customers. Since changing the C-factor of the rack meant designing and tooling a new rack bar and pinion, we had to find another way of making the steering feel slower. Our solution was to change the roll understeer of the front suspension.
How Clever Engineering ‘Canceled’ Out The Land Rover’s Overly-Quick Steering Rack
Roll understeer (which David describes in the Instagram video above) is a characteristic of a suspension design that forces the wheels to steer in the direction opposite to the steering input when the vehicle body rolls in a turn. In most cars, both the front and rear suspensions are designed to have some nominal amount of roll understeer. Here is how it works.
Let’s say we are making a left turn. We turn the steering wheel to the left which turns the front wheels to the left. As the vehicle enters the turn, the body will roll to the right which moves the right-side suspension slightly upward and the left side slightly downward. As the right-side suspension moves up, the roll understeer forces the right front wheel to steer slightly to the right, effectively subtracting some of the steering we put in at the steering wheel. The suspension actually steers a small amount less than we asked for. It “under” steers a little.
The amount of steering we get as the body rolls is controlled by the suspension design. Let’s take an example front suspension:
When viewed from the front:
you can see the upper and lower control arms as well as the tie rod and, in each case, the arcs each will trace as the suspension moves up and down. In this design, the arcs are all very similar, so the three arms will move relatively parallel to each other. But what if we moved the outer tie rod end down like this?
Now we see that the arc the outer tie rod end follows as the wheel moves up and down is tilted relative to the upper and lower ball joint arcs. This means that as the wheel moves up, the outer tie rod end will move outward (i.e it will “extend”) while the upper and lower ball joints are just moving more or less straight up. Since the steering rack and tire rod are mounted in front of the axle centerline in this design, the effect will be to steer the wheel outward as it moves up and inward as it moves down.
By moving the outer tie rod end down, we have changed the way the wheel steers as it moves up and down. By varying the amount we move the outer tie rod up or down relative to the upper and lower ball joints, we can change the amount of steer we get during wheel movement. This is how we change the amount of roll understeer our suspension has.
And that is exactly what we did on the original Model S. We moved the outer tie rod down in order to increase the amount of roll understeer. Effectively, we used the roll of the body during a turn to subtract from the amount of steering we asked for at the steering wheel which had the effect of slowing the steering down to a level we felt would be acceptable to our customers.
Saving Money By Flipping Control Arms, Smart Manufacturing
Using an existing steering gear wasn’t the only we tried to save cost in the Model S suspension. One of the most expensive parts of a suspension design are the tools needed to make the control arms, especially if those parts are forged. Forging tools can cost many hundreds of thousands of dollars depending on how complex the part is. If we can use fewer tools, then the total bill will be that much lower. An easy way to do this is to use try to use common parts on the left and right side. That way, you need a single tool where you otherwise would have needed two. In many cases this just isn’t possible. A knuckle, for example, is almost impossible to make the same on the left and right side. I’ve seen it done but in my opinion it seriously compromises the design.
But in other cases, it can be done. The lateral link in the Model S front suspension is a good example. This part is the same on both the left and right sides.
The rear suspension is where we were able to really put this philosophy to good use. While the rear knuckles are unique left to right, all the other suspension parts are common. Even the big lower link is the same left and right. The left side is simply the right-side part flipped over.
This link is actually quite complex since it is a hollow casting. Hollow castings use a sand core placed inside a steel mold leaving a small gap all around the edges where the molten aluminum is poured. Once the aluminum solidifies and the part is removed from the mold, the sand is trapped inside the part. It is then removed by shaking it violently so the sand falls apart and is poured out through a large hole left for that purpose. Once all the sand is gone, the hole is plugged or in some cases simply left open.
The knuckles are standard aluminum castings, nothing special there, but the other links were made in a way that also saved us a lot of money. They are extrusions.
This video does a good job of explaining what extrusions are and how they are made:
I mentioned earlier that forging tools are extremely expensive and while tools to make aluminum castings are a lot cheaper, they are still quite expensive. Extrusion tools, by comparison, are extremely cheap. The die to make the small integral link only cost a few thousand dollars while the die to make the upper link was on the order of 3-4 times that much. And since each of these parts is common on the left and right sides, it meant one set of dies for both sides of the suspension. A significant savings.
Using Common Parts Left-Right And Borrowing The Land Rover Rack Was Important
It may seem counter-intuitive to think that a car as expensive as the Tesla Model S would require us to be so concerned about cost, but when you are a start-up company with limited funds, every penny counts. The Model S was and remains an expensive car, so its functionality had to be at a commensurately high level. Achieving that with the suspension meant getting creative and saving cost wherever we could while still delivering the performance. It was a fun challenge.
The 1963 Corvette design borrowed stamped-steel front control arms from the ’58-’64 full-size Chevy passenger car, switched left and right. The tire clearance indents on what would normally be the rear of the lowers face forward on the ‘Vette. (Add’l weld-on brackets such as sway bar attachment were different.) Used on C2-C3 up to 1982. Kinda fun imagining some manager plopping down a box of Impala parts in front of Zora Arkus-Duntov and saying, “Here’s what you get” and Zora thinking “Oh yeah, well then I’ll use them backwards”.
This article is near and dear to me because I was about a year ahead of you in 2007, working for a major tier-1 supplier, but doing my best work on a home project involving sourcing a steering rack. This turned out to be very time consuming for all the reasons you describe, and once decided upon and sourced, then creating the mounts to properly place the rack considering all the pertinent dimensions and interferences considering ackermann, roll steer, etc. Here’s a few pics…
https://flic.kr/p/4D2qxp
https://flic.kr/p/4D6Hcw
To fix the fast steering, couldn’t you just move the location of the outer tie rod end forwards, closer to the front of the car to give it a longer moment arm about the steering axis? This way, more rack movement would give less change in steering angle while reducing steering effort (at least with the same steering assistance calibration). Or did this start interfering with brake caliper mounting or inner wheel barrel?
That would work but then we would lose turning radius. I don’t talk about the track travel but that is something else you can’t change one the track bar and housing are designed.
Ah, ok, so at low speeds when there isn’t appreciable roll, it’s still a fast ratio but as speeds increase and thus cornering forces and roll per degree also increase, it effectively becomes a slower ratio.
As a powerful rwd or at least rear biased awd car…. You can also achieve a tight turning radius in a more… Adventuresome…. Way.
Fun fact: the Datsun 240z uses the same stamped rear lower control arm on both side for the same reason, making the rear suspension asymetric.
Apologies if I missed this, but the Evoque obvs also came in LHD; what was different about that rack that it wouldn’t have worked?
Placement of the rack ahead of the axle centerline in the Model S vs behind it in the Evoque.
Because we needed the rack to operate in reverse of how it was designed, we used the RHD rack and flipped it over. If we had used the LHD Evoque rack in our car, the steering would have been backwards.
Ah got it—thanks!
It certainly worked out well. The S has great steering and suspension.
Porsche did something similiar with reusing control arms on the 996/986 cars. Front/back, left/right all of the lower control arm assemblies are the same parts.
Lotus did something similar with the Evora steering rack.
Lotus is another company that is perpetually short of cash so I’m not surprised they did that.
Great article Huibert.
I was thinking of you the other night when I read the article on crash safety, I believe. EDIT: on Discord there was a link to a vehicle that had been split in two in a crash. It got me thinking about how week getting hit at that point in a vehicle is, as they are not designed for it (much, due to space). In my mind was was running the cross section view of a car for and aft in my brain-cad. There isn’t much structure at that point. AND, I was thinking about the doors….
tldr; How much do doors add to the stiffness of a vehicle, percentage-wise?
I’m an engineer, studied automotive engineering in school. Went down the semiconductor path. So, I’m not a complete fool in this realm. But, doors.
Doors attach at 3 points, 2 at the A-pillar, 1 at the B-pillar. PLUS, the addition of the seal interface. The seal interface makes this a compliant member, only held at 3 points. So, I suppose my real question is how much does the door seal add to the stiffness of the car.
This is obviously not relevant to your article, but I didn’t know of a better way (I suppose I could email you?) to ask you this. Perhaps a topic for another post.
The answer is basically nothing. The doors are rigid at the hinges but not the latch and the seals are by their nature flexible. The rigidity of the body is due to the sheet metal and the glass, not the doors. Having said that, doors are very much a part of the crash structure and will carry a LOT of load in an impact, both frontal and side.
That’s what I thought. Thanks. I know they have are integral to crash safety. It’s one of the first things I look at in those crash video slow-mo’s. Still be best analogy I remember about stiffness is the shoebox with and without the lid. So great.
Thanks, I’m glad I asked.
Doesn’t having a high amount of roll understeer cause the front tires to toe out under heavy braking making it feel very squirrely or pull to the side when you hit a bump?
Yes, that is a danger since the nose dives during braking and would put the wheels into a toe-out condition. There are ways to mitigate this with bushings though. The right way to fix steering that is too fast is to use a lower C-factor so you don’t have to compromise anything else, but when you’re faced with limited cash, you have to make do.
Curious about a detail that was probably cut for length: what’s the tradeoff for using aluminum extrusions for suspension components? Lower tooling cost in exchange for (?)
I hadn’t known tooling costs differed so drastically. It explains why more extrusions are seen in low volume cars. Honda Insight (first gen), GM EV1, Toyota Mirai, etc.
The trade-off is that you are limited in the shapes you can create. It has to be planar, for example. You can’t really make 3D shapes.
Great article. Did it affect longevity? Wondering if a steering rack designed for operating in one orientation is sufficiently lubricated when flipped. Are the seals holding up?
Fortunately Land Rover designed the party upside down.
No effect on longevity. There is no oil reservoir so the gear doesn’t care how it is mounted.
Really cool! I own one of these early Model S’s, a P85 made in Feb, 2013. Every year when I swap tires (summer to winter and winter to summer) I take the opportunity to clean the wheel wells and as much of the suspension components that I can reach. I’ve always admired the manufacturing in there compared to other cars I’ve owned. Cool to see the background of how its made!
This was an awesome insight into how the sausage is made. In the end, engineering is as much art and creativity as it is science, and this illustrates it perfectly.
Only here at the Autopian can we move so easily from the utterly silly to the tremendously profound. I’m totally here for it in both directions.
This is pretty brilliant, I love it when engineers work out clever/simple solutions to problems like this.
I’m not 100% sure, but I think I remember hearing that when the Model S was initially designed and put in to production, it wasn’t initially engineered for AWD capability, just RWD (?). To make the dual motor/AWD cars there was some significant rework to the front structures, etc…
I could be totally wrong though.
You are correct. The suspension was designed with AWD in mind from the beginning, but the body structure had to be heavily modified to allow a driveshaft to fit in the front.
Flipping the rack was my immediate thought, but only because I spent years working out designs for my own cars and tearing apart and inspecting cars to see how they’re built. The roll understeer was a clever solution to the rack ratio problem (though I find it funny that a Land Rover had too quick a rack for a performance sedan). I love articles like this because they demonstrate just how expensive it is to make cars and the often unthought of engineering that goes into systems most of us take for granted. When people complain about the prices of cars, I shake my head thinking about the incredible value per dollar and general greater longevity compared to nearly everything else we buy in spite of their complexity and more varied and abusive operational environment.
What I like about articles like this is you learn more than just the subject written about. I got an entertaining and informative article about Tesla. In addition it explains why new car companies have to start with experience models and hopefully expand to cheaper. It alludes to why an inexpensive car will likely only be able to be developed by an established auto manufacturer.
But one thing is missing. I want to know about how the idea to flip the Rover suspension was arrived at? Did the manufacturer come back and say this will work if flipped? Did the engineers look at it and realize immediately it would work? Or was everyone sitting around and contemplating until one ĝuy or girl said Hey you know what?
Once we knew the rack was available it didn’t take long to come up with this solution. We had some very good, experienced engineers on the team and a little brainstorming brought out the answer.
Thanks that would have bothered me. I have seen teams with same mind miss obvious ideas. Only to be embarrassed by a new member to the team in 30 seconds. It was used I was unpopular.
Hopefully the upper folks that have influence on your career realize it’s not a popularity contest 🙂
Nope i save them $80,000 and I get the same raise they give the potted plant in the corner of the room.
Does anyone know what factors result in the dogshit Model 3&Y turning radius?
Saw a lady fail to turn into what appeared to be easy parking spot, initial thought was she was a bad driver, reality is their turning radius is worse than giant SUVs.
Interestingly, my 2004 Cayenne has the best turning radius of any vehicle I’ve ever driven. Not sure how Porsche pulled that off but I’d be interested to know the mechanics behind it too. Perhaps how far the wheels are able to be turned side-to-side?
That’s interesting, didn’t know Cayenne’s had a good turning radius. LR3’s of the same era also had shockingly good turning radiuseseses.
Many European cars have tight turning radii because roads are so much narrower there than here, especially in cities. You can’t get away with a big turning circle. It’s all about how far the front wheels turn. But that gets into the whole package of the front of the car: crash rail placement, engine package (for ICE cars), tire size, and finally overall vehicle width.
I’ll agree that the 3/Y radius is shit. That was one of the biggest things I had to get used to when I got a 2020 3, coming from my 2017 Lexus IS. U-Turns I had no issues with in the Lexus quickly became issues in the 3 and later my Y.
I now have an S,, but really haven’t had many damming U turns in it yet that have stood out.
If you liked this article, please share it on social media!
My mind remained continuously blown throughout the article.
Which is why, when Huibert first told me about it, I said: We’ve got to get this on the site!
The quality of writing and content here vs. The Old Lighting Site these days is now big enough to drive a Cybertruck through. This article reminds me a lot of one of the first ones you guys published just after the March 32nd launch when Thomas did a deep dive into how factory car audio systems are tuned. Love it.
Fantastic article. Very interesting and extremely well written. Thanks!
never knew the rack width made a difference, saw a guy with a home built truck frame- his rack was off a Fiat and was about 3 feet wide
Great article. And incidentally, my next car just might be a used Model S.
Great article
Fascinating and clever. Thanks for sharing.
Its really interesting to see that a manufacturer was forced to use the same techniques as hot rodders. I spent a fair amount of time helping a team who was developing a bolt on rack/pinion solution for datsun 510s, trying to find a common and easily available rack that had the correct measurements to have good bump steer characteristics.
A youtuber I follow called AaronCake updated an early 70s Mazda Cosmo by flipping a steering gear, for the same reasons you were forced to.
Keep these articles coming. I enjoy reading them.
Hot Rodders are in the same boat as we were. You can’t just fabricate or tool up complex things like steering gears so you have to use what you can find.
Their advantage is buying off the shelf without needing to negotiate a license for production.
It seems that the bigger challenge for hotrodders would be to find documentation of the existing parts in the catalog without (much?) help from the suppliers. Even if you know exactly what you need, how are you going to find who has one like it out there?
Does anyone in the hot rod community know: Is there a master database of steering rack specifications for customizers and hotrodders, or is this just more “the value of an experienced builder” kind of thing?
Mustang II
Electric Boogaloo
NO, it takes a lot of digging. Same with rotor sizes and any other instance where you are trying to source an OE part for a special purpose.
A fair amount of detail can be found by digging in to the listed specifications on various websites. RockAuto often carries a surprising amount of detail, but also even Oreilly and Autozone websites depending on the specific part number. I was working as a counterman at Oreilly at the time, and we had access to a bunch of details, like total length, lock to lock, etc. which I was using to answer questions.
I wonder if you could set an AI loose on that type of question. They do flat BS, but forced to provide a link they also turn up what I’m looking for pretty often.
My pics I linked to above were for a MR2 rack on a 510. This was done around 2007 and we had a good discussion with a few folks looking for a bolt-on option on The 510 Realm, where you part of this too?
On Ratsun for me. I think it was a few years later than ’07. Probably the same guys though. The conversion ended being sold under the JBC moniker.